Remote sensing of the environment with smallunmanned aircraft systems (UASs), part 2:scientific and commercial applications1

KenWhitehead, Chris H. Hugenholtz, Stephen Myshak, Owen Brown,Adam LeClair, Aaron Tamminga, Thomas E. Barchyn, Brian Moormanand Brett Eaton

Abstract: Small unmanned aircraft systems (UASs) are often suited to applications wherethe cost, resolution, and (or) operational inflexibility of conventional remote sensing plat-forms is limiting. Remote sensing with small UASs is still relatively new, and there is limitedunderstanding of how the data are acquired and used for scientific purposes and decisionmaking. This paper provides practical guidance about the opportunities and limitations ofsmall UAS-based remote sensing by highlighting a small sample of scientific and commer-cial case studies. Case studies span four themes: (i) mapping, which includes case studiesto measure aggregate stockpile volumes and map river habitat; (ii) feature detection, whichincludes case studies on grassland image classification and detection of agricultural cropinfection; (iii) wildlife and animal enumeration, with case studies describing the detectionof fish concentrations during a major salmon spawning event, and cattle enumeration ata concentrated animal feeding operation; (iv) landscape dynamics with a case study of arcticglacier change. Collectively, these case studies only represent a fraction of possible remotesensing applications using small UASs, but they provide insight into potential challengesand outcomes, and help clarify the opportunities and limitations that UAS technology offersfor remote sensing of the environment.

Key words: UAS, remote sensing of the environment, case studies, unmanned aerial vehicles,remotely piloted aircraft, remote sensing.

Résumé : Les petits systèmes d’aéronef sans pilote (UAS) sont principalement utilisés pourdes applications dont les coûts, la résolution et (ou) l’inflexibilité opérationnelle des plate-formes de télédétection conventionnelles sont limités. La télédétection à l’aide de petitsUAS est encore relativement récente et on connaît mal la façon dont les données sontrecueillies puis utilisées à des fins scientifiques et décisionnelles. Le présent exposé fournitdes conseils pratiques sur les possibilités et les limites de la télédétection à partir de petitsUAS en se servant d’un petit échantillon d’études scientifiques et commerciales. Quatrethèmes sont examinés dans ces différentes études de cas : (i) la cartographie, afin de mesurerles volumes d’aire de stockage des agrégats et établir une représentation graphique de l’ha-bitat d’un cours d’eau; (ii) la détection des caractéristiques, afin d’effectuer la classificationdes images de dunes et la détection de l’infection de terres agricoles; (iii) le dénombrementde la faune et du bétail, afin de décrire la détection des concentrations de poissons lors de

Received 27 May 2014. Accepted 2 July 2014.

K. Whitehead, C.H. Hugenholtz, A. LeClair, T.E. Barchyn, and B. Moorman. Department of Geography, University ofCalgary, 2500 University Drive NW, Calgary, AB T2N 1N4, Canada.S. Myshak and O. Brown. Isis Geomatics, 3582 30 St. N., Lethbridge, AB T1H 6Z4, Canada.A. Tamminga and B. Eaton. Department of Geography, University of British Columbia, 1984 West Mall, Vancouver, BC V6T1Z2, Canada.Corresponding author: K. Whitehead (e-mail: kwhitehe@ucalgary.ca).1This paper is part two of two companion papers published in this issue of J. Unmanned Veh. Syst. (Whitehead and Hugenholtz.J. Unmanned Veh. Syst. This issue. 2014. doi:10.1139/juvs-2014-0006).

ARTICLE

86

J. Unmanned Veh. Syst. 2: 86–102 (2014) dx.doi.org/10.1139/juvs-2014-0007 Published at www.nrcresearchpress.com/juvs on 30 July 2014.

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

grand frai du saumon et de dénombrer les animaux dans une exploitation intensive d’en-graissement du bétail; (iv) la dynamique des paysages, afin d’examiner les changementsdes glaciers de l’Arctique. Dans leur ensemble, ces études de cas ne représentent qu’unefraction des applications de télédétection possibles, utilisant des petits UAS; toutefois, ellesdonnent un aperçu des défis et résultats éventuels et aident à clarifier les possibilités et leslimites que la technologie des UAS offre pour la télédétection de l’environnement.

Mots-clés : UAS, télédétection de l’environnement, études de cas, véhicules aériens sans pilote,aéronef téléguidé, télédétection.

1. Introduction

Remote sensing from satellites and from manned aircraft is a valuable tool that can provide sum-mary measurements of land cover, surface conditions, and related changes at a variety of spatial andtemporal scales. Remote sensing is continuously and rapidly evolving with the advent of new technol-ogy. In the past, most remote sensing relied upon medium-resolution imagery from sensors, such asLandsat ETM and SPOT (ground resolutions ≈5–30 m). The latest generation of optical satellites (e.g.,GeoEye, WorldView) have ground resolutions in the range of 0.5–1 m, which is starting to rival tradi-tional aerial photography (resolutions typically 0.1–0.5 m). This “resolution creep” is also blurring thedistinction between photogrammetry and remote sensing. Satellite imagery is now often used in tra-ditional photogrammetric roles, such as the generation of digital elevation models (DEMs) andorthoimages.

Further evidence of the integration of photogrammetry and remote sensing can also be seen intechnologies such as airborne LiDAR (e.g., Glennie et al. 2013) and synthetic aperture radar (SAR;e.g., Crosetto and Aragues 2000). These are active systems that allow the production of DEMs throughcompletely different workflows than those used in traditional photogrammetry. Frequently, spectralinformation for analysis is being extracted from aerial photography (e.g., Hunt et al. 2010; Knoth et al.2013). Similarly, the development and refinement of structure frommotion packages, such as Bundler(e.g., Fonstad et al. 2013) and Photoscan (e.g., Turner et al. 2012) now make it possible to produce accu-rate digital surface models (DSMs), digital terrain models (DTM) and orthoimage mosaics quickly andaffordably.

The resolutions and capabilities of different sensors have seen major improvements over the lastfew years; however, until recently, remote sensing surveys were largely carried out from satellitesor manned aircraft. This is changing with the proliferation of small (<25 kg) unmanned aerial systems(UASs). Such systems typically consist of a lightweight unmanned aerial vehicle (UAV), imaging pay-load, controller, navigational computer, and ground-based pilot (and spotters, if required). Distinctionof the “small” class of UASs came about through the U.S. Federal Aviation Administration’s Moderni-sation and Reform Act of 2012 (Hugenholtz et al. 2012). The use of UASs smaller than 15 kg is expectedto expand rapidly for civil, commercial, and scientific uses, because they are less expensive and moreversatile than larger UASs. They are also capable of acquiring high-resolution imagery safely frommuch lower altitudes than those associated with manned aircraft and larger UASs.

There has been a great deal of interest and speculation about the potential applications of smallUASs in the scientific community (Hugenholtz et al. 2012); however, much of the literature publishedto date only addresses potential applications, rather than analysing completed case studies and pre-sently viable uses. Other published studies relate to experimental applications of new sensors, suchas UAS-mounted LiDAR, (e.g., Lin et al. 2011), which are of interest for the future, but are insufficientlymature at present to allow routine operation. To address this, we present seven case studies derivedfrom our collective experience. These studies represent applications for which the current generationof small UASs are well suited, and that can be carried out on a routine basis without the need for cus-tomisation. These applications can therefore be considered representative of the current range ofapplications for which small UASs are being used. Although the examples presented here are allfrom Canada, similar applications are possible anywhere in the world (subject to local regulations).The case studies focus on four broad themes:

1. Mapping: Case studies include the determination of aggregate stockpile volumes and mapping ofriver habitat.

2. Feature detection: This section includes case studies on image classification of pocket gophermounds and the detection of agricultural crop infection.

