Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
DR. SANJIB K. GHOSH
Ass't. Proj., Ohio State Univ.Columbus, Ohio 43210
x- and Y-Parallax
Observations
ABSTRACT: A n experimental study determines the comparative preczswn ofX- and V-parallax observations in stereo-photogrammetric models. Supportingtheories are presented. It was determined empirically that the standard error ofX-parallax observations is about 0.4 times the standard error of V-parallax observations. The effect of nsing dove prisms to change V-parallax to X-parallaxis also studied.
The failure of obtaining an intersectionmeans the existence of parallaxes (X or Y),which are defined by:
Such parallaxes are due to the errors in theorientation elements. The parallaxes existingat a model-point expressed as functions of theerrors in the orientation elements have beenderived and established by various photogram metrists earlier (e.g., Branden berger).Any text book may be referred to for suchderivations (e.g., Hallert, Schweidefsky, orZeller). Considering a right-handed coordinate system with clockwise rotations, withthe origin of the system at the left side projection center and with regard to the use ofnegatives (and not diapositives) the functions are:
INTRODUCTION
T HE FORMATION of a three dimensionalstereo-model involves the intersection
of the optical rays, originating from the twocorresponding picture points, at each point inthe surface of the model. The intersection isperformed when at the model-point (in themodel space) both the rays have the sameX-coordinate (model coordinate) and also thesame V-coordinate. The mathematical expression for this is that at a point P in themodel-space, XPX=XPII' where X PI is theX-coordinate with respect to the ray fromthe left-camera of the stereo-plotting instrument and X pu is the X-coordinate withrespect to the ray from the right-camera.Similarly, for the V-coordinates, Y pl = YpIl'
With a view to studying the feasibility ofusing X-parallaxes for the relative orienta-
(1)
(3)b
-Px = 5ZZ
\ (X - b)'(- Z/1 + Z' Idq,1l
X XY-Px = db~'1 - - dbz, - Y ·dK' - -- dWI
Z Z
r (X') (X - b)+ Z 1 +-- dq" - db~'11 +--- dbz llZ' Z
(X - b) Y+ Y·dKIl + Z dWIl
-Py = dbYI - ~ dbz l +X·dKI - Z (1 +~) dWIZ Z'
XY Y+ - dq" - db)'f1 + - dbz!1 - (X - b)dKIlZ Z
( Y') (X - b)Y+ Z 1 + Z' dWIl - Z dq,f1 (2)
rtis apparen t that the use of ei ther or bothof the parallaxes would result in solving theerrors in the elements of relative orientation.The author was curious about it and wentdeeper into these aspects in a researchsponsored by the U. S. Army Engineer(GIMRADA, Fort Belvoir, Virginia).
The X-parallax at a model point is relatedto the corresponding elevation error according to an expression -Px=oZ(bx/Z); whereOZ is the elevation error. For all practicalpurposes, one can assume that the modelbase (b) is identical with the X-component(bx) of the base. I n practice, this is almostalways true. Then
Px = Xp! - X PIl and,Py = Y PI - Y PIl .
X-parallax (error):Y-parallax (error):
667
668 PHOTOGRAMMETRIC ENGINEERING
tion of a stereo-model and to analyze theaccuracy obtained thereby as compared tothe situation with the' use of V-parallaxes, aseries of experimental observations weremade in the Photogrammetric laboratory ofthe Department of Geodetic Science, TheOhio State University. Below is given an account of the studies.
DESCRIPTION OF THE EXPERIMENT
For these investigations a \Vild Autogra ph A7 was used. The photographs weretaken wi th a Wild RC5 Aerial Camera;1=152.36 mm.; format: 23 cm.X23 cm.;standard photography. Picture scale was1 :8000 approximately. Model Scale was1 :4000.
