23
1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc Geomatics for the Built Environment: (1) Geodata Acquisition Technology & (2) Geodata Quality) Senior Editor GIM International International Consultant (2014 -2015: WB, Kenya) [email protected]

1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

Embed Size (px)

Citation preview

Page 1: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

1Challenge the future

Point Clouds from Lidar and Imagery – Status and Trends

Mathias J.P.M. LemmensDelft University of Technology, The Netherlands (MSc

Geomatics for the Built Environment: (1) Geodata Acquisition Technology &

(2) Geodata Quality)

Senior Editor GIM InternationalInternational Consultant (2014 -2015: WB, Kenya)[email protected]

Page 2: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

2Challenge the future

Racurs 2014 Conference, HainanFeatures of Point Clouds and Functionalities of Processing Software

ISPRS 1988 Congress, KyotoA SURVEY ON STEREO MATCHING TECHNIQUES

IGARSS'97, Singapore Accurate height information from airborne laser-altimetry

ISPRS 1997, Stuttgart,Building detection by fusing airbornelaser-altimeter DEMS and 2D digital maps.

1997 onwards:Over 100 papers in GIM Internationalon Photogrammetry and Lidar

Page 3: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

3Challenge the future

Agenda

• General Developments• Dense Image Matching on Oblique Images• Multispectral Airborne Lidar• Airborne Radar• SLAM: Indoor 3D Modelling of Indoor Scenes

Page 4: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

4Challenge the future

Nucleus of Point Clouds

Page 5: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

5Challenge the future

Developments

Applications are steadily growing

Variety of Sensors create increasingly dense point clouds

Variety of Processing Software

How to tackle the storage and fast retrieval problem?

Page 6: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

6Challenge the future

DIM allows point densities similar to the ground sampling distance (GSD) of the imagery: GSD of 10cm 100 height points per square meter.

Driving Forces - Programmable graphical processing units (GPUs) - New algorithms Semi-global matching (SGM) algorithm introduced by Hirschmüller (2008) - Cheap computer power - Cheap digital cameras provide high-quality imagery, while large overlaps do not add to costs - Open source packages available from computer vision.

Dense Image Matching (DIM)

Page 7: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

7Challenge the future

Aerial multi camera systems capture oblique and nadir imagery at the same time full and intuitive view on both building footprints and facades beneficial for creating 3D city models.

Maltese crossconcept.

Oblique Images

Page 8: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

8Challenge the future

Oblique images allow to extract denser point cloudswith façades and building completed reconstructedCourtesy: Remondino et al., 2014

Page 9: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

9Challenge the future

Page 10: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

10Challenge the future

Oblique Images

Challenging for oblique imagery:• Large scale variations• Illumination changes• Many occlusions

Many questions are still open (Remondino et al., 2014): - when to use oblique imagery; - what are its strengths and weaknesses; - what is the optimal acquisition patterns for metric mapping; - how to deal with illumination and scale changes - which processing software is reliable and efficient?

Need for performance measures of DIM software for obliqueimagery (See Deuber et al., 2014)

Page 11: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

11Challenge the future

Airborne Lidar

Routinely used for: - 3D modelling of urban areas - capturing boreal forests - Mapping Power Lines, etc. Status: - Increasing laser pulses frequencies (up to one million per second) - Multiple pulses in air - (full) waveform digitization

Trends: - Multispectral Lidar - Photon Lidar - Lidar on a UAS

Riegl-VQ-820 G

Page 12: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

12Challenge the future

The Titan, introduced December 2014, emits independentpulses in 3 narrow spectral bands.(Courtesy: Optech).

Multispectral Lidar

Page 13: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

13Challenge the future

Multispectral Lidar

The 3 beams do not pass the exact same path the 3 multispectral points do not refer to the same terrain point.

Envisioned applications - topographic surveying - shallow water bathymetry - environmental modelling - urban surface mapping - land cover classification.

Combination through gridding raster rather than a point cloud.

False-colour raster image generated using Titan Lidar wavelength combinations(Courtesy of Laserdata GmbH and Optech).

Page 14: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

14Challenge the future

Further ImprovementsEach terrain point is recorded in each of the three wavelengths Manufacturing a system where the beams overlap precisely and the returns are measured simultaneously.

Multispectral Lidar

Page 15: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

15Challenge the future

BradarSAR Brazil

Rockwell OrbiSAR P-band – wavelength 75cm – can penetrate the foliage and reach terrain underneath vegetation

World´s largest aerial mapping project with X- and P-band. Products: DTMs, DSMs, X- and P-band orthoimages and 2880 maps at scale 1:50,000.

Page 16: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

16Challenge the future

P-band shows a road not visible in X-bandCourtesy: Sambatti and Lübeck, 2015

P-band (75cm) X-band (3cm)

Page 17: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

17Challenge the future

Double bounce: P-band signals reflect on the terrain and onlyreturn as backscatter when reflecting again on tree trunks.Courtesy: Sambatti and Lübeck, 2014

Page 18: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

18Challenge the future

Radar DTMs and DSMs provide vegetation height maps and combined with biomass ground truth the biomass/carbon stocks can be estimatedCourtesy: Sambatti and Lübeck, 2014

Page 19: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

19Challenge the future

SLAM

GNSS signals are blocked indoors.

Solution: ‘guessing’ the position and representing the space based on sensor data and prior knowledge.Guesses are iteratively refined using data collected while the robot is moving.Algorithms based on the iterative closest point (ICP) algorithm aimed at minimizing the difference between successive point clouds and the extended Kalman filter.Central is the use of landmarks; features distinct from the background.

Positioning sensors: odometers, INS and lasers.

Simultaneous Localisation And Mapping

Page 20: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

20Challenge the future

Founders of NavVis: Robert Huitl, Sebastian Hilsenbeck, Dr. Georg Schroth,Dr. Felix Reinshagen.

3D Mapping Trolley

SLAM

Page 21: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

21Challenge the future

Point cloud overlaid with images of the shipping hall of the Deutsches Museum

(Courtesy: Reinshagen et al., 2015)

Page 22: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

22Challenge the future

Moscow by Bicycle?

Page 23: 1 Challenge the future Point Clouds from Lidar and Imagery – Status and Trends Mathias J.P.M. Lemmens Delft University of Technology, The Netherlands (MSc

23Challenge the future

Thank you so much for your attention.