3. Wildlife: Examples include a survey of a major salmon spawning event and the use of thermalinfrared imagery for cattle enumeration.

Whitehead et al. 87

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

4. Landscape dynamics: This section features a case study in which imagery from a small UAS wasused to study glacier dynamics.

While these case studies illustrate only a small number of the remote sensing applications that arepossible using small UASs, they serve to illustrate both the advantages and the challenges of workingwith this type of data. Successful integration of data gathered by small UASs will ultimately depend onwhether the advantages of such data (e.g., high spatial resolution, affordability, flexibility of acquisi-tion) outweigh the disadvantages (e.g., variable illumination, inconsistent geometry, low qualityimagery).

This review is the second part of a two-part paper. Part 1 (Whitehead and Hugenholtz 2014)addresses the technology and present issues associated with spectral and topographic data collectedfrom UASs. For more basic primers on the use of UASs and complementary perspectives, we refer read-ers to reviews by Hardin and Jensen (2011) or Watts et al. (2012).

2. Regulations governing the use of small UASs in Canada

As described in the companion paper (Whitehead and Hugenholtz 2014), regulations are in place inmany countries governing non-military uses of small UASs. These regulations influence the remotesensing data collected by UASs. Because this review features case studies performed in Canada, inthis section we outline the regulatory framework and some of the rules that distinguish remote sen-sing data acquired by small UASs to provide necessary context.

In Canada, the operation of UASs falls under the Canadian Aviation Regulations, which are admi-nistered by Transport Canada — the federal agency overseeing Canada’s transportation systems. Tocarry out non-recreational remote sensing surveys with small UASs, an individual must hold insuranceand a Special Fight Operations Certificate (SFOC; Transport Canada 2008). This permits the operationof a UAS, subject to several conditions, some of which are described later. Although Transport Canadamay issue blanket SFOCs for organisations to operate in certain parts of the country, new applicationsare always dealt with on an individual basis.

The paperwork and time involved in completing SFOC applications can slow entry for manyresearchers and professionals wishing to become involved in UAS operations. To meet the demandfrom such users, Transport Canada makes provision for a simplified application process (TransportCanada 2008). This process streamlines the application and is designed to meet the needs of the major-ity of such users. However users who require greater operational flexibility, such as police and searchand rescue, may need to apply through the standard SFOC process, as the conditions imposed by thestreamlined application process may be too restrictive to meet their needs. The following discussionrefers to SFOCs issued under the simplified application process, unless otherwise stated.

One important restriction that impacts remote sensing with small UASs is the regulatory require-ment that they are operated within visual range at all times, even though the telemetry between theUAS and ground control system may reach well beyond this limit. In practice, this means that UASsmust remain relatively close to the takeoff point, limiting the area that can be surveyed in a singleflight. The maximum distance will vary according to the size, shape, and colour of the UAS, as wellas the weather conditions and lighting. Although Transport Canada may specify otherwise, formany UASs the maximum visual range is <1 km, which yields a maximum surveyable area of approxi-mately 3 km2 from one location.

A secondary restriction affecting remote sensing surveys is that the flying height of UASs is typi-cally limited by Transport Canada to a ceiling of 400 feet (122 m) above ground level (Transport Canada2008). This allows for very high ground resolutions in imagery, but makes it necessary to collect manymore images than would otherwise be required, which can in turn introduce problems at the proces-sing stage. At low flying heights, the effects of relief displacement and tree lean are often greatly exag-gerated, making it extremely difficult to produce good orthoimages for some areas. In areas of highrelief, the 400 feet limitation may not be practical, and may thus rule out the use of a UAS. Undersuch circumstances, Transport Canada may allow a higher flying height; however the additionaltime involved in obtaining the certification may impact the viability of a project.

The simplified SFOC application process currently limits UAS operation in Canada to daylight useonly. For photogrammetric and most remote sensing applications this is not a problem. However thedaytime-only requirement can be a limitation for thermal surveying. This particularly affects applica-tions, such as search and rescue operations, and heat loss and energy efficiency audits (e.g., Martinez-De Dios and Ollero 2006) as night enhances thermal contrasts. With the development of reliable senseand avoid systems, night flights may become more commonplace in the future; however, at presentnighttime operation is only permitted when a full SFOC application has been lodged.

88 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

Transport Canada also imposes restrictions on the size of UAS platforms that can be flown. In gen-eral, only aircraft with a take-off weight of less than 35 kg are permitted (Transport Canada 2008).While aircraft below this weight have the advantages of portability and safety, this restriction effec-tively limits the duration of flights, as well as impacting aircraft stability. The most serious impact,however, is on the size of remote sensing payloads that can be carried. Payload limitations restrictmost UASs to carrying only lightweight compact cameras. Larger sensors used on piloted aircraft,such as metric cameras, hyperspectral scanners, SAR, and light detection and ranging (LiDAR) sys-tems, are generally too heavy at present to be carried on most UAS platforms meeting this weightrestriction, and although some proof-of-concept research has mounted some of these sensors tolightweight UASs (e.g., Turner et al. 2012; Wallace et al. 2012), the flying time is generally too shortfor practical applications. The development of miniaturised “micro-electromechanical systems” mayusher in a new generation of smaller, lighter sensors in the next few years, but at present the takeoffweight restriction imposes practical limitations on the type of remote sensing data that can beacquired from small UASs.

Because of safety considerations, there are also restrictions on where small UAS surveys can be car-ried out. Currently in Canada, no UAS overflights are allowed in urban areas, or within 100 feet of peo-ple or inhabited structures unless consented (Transport Canada 2008). The high spatial resolution thatcan be obtained from small UAS surveys would be very useful in many urban settings; however, forpublic safety and perhaps privacy concerns, this restriction is likely to remain in place for researchand commercial applications. Further restrictions on location apply in restricted airspace, and closeto airports.

A recent report by the Canadian UAV Systems Program Design Working Group proposes majorchanges for regulation of lightweight UASs under 25 kg in weight (Transport Canada 2012). In particu-lar, this report proposes that UASs classified as “low energy” should be exempted from SFOC regula-tions. The proposed definition of low energy is that the kinetic energy imparted by the platform toa stationary person or object in the event of a crash is less than 12 J/cm2, which is not considered tobe a dangerous impact. The implications of this change are that many small UASs could potentiallybe operated without restriction, particularly those optimised for surveillance. This recommendationhas been criticised for ignoring privacy concerns, and has prompted calls for a wider public debateon the implications of regulatory change (Gersher 2014).

3. Case studies

In this section we outline seven case studies. While not exhaustive, these applications represent across section of the types of applications for which small UASs are well suited. Many proposed applica-tions of small UASs are experimental, or require expensive customisation of the UAS platform; how-ever, the case studies described here were all carried out using commercially available systems,equipped with standard “off the shelf” components and cameras. As such, they can be consideredto be representative of the current state of the industry. In all cases, photography or video is the pri-mary data source. While there are a number of more advanced sensors that have been developed foruse with small UASs, these are still largely experimental. The primary use of small UASs to date hasbeen for the production of DEMs and orthoimage mosaics. This represents an extension of the tradi-tional photogrammetric mapping process. It is expected that this emphasis will change as the indus-try matures and new sensors and analysis techniques become available. However the case studiespresented reflect our intention to portray the current state of the industry.

For each survey we used one of four battery-operated UASs (Fig. 1), which included one rotary-wingplatform (Aeryon Scout Pro quadcopter) and three fixed-wing UASs (RQ-84Z Aereohawk, OutlanderUAS, Ebee). The rotary-wing platform has a maximum flying time of about 20 min or less, dependingon wind, while the fixed wing UASs range from 40 (Ebee) to 90 min (Areohawk). The cameras used toacquire the imagery included consumer-grade RGB, near-infrared (NIR) filters, and a forward-lookinginfrared thermal. The purchase price for these UASs ranges from CAN$7000 for the Outlander UAS toCAN$107 500 for the Aeryon Scout Pro.

3.1. Mapping

3.1.1. Case study 1: aggregate (sand and gravel) volumetricsOne of the leading commercial applications of UAS-based remote sensing is measuring aggregate

stockpile volumes (e.g., Eisenbeiss 2009; Haala et al. 2011; Sauerbier et al. 2011). In Canada, aggregatecompanies have been early adopters of UAS surveys for several reasons. Aggregate quarries are usuallyself-contained, and encompass areas that are within or close to the line-of-sight limit. The areainvolved can easily be covered by a small UAS, although larger quarries may require two or three

Whitehead et al. 89

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

flights for complete coverage. Quarries are also usually free of vegetation, making the production ofaccurate DEMs comparatively straightforward.

For this case study, two UAS surveys were carried out of a small stockpile in June and again inNovember 2012, using the Aeryon Scout Pro quadcopter. This UAS uses a proprietary camera (Photo3S),which has a focal length of 7.5 mm and a field of view of 37° by 29°. For both surveys, 62 images werecollected, spread over 6 flight lines, at a flying height of 100 m. In both cases, the aircraft flew at aspeed of 5 m s–1 and the survey was completed in approximately 20 min. The same flight plan wasused for both surveys to ensure that the imaging geometry remained as consistent as possible. Foreach survey, 10 ground control points (GCPs) were surveyed around the stockpile using a real-timekinematic global navigation satellite system (GNSS). On each occasion the stockpile was also surveyedusing the GNSS, to assess the vertical accuracy of the UAS-derived DEMs relative to the conventionalfield-based survey approach.

Processing for both jobs was carried out using EnsoMOSAIC software. Initial estimates of imagecentres and camera orientations were obtained from the aircraft log file. After aerial triangulationwas completed, a 3.5 cm resolution DEM was constructed for each survey. Both DEMs were edited toremove the effects of vegetation and were then used to produce final orthoimage mosaics. TheDEMs and orthoimage mosaics generated are shown in Fig. 2. Volumetric estimates were producedin each case using ArcGIS.

For the June survey, the root-mean-square error (RMSE) of vertical difference between the UAS-derived DEM and 126 GNSS points surveyed on the stockpile was 0.106 m, with the UAS-derivedvolume estimate being 2.6% lower than that estimated from the GNSS survey of the stockpile. Forthe November survey the RMSE of the difference between the UAS-derived DEM and 107 surveyedGNSS points was 0.097 m, with the UAS and GNSS-derived volumes differing by 3.9%. The volumeremoved from the stockpile between the two surveys was estimated at 1521 m3, which was 2.5%greater than that estimated from the haul weight. Conversion of haul weight to volume necessitates

Fig. 1. Four small UASs were used in the seven case studies presented herein: (a) RQ-84Z Aerohawk (Hawkeye UAVLtd., New Zealand); (b) Outlander UAS (CropCam Inc., Canada); (c) Aeryon Scout Pro quadcopter (Aeryon Labs Inc.,Canada), and (d) Ebee (senseFly Inc., Switzerland).

90 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

the application of a conversion factor, so haul weight derived estimates are unlikely to agree exactlywith the UAS data. Overall, this case study demonstrates that photogrammetric surveying using asmall UAS can yield measurements that are comparable to conventional methods of measuring stock-pile volume and volume changes.

3.1.2. Case study 2: river habitat mappingOne of the key advantages of remote sensing with small UASs is that the data can bridge an impor-

tant gap between ground-based surveys and remote sensing data from satellites or piloted aircraft. Forexample, ground-based techniques are very effective for developing high-resolution maps over smallareas, but impractical for large areas due to the time and physical effort required. Conversely, remotesensing data from satellites and piloted aircraft can be used to map large areas, but they either do nothave adequate spatial resolution or are too expensive to map fine-scale features within small to inter-mediate areas on a routine basis. Small UASs are particularly well-suited to mapping at an intermedi-ate spatial scale (i.e., 1–10 km2).

The case study for this application was carried out along a 1 km segment of the Elbow River,approximately 21 km west of Calgary, Alberta. The UAS survey was conducted during low-flowconditions (5.9 m3s−1) in late September 2012. The purpose of the UAS survey was to evaluate the abil-ity of the imagery to measure key metrics of reach-scale river morphology. At the field site (50.988°N,114.509°W), the river flows through a ∼200 m wide active channel belt with extensive gravel bars. In anumber of places the river splits into separate channels. The channel belt consists of mixed sediment(silt – boulders) and was covered in places by large wood, which included individual logs, trees, root-wads, and large jams. Large wood has an important influence on the morphological and biologicalquality of aquatic ecosystems and as such, is a key component of fish habitat.

The Aeryon Scout Pro quadcopter was used for the survey with its proprietary Photo3S RGB cam-era. A total of 45 highly visible GCPs were distributed throughout the study area to assist with thephotogrammetric processing. In total, the survey yielded 192 images, which were used to develop aDEM and orthoimage. One of the unique aspects of this survey was that the water in the active chan-nels was relatively clear, which allowed extraction of the bathymetry. We then applied a two-dimen-sional hydrodynamic model (River2D) to estimate spatial variability in the flow (see details inTamminga et al. 2014). We mapped the median grain size of surface sediment (D50) using an empiricalrelation between image texture and field measurements of D50 (R

2 = 0.82).Results from the UAS survey are shown in Fig. 3. The 5 cm spatial resolution orthoimage mosaic

(Fig. 3a) and DEM (Fig. 3b) allowed for the detailed characterisation of channel morphology and aqua-tic habitat features (Fig. 3c). Based on 297 check points, the vertical RMSE of the DEM was calculated tobe 0.088 m in dry, exposed areas, and 0.119 m in submerged areas, following a correction procedure forthe refractive effects of the water–air interface (Westaway et al. 2000). This topographic informationwas sufficient to apply the River2D model, which allowed for determination of depths and velocitiesthroughout the study reach (Fig. 3d, 3e). When superimposed on the base orthoimage, these depth andvelocity distributions provide an intuitive method with which to assess reach-scale flow patterns andtheir relationship to geomorphic context.

Fig. 2. UAS orthomosaics and DEMs used for determining changes in stockpile volume following excavation (casestudy 1). Orthomosaic for (a) the June survey and (b) the corresponding DEM and (c) orthomosaic for the Novembersurvey and (d) the corresponding DEM. The outline of the stockpile prior to excavation is denoted by the dashedline in each image.

Whitehead et al. 91

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

Fig.3.

River

morpholog

yan

daq

uatic

hab

itat

analyses

oftheElbow

River,20

12(casestudy2),sh

owing(a)theorthoimag

e,(b)theDEM,(c)D50va

lues

anddigitised

cove

rfeatures,

(d)mod

eled

dep

th,(e)mod

eled

velocity,an

d(f)co

mpositesu

itab

ilityindex

values

forad

ult

browntrou

thab

itat.

92 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

To evaluate the utility of the UAS survey data for aquatic habitat assessment, we then combinedthe modeled hydraulic data with features that provide visual overhead cover for fish species, includinglarge wood, overhanging vegetation, undercut banks, water surface turbulence, and pools. Such fea-tures are strong determinants of fish habitat availability and were digitised directly from the ortho-image. Using preference curves developed for the nearby, ecologically similar Kananaskis River thatdescribe the theoretical suitability of depths, velocities, and cover accessibility for adult brown trout,we were able to map out spatial patterns of habitat suitability (Fig. 3f) in terms of a composite suitabil-ity index that ranges from 0 (least suitable) to 1 (most suitable). Although not validated directly, thismethod highlights distributions of potentially desirable habitat hotspots and can be used to estimatethe total habitat availability for the reach. It also demonstrates how the perspective offered by UASdata can show the spatial complementarity of habitat features: deep low-velocity pools with abundantoverhead cover sources were identified as highly suitable areas and easily mapped throughout thereach. Overall, the combination of the high-resolution orthoimage and detailed topographic data pro-vided by small UAS-based remote sensing provides an ideal method for intermediate scale analyses ofriver morphology and aquatic habitat.

3.2. Feature detection

3.2.1. Case study 3: precision agriculturePrecision agriculture is a sophisticated method of actively managing within-field variability in

crops, which often makes use of remotely sensed imagery to provide data on spatiotemporal variabil-ity in crops (Mulla 2013). Remote sensing is especially useful for identifying areas needing additionaltreatment with water, chemicals, pesticides, and herbicides. However, widespread adoption of remotesensing in precision agriculture is contingent on affordability, accessibility, and the resolution (spa-tial, temporal, spectral, and radiometric) of the imagery. For small-scale operations, the cost of datafrom conventional remote sensing platforms may be too high. Small UASs potentially offer a low-cost alternative to conventional remote sensing platforms, and research to date shows promise(e.g., Berni et al. 2009; Baluja et al. 2012; Zhang and Kovacs 2012; Bellvert et al. 2013; Calderón et al.2013; Corcóles et al. 2013; Garcia-Ruiz et al. 2013; Gonzalez-Dugo et al. 2013; Huang et al. 2013; Urbahsand Jonaite 2013; Zarco-Tejada et al. 2013; Duan et al. 2014).

The case study for this application involved using imagery from a UAS survey to identify and trackinfections due to late blight (a pathogenic organism that occurs in high-value potato crops). We usedvisible (VIS) and NIR imagery to identify areas of late blight. The primary characteristic of late blight ischlorosis, which occurs when leaves produce insufficient chlorophyll. To detect late blight we used thenormalised difference vegetation index (NDVI), defined by:

NDVI ¼ NIR � VISNIR þ VIS

(1)

We used the Ebee fixed-wing UAS equipped with a Canon IXUS/ELPH 125HS compact camera (16.1MP resolution). To calculate and map NDVI, the UAS was flown twice over the ≈1 km2 field locatedin southern Alberta: once with the camera setting for visible (RGB) light, and once with a NIR filter.A total of 12 flight lines were used to cover the field, resulting in 162 images. The altitude of theUAS was 120 m, yielding a ground resolution of 4.5 cm, which was resampled to 25 cm during the pro-cessing. To expedite the image processing and provide the farmer with quicker results, direct geore-ferencing, using GNSS positions from the UAS log file, was used instead of the more rigorousapproach with GCPs. Typical levels of accuracy obtainable from direct georeferencing without differ-ential correction are in the range of 2–5 m (Turner et al. 2014).

The first survey was flown in mid-July, with a follow-up survey two weeks later. The image mosaicsof the visible and NIR imagery from both surveys are shown along with the computed NDVI maps inFig. 4. Imagery from the first flight in mid-July showed that late blight was present in a small patchlocated in the southeast corner of the field. Identification of late blight is fairly straightforwardbecause it typically begins in small patches, whereas water stress is not as isolated. The presence oflate blight was later confirmed by the farmer, who then applied fungicide in an attempt to managethe infection and limit its spread. A second UAS survey was then performed in late July to determineif the fungicide had been successful. This survey was identical to the first and used the flight para-meters and configuration from the first survey, two weeks earlier. In this case the imagery revealedthat late blight had spread to other patches, thus requiring more intensive fungicide applications tolimit further spread.

Whitehead et al. 93

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

Fig.4.

Theap

plication

ofUAS-based

remotesensingto

iden

tify

late

bligh

tin

apotatocrop

(casestudy3).T

heim

agesequen

cesarefrom

twoseparateflights

in(a–c)

mid-an

d(d–f)late

July

2013.Fo

rthemid-Ju

lyflight,(a)thevisible

imag

ery,

(b)theNIR

imag

ery,

and(c)thecalculatedNDVIaresh

own.Fo

rthelate

July

flight,(d)the

visible

imag

ery,(e)theNIR

imag

ery,an

d(f)thecalculatedNDVIaresh

own.A

reas

withco

nfirm

edor

suspectedlate

bligh

tarecircled.IntheNDVIim

ages

thearea

outside

thefieldwas

masked

.

94 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

3.2.2. Case study 4: pocket gopher mound feature detection and image classificationOne of the unique characteristics of remote sensing imagery from small UAS is that it is able to

resolve small-scale features that are otherwise difficult or impossible to resolve with imagery fromsatellites and manned aircraft. For example, after processing it is not uncommon to obtain groundresolutions of 5 cm or less with RGB imagery acquired from a UAS equipped with an off-the-shelf digi-tal camera. This is about an order of magnitude smaller than the ground resolution of the highest-resolution satellite imagery that is available commercially (e.g., GeoEye-1: 0.41 m). However, the highground resolution also brings a whole host of challenges in terms of adapting image classificationtechniques developed for low-resolution imagery (Laliberte et al. 2010; Laliberte and Rango 2011).

This case study illustrates feature detection of pocket gopher mounds in a grassland region ofsouthwestern Saskatchewan. In June 2012 we acquired RGB imagery from the RQ-84Z AerohawkUAS over an area of roughly 2 km2. The camera used was an Olympus PEN Mini E-PM1 with a 14 mmlens. This camera was precalibrated prior to the survey. A total of 280 images were gathered along14 flight lines, with the overlap and sidelap between strips both set to 65% to ensure full stereo cover-age. The flight was completed in 50 minutes under ideal conditions: clear skies and light winds.

Ground control for the survey consisted of 28 yellow plastic targets distributed throughout thesite. Processing was carried out using the Match-AT module of Trimble’s Inpho software. Followingaerial triangulation, 1 m resolution DTMs and DSMs were created for the survey area. The processused by Inpho to create a DTM involves interpolation between points within a sparse point cloud(Hugenholtz et al. 2013). Consequently it is less sensitive to the effects of vegetation than a surfacemodel produced using a dense point cloud, and will often provide a better representation of the ter-rain in vegetated areas. In contrast a DSM better represents bare areas and areas that have low con-trast variations. The DSM covering the bare areas, and the DTM covering the vegetated areas, werecombined to form a composite DEM, which was used to generate a set of 0.1 m resolution orthoimages.The final step was to combine these to form a seamless orthoimage mosaic (Fig. 5).

While the primary goal was to test the vertical accuracy of a DEM derived from the UAS imageryrelative to airborne LiDAR data (Hugenholtz et al. 2013), the final mosaicked orthoimage revealedthe pervasiveness of a distinct feature that was otherwise unremarkable from the ground. Litteredextensively across the survey area were small mounds of sand, usually 1 m in diameter or less, causedby the digging activities of fossorial mammals like the northern pocket gopher (Thomomys talpoides)and thirteen-lined ground squirrel (Ictidomys tridecemlineatus). These are highly active burrowing mam-mals that leave numerous, but relatively small, mounds of excavated sand above their burrownetworks.

While there has been a significant amount of research on the impact fossorial mammal distur-bance has on ecosystems at a localised scale, there is little in the literature to suggest how widespreadsuch “bioturbation features” actually are. For the most part, their intensity of disturbance in any onelandscape has largely relied on proxies, such as population censuses or rough estimates of moundcoverage, with 5%–25% being commonly reported numbers (see, e.g., Richens 1966; Turner et al.1973; 1980; Grant et al. 1980; Hobbs and Mooney 1985; Spencer et al. 1985; Huntly and Inouye 1988;Reichman et al. 1993). The majority of mounds are too small to distinguish using traditional remotesensing imagery, limiting all research to what is logistically feasible to achieve at the scale of a researchplot (i.e., <1 acre). With 10 cm or better resolutions over relatively large areas, imagery from small UASspresents an opportunity to directly measure the footprint of fossorial mammal disturbance at the scaleof a landscape for the first time.

An automated workflow was developed using object-based image classification to avoid manuallydigitizing the mounds (estimated to take >3 months). Using the feature extraction module in ENVI 5,groups of similar pixels in the image were segmented to create image objects at a scale fine enough torecognise the mounds. Spatial, spectral, and textural attributes for each image object were then calcu-lated based on both the RGB bands in the original mosaic, and several custom bands generated in apreprocessing step that enhanced different aspects of mound characteristics (edges, brightness,etc.). Classification rule sets to isolate the mound objects from the background grassland matrixwere then constructed based on the object attributes. Due to variability in focus and issues with bidir-ectional reflectance, no single rule set could accurately separate mounds across the entire mosaic. As aresult, the image was broken into five locally similar subzones, with separate classification rule setsbuilt for each.

The final classification produced a binary vector map of mounds at the study site (Fig. 5). Thisallowed total area of disturbance to be assessed at a landscape scale, showing that while moundscan be locally extremely abundant, in aggregate they only cover ∼1% of the landscape; one fifth ofthe area than has previously been estimated using ground-based methods.

Whitehead et al. 95

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

3.3. Animal and wildlife monitoring

3.3.1. Case study 5: monitoring a salmon spawning eventBecause of logistical challenges, cost, and the requirement for high-resolution imagery, small UASs

can fill an important niche for wildlife researchers. The imagery or video collected by the UAS can beused for wildlife enumeration or detecting the presence or absence of wildlife. This case studydescribes efforts to map the annual salmon run in October 2010 along the Adam’s River in southernBritish Columbia. The goal was to provide an overview of salmon locations; salmon have importantecological and socioeconomic implications in the region.

The October 2010 sockeye salmon run along the Adam’s River was believed to have been the lar-gest in the last 100 years, with an estimated four million salmon returning to the river (Adams RiverSalmon Society 2013). During the peak of this run, a UAS survey was carried out along a 1 km segmentof the river near Shuswap Lake. The survey involved the Outlander UAV, which is a lightweight fixed-wing UAV with a 2.5 m wingspan. The aircraft carried a Panasonic Lumix-LX3 camera, which is light-weight, compact, and has a retractable lens assembly. The camera was used at its widest angle setting,giving it an effective focal length of 5.1 mm.

The main purpose of the survey was to produce an image mosaic that would allow the majorconcentrations of salmon, and individual salmon on the spawning grounds, to be clearly identified.

Fig. 5. Object-based classification of gopher mounds in the Great Sand Hills, Sask. (case study 4). (a) Colourcomposite of the 1.92 km2 study site. (b) Detail close-up of a roughly 60 m × 60 m area with a linear stretch tohighlight the bright yellow mounds. (c) Result of the binary object-based feature extraction (mounds in black). (d)Cluster of 1–2 year old mounds as they appear on the ground.

96 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

The river survey was flown in four strips at a height of 150 m above ground level, which gave a spatialresolution of better than 5 cm. The aircraft was launched from a sandy beach adjacent to the river, andwas recovered after the survey by flying it into a net.

Five GCPs were placed at accessible locations along the river. For this application, high spatial accu-racy was not required, and horizontal errors of up to 2 m were considered acceptable. Processing wascarried out using Trimble’s Inpho software, with all GCPs being given the same nominal elevation.Rather than using a DEM to orthorectify the images, a simple flat surface was used to rectify the ima-gery. This process is sufficient to remove scale and perspective distortions, but residual relief distor-tions within the imagery are not corrected for. The final mosaic was assembled from the rectifiedimages, and is shown in Fig. 6.

The imagery acquired provides a unique perspective of the distribution of salmon within the river,and clearly shows individual salmon occupying the spawning beds. Conventional aerial photographywould not be able to provide the level of resolution, nor the flexibility in the timing of image

Fig. 6. Georeferenced image mosaic of the lower Adams River, B.C. (case study 5). Insets show largeconcentrations of salmon. Spawning salmon can be seen as individual dots adjacent to the main concentrations.Imagery provided by Accuas Services, Inc.

Whitehead et al. 97

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

acquisition that was achieved using the small UAS. When combined with external information fromsources, such as fish tagging, UAS imagery could potentially be used to arrive at an estimate of salmonnumbers at the time of the survey. In this case there was no such external information available, but itis estimated that the salmon shown in Fig. 6 number in the hundreds of thousands.

3.3.2. Case study 6: cattle enumerationFor larger mammals, imagery acquired from small UASs has been quite successful at detecting

individuals, for example: elephants (Vermeulen et al. 2013), rhinoceros (Mulero-Pazmany et al. 2014),manatees (Jones et al. 2006), birds (Jones et al. 2006) (Chabot and Bird 2012) (Sarda-Palomera et al.2012), and dugong (Hodgson et al. 2013). In most published examples the enumeration has been con-ducted with RGB imagery, but to improve detection against a complex background, Mulero-Pazmanyet al. (2014) suggest it may be preferable to use thermal imagery.

Our animal enumeration case study was carried out with the Aeryon Scout Pro quadcopter and anintegrated thermal forward-looking infrared camera. The application involved beef cattle enumera-tion at a concentrated animal feeding operation (CAFO) or feedlot. The CAFO consisted of separateholding pens where different groups of cattle were held. The flight was performed in late fall 2012,which provided reasonable temperature contrast between the ground and the cattle. The UAS wasflown approximately 100 m above the ground along five flight lines. Individual frames from the ther-mal forward-looking infrared video were extracted and stitched together with Microsoft’s Image Com-posite Editor to produce the final mosaic. Given the application, georeferencing was not required.Individual cows were then identified from manual interpretation of the imagery.

A portion of the CAFO is shown in Fig. 7a, with individual animals denoted by points in Fig. 7b. Inmost cases, the animals are clearly resolved from the background soil, which makes identificationstraightforward. There is some distortion in parts of the image caused by animals moving in betweenflight lines, which introduces some uncertainty to the counts. Capturing a single frame over each penwould likely reduce uncertainty, but this would also require an intervention by the operator to posi-tion the UAS manually, because the footprints of images acquired during autonomous operation areunlikely to be optimal for this purpose.

3.4. Landform dynamics

3.4.1. Case study 7: repeat survey of an arctic glacierSmall UASs are ideal for multitemporal remote sensing aimed at identifying and quantifying envir-

onmental changes. Under appropriate weather conditions, the baseline interval between UAS surveyscan be specifically tailored to match the pace of the environmental changes under investigation, orcan match the interval required for decision making. Furthermore, the ability to develop multitem-poral DSMs and DTMs enables researchers and professionals to track and quantify morphodynamicchanges in three dimensions. Already, researchers have used multitemporal imagery from smallUASs to track landslides (Niethammer et al. 2012; Lucieer et al. 2014; Rothmund et al. 2013), monitor

Fig. 7. Cattle enumeration based on thermal imaging from the Aeryon Scout Pro UAS (case study 6). (a) Portion ofthe CAFO; (b) oblique photo with individual cows identified. Two half-ton pickup trucks are circled in (a) for scale.

98 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

crop growth (Bendig et al. 2013), count birds (Sarda-Palomera et al. 2012), monitor crop water stress(Stagakis et al. 2012), and monitor changes in various vegetation indices (Zarco-Tejada et al. 2013).

For this case study, a UAS survey was carried out in the summer of 2010 over the terminus of anarctic glacier on Canada’s Bylot Island. A follow-up survey was carried out a year later, using imagerycaptured from a piloted helicopter. Further details on this study are described by Whitehead et al.(2013). The resulting DEMs and orthoimage mosaics allowed detailed estimates to be made of surfacethinning, marginal retreat, and flow rates at the glacier terminus, over the one year interval betweenthe surveys.

The 2010 UAS survey was carried out using the Outlander fixed-wing UAS, which carried a LumixLX3 compact camera with an effective focal length of 5.1 mm. The aircraft flew 16 north–south flightlines, and collected 148 photos from a nominal altitude of 300 m above the glacier surface. The surveycovered the lower 1.5 km of the glacier adjacent to the terminus and took approximately 30 min tocomplete. Prior to the aerial survey, 16 targets were placed across the glacier surface, and in the adja-cent moraine regions.

This job was processed using Trimble’s Inpho software, with the photos from every second stripbeing used, except for the steeply sloping valley sides, where extra photos were added from theunused strips to ensure full stereo coverage. Eight of the targets were used as horizontal and verticalGCPs, with the remaining targets being used as check points. The accuracy assessment showed root-mean-square RMSEs for the check points of 0.18, 0.21, and 0.42 m in X, Y, and Z directions, respectively.After the triangulation was completed, a 1 m resolution DEM was created. This was used to orthorec-tify each image, prior to the creation of an orthoimage mosaic, with a resolution of 10 cm.

For the 2011 survey, a Lumix GF-1 camera, equipped with a 14 mm lens was attached to the landinggear of a manned helicopter. The helicopter flew over the glacier in a series of north–south flight lines,with the camera being triggered manually every four seconds. In this case the helicopter had a nominalaltitude of 400 m above the glacier surface and preliminary photo centre coordinates were estimatedfrom the 2010 mosaic. The processing chain was similar to that for the previous year, although in thiscase natural features were used for GCPs and check points, because of time constraints. HorizontalRMSE at the check points were higher than for the previous year at 0.63 and 0.52 m, in X and Ydirections, respectively, likely because target positions could not be as accurately identified. However,the vertical RMSE of 0.19 m was better than for the previous year. As with the previous year, a 1 mresolution DEM and a 10 cm resolution orthoimage mosaic were produced. The orthoimage mosaicsproduced from the 2010 and 2011 surveys are shown in Fig. 8.

Using the DEMs generated from the two surveys, it was possible to quantify the thinning of the gla-cier from 2010 to 2011. The orthoimage mosaics also allowed changes to the terminus extents to bedirectly measured, making it possible to quantify glacial retreat over a one year period. The magni-tude and direction of surface flow was also measured, by comparing the positions of approximately400 points on the glacier surface. This study shows how comparative measurements can be made of

Fig. 8. Fountain Glacier orthoimage mosaics from (a) 2010, and (b) 2011 (case study 7). Both images have 10 mcontours superimposed.

Whitehead et al. 99

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

rapidly changing landscape features. It is likely that glaciology is one field that will benefit signifi-cantly from the availability of low-cost, on-demand imagery acquired from small UASs.

4. Discussion and conclusions

While the case studies described herein are selective in their application, they are sufficientlydiverse to illustrate many of the major benefits and challenges currently associated with the use ofsmall UASs. They also provide a good snapshot of the present state of the industry. It can be seenthat photogrammetric and mapping-type applications are well suited to the capabilities of the currentgeneration of small UASs. The concomitant development of structure frommotion software packages,such as Photoscan and Pix4D, is also helping improve the application of small UASs for mapping.

The use of small UASs for photogrammetric surveying can meet a need for detailed, high-accuracysurveys of areas in the range of 1–10 km2. This has traditionally marked the boundary between ground-based surveys and photogrammetric surveys from manned aircraft. For areas of this size, groundsurveying is typically too time-consuming to be cost effective. The point density for ground surveysis generally much lower than that collected from a UAS, and while the accuracy of individual GNSSpoints can be very high, a survey with a small UAS will typically produce a DEM that contains muchmore detail and higher accuracy overall. Traditional photogrammetric surveying is typically inflexiblein terms of scheduling. It is also very expensive, and only becomes economical for high value projects,or where large areas need to be covered. Because of the higher flying height, the ground resolution ofphotography acquired from manned aircraft is typically much lower than that gathered by smallUASs. Surveying using small UASs is cost effective, flexible in terms of scheduling, and provides thehighest resolution imagery available.

What these case studies also show is that currently the main use of the orthoimagery from smallUASs is for qualitative analysis. There are many challenges, some of which are outlined in Whiteheadand Hugenholtz (2014), that limit that ability to straightforwardly adapt automated mapping techni-ques developed for satellite and manned aircraft. Thus far, the imaging technology adapted for smallUASs has not kept pace with the advances in UAS platforms and capabilities. Problems relating toimage quality and illumination differences between images generally make imagery acquired fromsmall UASs unsuitable for quantitative analysis. While the land classification study described hereindid involve some automated image classification, illumination differences meant that the job hadto be separated into five independent zones prior to classification, with unique classification rulesdeveloped for each zone. This example also serves to illustrate how object-based classification algo-rithms, which take into account shape, textural, and contextual attributes, are likely preferable topixel-based classification for the analysis of imagery from small UASs.

With the continued expansion of small UASs for remote sensing and mapping, it is likely that newsensors will be developed that better match the needs of researchers and the market. For imaging sen-sors, key areas that need to be addressed include size and weight, image quality, sensor stability, sig-nal-to-noise ratios, and the ability to record imagery across multiple wavebands simultaneously. Thenext generation of sensors, designed specifically for incorporation within small UASs, are likely toaddress many of these shortcomings. The capability for automated image analysis will greatly increasethe utility of small UASs for many applications, and is likely to lead to greater uptake in many sectors.

In the case studies described here, we have attempted to show the types of applications for whichsmall UASs are currently suitable. At present these are heavily biased towards photogrammetric appli-cations. While small UASs do not provide a universal solution to all mapping requirements, they offera cost-effective alternative to ground surveying and traditional photogrammetry for many jobs. Withthe continued evolution of platforms and sensors, it is likely that the range of applications for whichsmall UASs are suitable will continue to expand.

Acknowledgements

The authors are grateful for the financial support received for the case studies. Funding was pro-vided by the Natural Sciences and Engineering Research Council of Canada, Cenovus Energy, AlbertaInnovates, the Canadian Foundation for Innovation, and the University of Calgary. J. Walker (Isis Geo-matics) is acknowledged for providing assistance with some of the data acquisition and image proces-sing. Imagery of the Adams River was provided by Accuas Services, Inc.

ReferencesAdams River Salmon Society. 2013. 2010 Late Run Sockeye Salmon Preliminary Escapement Estimates. Available from http://www.salmonsociety.com/DFO%20salmon%20count.pdf. Accessed 26 June 2014.

100 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

Baluja, J., Diago, M.P., Balda, P., Zorer, R., Meggio, F., Morales, F., and Tardaguila, J. 2012. Assessment of vineyard water status varia-bility by thermal and multispectral imagery using an unmanned aerial vehicle (UAV). Irrig. Sci. 30(6): 511–522. doi: 10.1007/s00271-012-0382-9

Bellvert, J., Zarco-Tejada, P.J., Girona, J., and Fereres, E. 2013. Mapping crop water stress index in a Pinot-noir vineyard: comparingground measurements with thermal remote sensing imagery from an unmanned aerial vehicle. Precision Agriculture. 15(4): 361–376. doi: 10.1007/s11119-013-9334-5

Bendig, J., Bolten, A., and Bareth, G. 2013. UAV-based Imaging for Multi-Temporal, very high resolution crop surface models tomonitor crop growth variability. Photogrammetrie-Fernerkundung-Geoinformation. 2013(6): 551–562

Berni, J., Zarco-Tejada, P.J., Suarez, L., and Fereres, E. 2009. Thermal and narrowband multispectral remote sensing for vegetationmonitoring from an unmanned aerial vehicle. IEEE Trans. Geosci. Remote Sens. 47(3): 722–738. doi: 10.1109/TGRS.2008.2010457

Calderón, R., Navas-Cortés, J.A., Lucena, C., and Zarco-Tejada, P.J. 2013. High-resolution airborne hyperspectral and thermal imageryfor early detection of Verticillium wilt of olive using fluorescence, temperature and narrow-band spectral indices. Remote Sens.Environ. 139: 231–245. doi: 10.1016/j.rse.2013.07.031

Chabot, D., and Bird, D.M. 2012. Evaluation of an off-the-shelf unmanned aircraft system for surveying flocks of geese. Waterbirds.35(1): 170–174. doi: 10.1675/063.035.0119

Corcóles, J.I., Ortega, J.F., Hernández, D., and Moreno, M.A. 2013. Estimation of leaf area index in onion (Allium cepa) using anunmanned aerial vehicle. Biosystems Eng. 115(1): 31–42. doi: 10.1016/j.biosystemseng.2013.02.002

Crosetto, M., and Aragues, F.P. 2000. Radargrammetry and SAR interferometry for DEM generation validation and data fusion. Pro-ceedings of the CEOS SAR Workshop, Toulouse, 26–29 October 1999, ESA SP-450. Available from http://icc.cat/index.php/content/download/3860/12916/file/radargrammetry_and_sar_interferommetry.pdf. Accessed 26 June 2014.

Duan, S.B., Li, Z.L., Wu, H., Tang, B.H., Ma, L., Zhao, E., and Li, C. 2014. Inversion of the PROSAIL model to estimate leaf area index ofmaize, potato, and sunflower fields from unmanned aerial vehicle hyperspectral data. Int. J. Appl. Earth Obs. Geoinf. 26: 12–20.doi: 10.1016/j.jag.2013.05.007

Eisenbeiss, H. 2009. UAV photogrammetry. Eidgenössische Technische Hochschule, Zürich.Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L., and Carbonneau, P.E. 2013. Topographic structure from motion: a newdevelopment in photogrammetric measurement. Earth Surf. Processes Landforms. 38(4): 421–430. doi: 10.1002/esp.3366

Foster, M.A., and Stubbendieck, J. 1980. Effects of the plains pocket gopher (Geomys bursarius) on rangeland. J. Range Manage. 33(1):74–78. doi: 10.2307/3898233

Garcia-Ruiz, F., Sankaran, S., Maja, J.M., Lee, W.S., Rasmussen, J., and Ehsani, R. 2013. Comparison of two aerial imaging platformsfor identification of Huanglongbing-infected citrus trees. Comput. Electron. Agric. 91(0): 106–115. doi: 10.1016/j.compag.2012.12.002

Gersher, S. 2014. Canada’s domestic regulatory framework for RPAS. A call for public deliberation. J. Unmanned Veh. Sys. 2(1): 1–4.doi: 10.1139/juvs-2013-0016

Glennie, C.L., Carter, W.E., Shrestha, R.L., and Dietrich, W.E. 2013. Geodetic imaging with airborne LiDAR: the Earth’s surfacerevealed. Rep. Prog. Phys. 76(8): 086801. PMID: 23828665. doi: 10.1088/0034-4885/76/8/086801

Gonzalez-Dugo, V., Zarco-Tejada, P., Nicolás, E., Nortes, P.A., Alarcón, J.J., Intrigliolo, D.S., and Fereres, E. 2013. Using high resolutionUAV thermal imagery to assess the variability in the water status of five fruit tree species within a commercial orchard. PrecisionAgriculture. 14(6): 660–678. doi: 10.1007/s11119-013-9322-9

Grant, W.E., French, N.R., and Folse, L.J. Jr. 1980. Effects of pocket gopher mounds on plant production in shortgrass prairie ecosys-tems. Southwestern Naturalist. 25(2): 215–224. doi: 10.2307/3671243

Haala, N., Cramer, M., Weimer, F., and Trittler, M. 2011. Performance test on UAV-based photogrammetric data collection. Proceed-ings of the International Archives of the Photogrammetry. Remote Sensing and Spatial Information Sciences. 38(1/C22): 7–12

Hardin, P.J., and Jensen, R.R. 2011. Small-scale unmanned aerial vehicles in environmental remote sensing: challenges and oppor-tunities. GIScience & Remote Sensing. 48(1): 99–111

Hobbs, R.J., and Mooney, H.A. 1985. Community and population dynamics of serpentine grassland annuals in relation to gopher dis-turbance. Oecologia. 67(3): 342–351. doi: 10.1007/BF00384939

Hodgson, A., Kelly, N., and Peel, D. 2013. Unmanned aerial vehicles (UAVs) for surveying marine fauna: a dugong case study. PLoSOne. 8(11): e79556. PMID: 24223967. doi: 10.1371/journal.pone.0079556

Huang, Y., Thomson, S.J., Hoffmann, W.C., Lan, Y., and Fritz, B.K. 2013. Development and prospect of unmanned aerial vehicle tech-nologies for agricultural production management. Int. J. Agric. Biol. Eng. 6(3): 1–10

Hugenholtz, C.H., Moorman, B.J., Riddell, K., and Whitehead, K. 2012. Small unmanned aircraft systems for remote sensing andearth science research. Eos, Transactions American Geophysical Union. 93(25): 236, doi: 10.1029/2012EO250005

Hugenholtz, C.H., Whitehead, K., Barchyn, T.E., Brown, O.W., Moorman, B.J., LeClair, A., Hamilton, T., and Riddell, K. 2013. Geomor-phological mapping with a small unmanned aircraft system (sUAS): feature detection and accuracy assessment of a photogram-metrically-derived digital terrain model. Geomorphology. 194: 16–24. doi: 10.1016/j.geomorph.2013.03.023

Hunt, E.R., Hively, W.D., Fujikawa, S.J., Linden, D.S., Daughtry, C.S.T., and McCarty, G.W. 2010. Acquisition of NIR-green-blue digitalphotographs from unmanned aircraft for crop monitoring. Remote Sens. 2(1): 290–305. doi: 10.3390/rs2010290

Huntly, N., and Inouye, R. 1988. Pocket gophers in ecosystems: patterns and mechanisms. BioScience. 38(11): 786–793. doi: 10.2307/1310788

Jones, G.P., Pearlstine, L.G., and Percival, H.F. 2006. An assessment of small unmanned aerial vehicles for wildlife research. Wildl.Soc. Bull. 34(3): 750–758. doi: 10.2193/0091-7648(2006)34[750:AAOSUA]2.0.CO;2

Knoth, C., Klein, B., Prinz, T., and Kleinebecker, T. 2013. Unmanned aerial vehicles as innovative remote sensing platforms for high-resolution infrared imagery to support restoration monitoring in cut-over bogs. Appl. Veg. Sci. 16(3): 509–517. doi: 10.1111/avsc.12024

Laliberte, A.S., Herrick, J.E., Rango, A., and Winters, C. 2010. Acquisition, orthorectification, and object-based classificationof unmanned aerial vehicle (UAV) imagery for rangeland monitoring. Photogramm. Eng. Remote Sens. 76(6): 661–672. doi:10.14358/PERS.76.6.661

Laliberte, A.S., and Rango, A. 2011. Image processing and classification procedures for analysis of sub-decimeter imagery acquiredwith an unmanned aircraft over arid rangelands. GIScience & Remote Sensing. 48(1): 4–23. doi: 10.2747/1548-1603.48.1.4

Lin, Y., Hyyppa, J., and Jaakkola, A. 2011. Mini-UAV-borne LIDAR for fine-scale mapping. IEEE Geosci. Remote Sens. Lett. 8(3): 426–430. doi: 10.1109/LGRS.2010.2079913

Lucieer, A., de Jong, S., and Turner, D. 2014. Mapping landslide displacements using structure from motion (SfM) and image corre-lation of multi-temporal UAV photography. Prog. Phys. Geog. 38(1): 97–116. doi: 10.1177/0309133313515293

Martinez-De Dios, J.R., and Ollero, A. 2006. Automatic detection of windows thermal heat losses in buildings using UAVs. Automa-tion Congress, 2006. WAC ’06. doi: 10.1109/WAC.2006.375998.

Mulero-Pazmany, M., Stolper, R., van Essen, L.D., Negro, J.J., and Sassen, T. 2014. Remotely piloted aircraft systems as a rhinocerosanti-poaching tool in Africa. PLoS One. 9(1): e83873. PMID: 24416177. doi: 10.1371/journal.pone.0083873

Whitehead et al. 101

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.

Mulla, D.J. 2013. Twenty five years of remote sensing in precision agriculture: key advances and remaining knowledge gaps. Biosys-tems Eng. 114(4): 358–371. doi: 10.1016/j.biosystemseng.2012.08.009

Niethammer, U., James, M.R., Rothmund, S., Travelletti, J., and Joswig, M. 2012. UAV-based remote sensing of the Super-Sauze land-slide: evaluation and results. Eng. Geol. 128: 2–11. doi: 10.1016/j.enggeo.2011.03.012

Reichman, O.J., Bendix, J.H., and Seastedt, T.R. 1993. Distinct animal-generated edge effects in a tallgrass prairie community. Ecol-ogy. 74(4): 1281–1285. doi: 10.2307/1940496

Richens, V.B. 1966. Notes on the digging activity of a northern pocket gopher. J. Mammal. 47(3): 531–533. doi: 10.2307/1377704Rothmund, S., Niethammer, U., Malet, J.P., and Joswig, M. 2013. Landslide surface monitoring based on UAV-and ground-basedimages and terrestrial laser scanning: accuracy analysis and morphological interpretation. First Break. 31(8): 81–87

Sarda-Palomera, F., Bota, G., Vinolo, C., Pallarés, O., Sazatornil, V., Brotons, L., Gomariz, S., and Sarda, F. 2012. Fine-scale bird mon-itoring from light unmanned aircraft systems. Ibis. 154(1): 177–183. doi: 10.1111/j.1474–919X.2011.01177.x

Sauerbier, M., Siegrist, E., Eisenbeiss, H., and Demir, N. 2011. The practical application of UAV-based photogrammetry under eco-nomic aspects, International Archives of the Photogrammetry.. Remote Sensing and Spatial Information Sciences. ISPRS Zurich2011 Workshop, 14–16 September 2011. 38(1/C22): 45–50

Spencer, S.R., Cameron, G.N., Eshelman, B.D., Cooper, L.C., and Williams, L.R. 1985. Influence of pocket gopher mounds on a Texascoastal prairie. Oecologia. 66(1): 111–115. doi: 10.1007/BF00378561

Stagakis, S., González-Dugo, V., Cid, P., Guillén-Climent, M.L., and Zarco-Tejada, P.J. 2012. Monitoring water stress and fruit qualityin an orange orchard under regulated deficit irrigation using narrow-band structural and physiological remote sensing indices.ISPRS J. Photogramm. Remote Sens. 71: 47–61. doi: 10.1016/j.isprsjprs.2012.05.003

Tamminga, A., Hugenholtz, C., Eaton, B., and Lapointe, M. 2014. Hyperspatial remote sensing of channel reach morphology andhydraulic fish habitat using an unmanned aerial vehicle (UAV): a first assessment in the context of river research and manage-ment. River Res. Appl. doi: 10.1002/rra.2743

Transport Canada. 2008. The review and processing of an application for a Special Flight Operations Certificate for the Operation ofan Unmanned Air Vehicle (UAV) System. Available from http://www.tc.gc.ca/eng/civilaviation/opssvs/managementservices-referen-cecentre-documents-600623-001-972.htm. Accessed 26 June 2014.

Transport Canada. 2012. UAV systems program design working group phase 1 final report. Available from http://www.h-a-c.ca/UAV_REPORT.pdf. Accessed 26 June 2014.

Turner, D., Lucieer, A., and Wallace, L. 2014. Direct georeferencing of ultrahigh-resolution UAV imagery. IEEE Trans. Geosci. RemoteSens. 52(5): 2738–2745. doi: 10.1109/TGRS.2013.2265295

Turner, D., Lucieer, A., and Watson, C. 2012. An automated technique for generating georectified mosaics from ultra-high resolu-tion unmanned aerial vehicle (UAV) imagery, based on structure from motion (SfM) point clouds. Remote Sens. 4(5): 1392–1410.doi: 10.3390/rs4051392

Turner, G.T., Hansen, R.M., Reid, V.H., Tietjen, H.P., and Ward, A.L. 1973. Pocket gophers and Colorado mountain rangeland. Color-ado State University Agriculture Experimental Station Bulletin. 554S: 90p.

Urbahs, A., and Jonaite, I. 2013. Features of the use of unmanned aerial vehicles for agriculture applications. Aviation. 17(4): 170–175.doi: 10.3846/16487788.2013.861224

Vermeulen, C.D., Lejeune, P., Lisein, J., Sawadogo, P., and Bouché, P. 2013. Unmanned aerial survey of elephants. PLoS One. 8(2):e54700. PMID: 23405088. doi: 10.1371/journal.pone.0054700

Wallace, L., Lucieer, A., Watson, C., and Turner, D. 2012. Development of a UAV-LiDAR system with application to forest inventory.Remote Sens. 4(6): 1519–1543. doi: 10.3390/rs4061519

Watts, A.C., Ambrosia, V.G., and Hinkley, E.A. 2012. Unmanned aircraft systems in remote sensing and scientific research: classifi-cation and considerations of use. Remote Sens. 4(6): 1671–1692. doi: 10.3390/rs4061671

Westaway, R.M., Lane, S.N., and Hicks, D.M. 2000. The development of an automated correction procedure for digital photogram-metry for the study of wide, shallow, gravel-bed rivers. Earth Surf. Processes Landforms. 25(2): 209–226. doi: 10.1002/(SICI)1096-9837(200002)25:2<209::AID-ESP84>3.0.CO;2-Z

Whitehead, K., and Hugenholtz, C.H. 2014. Remote sensing of the environment with small unmanned aircraft systems (UASs), part1: a review of progress and challenges. J. Unmanned Veh. Syst. This issue. doi: 10.1139/juvs-2014-0006

Whitehead, K., Moorman, B.J., and Hugenholtz, C.H. 2013. Low-cost, on-demand aerial photogrammetry for glaciological measure-ment. The Cryosphere. 7(6): 1879–1884. doi: 10.5194/tc-7-1879-2013

Zarco-Tejada, P.J., Guillén-Climent, M.L., Hernández-Clemente, R., Catalina, A., González, M.R., and Martín, P. 2013. Estimating leafcarotenoid content in vineyards using high resolution hyperspectral imagery acquired from an unmanned aerial vehicle (UAV).Agric. For. Meteorol. 171: 281–294. doi: 10.1016/j.agrformet.2012.12.013

Zhang, C., and Kovacs, J.M. 2012. The application of small unmanned aerial systems for precision agriculture: a review. PrecisionAgriculture. 13(6): 693–712. doi: 10.1007/s11119-012-9274-5

102 J. Unmanned Veh. Syst. Vol. 2, 2014

Published by NRC Research Press

J. U

nman

ned

Veh

. Sys

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

UN

IV C

AL

GA

RY

on

09/2

7/18

For

pers

onal

use

onl

y.