Observations were made at or very neareach of three different points of the model\\·ith undulating terrain. One of the pointswas on a Aat side-walk that was runningEast-West (parallel to the X-axis), the secondpoint was on another side-walk that wasrunning North-South (parallel to the Y-axis)and the third point was on a grassy tennislawn with lines crossing each other at rightangles. Three hundred repeat-observationswere made at each point, thus making thetotal n'umber of observations 900. The 300observations at each point were broken downas follows:
Case AI: 50 observations of V-parallaxelimination with the element by" and corresponding readings of the by" counter bymoving the right side measuring mark fromfront to rear;
Case A2: 50 observations of V-parallaxelimination with the element by" and corresponding readings of the by" counter bymoving the right side measuring mark fromrear to front. lext, after changing the Yparallax into X-parallax with Dove-prisms,
Case Bl: 50 observations of V-parallaxelimination with the element by" and corresponding readings of the by" counter byapproaching the model surface with themeasuring mark from above;
Case B2: 50 observations of V-parallaxelimination with the element by" and corresponding readings of the by" counter byapproaching the model surface with themeasuring mark from below. Next, aftercoming back to the normal situation, bysetti ng the Dove-prisms back to their original situations,
Case Cl: 50 observations of X-parallaxelimination (i.e., reading of spot elevations)
with the foot-disk and the correspondingZ-counter readings by approaching themodel surface with the measuring markfrom above;
Case C2: 50 observations of X -parallaxelimination (i.e., reading of spot elevations) with the foot-disk and the corresponding Z-counter readings by approaching the model surface with the measuringmark from below.
The relevant elements were set at the instrument to the following values: j= 152.36mm.; Z~304.7 mm.; bx=157,20 mm.; by'=by"=bz'=bz"=100.00 (i.e., zero value)m m.
Assuming the arithmetic mean to be thecorrect value for each case (separately forAI, A2, B1, B2, C1 and C2), the standarderror for each case was computed. These are,as averages obtained from all the threepoints:
Case AI: 15.7 microns, and Case A2:13.5 microns; their average gives the standard error for Case A : 14.6 microns.
Case B1: 5.1 microns and Case B2: 5.7microns; their average gives the standarderror for Case B: 5.4 microns.
Case Cl: 10.9 microns and Case C2:18.8 microns; their average gives thestandard error for Case C: 14.8 microns.
However this Case C gives the standarderror of elevation determination in the model,which is related to the X-parallax accordingto the expression (3). This gives the standarderror of X-parallax determination from caseC to be, with proper substitutions in the expression (3):
157.214.8-- = 7.5 microns.
304.7
All the above are expressed as in the model.In these studies the model scale was approximately twice the picture scale. From thisconsideration, for the sake of standardization, it may finally be expressed that at thepicture scale (in this particular case):
The standard error of V-parallax determination is 7.3 microns.
The standard error of V-parallax determination, by changing them to X-parallax with Dove-prisms, is 2.7 microns.
The standard error of X -parallax deterfl1ination (from elevation observations)is 3.8 microns.
x- AND Y-PARALL.\X OBSERV.\TIONS 669
The above shows that the standard errorof X-parallax determination is between 0.37and 0.52 (i.e., between 2.7/7.3 and 3.8/7.3)~0.44, depending on the element used for thepurpose) times the standard error of Y-parallax determination. This means, generallyspeaking, the co-factors are related to eachother as follows:
Qpz,Pz = (O.44)2Qpy,py = O.2Qpy,py.
In case, however, Dove-prisms are uscd forchanging the Y-parallax to X-parallax, generally speaking, QPx,Px~Qpy,py.
The author wishes to thankfully acknowledge the guidance and support he received
from his adviser Dr. A. ]. Brandenberger inthis study, which was a part of a researchsponsored and supported by the U. S. ArmyEngineer, Geodesy, Intelligence and Mapping Research and Developmen t AgencyFort Belvoir, Virginia.
BIBLIOGRAPHY
1. Brandenberger, A. ]. "Fehlertheorie der ausseren Orientierung "on Steilaufnahlllcn. Diss.,E.T.H., ZUrich, J9117.
2. Ghosh, S. K. "J J1\'estigations into the problemof relati"e orientation." Diss., Ohio State University, 1964.
3. Jerie, H. G. Vorzeichenfragen an rallmlichenAliswertegeraten. Photogralllllletria, 1955-56,vol. 1.
ESSA, NEW FEDERAL !-\CE:-ICY,
INCLUDES THE COAST & GEODETIC SURVEY
President L. B. Johnson sent to Congressfor approval on May 13, 1965 a plan for thereorganization of two major agencies in theDepartment of Commerce: the \VeatherBureau and the Coast & Geodetic Survey.
Under the plan the \-\feather Bureau, theCoast & Geodetic Survey, and the CentralRadio Propagation Laboratory of thc National Bureau of Standards are to be combined in an agency known as the EI1\·ironmental Scicncc Scn'ices Administration(ESSA).
"The new agcncy will cnablc us to gi"e thcpublic, business and industry better andfaster service, and combining certain management functions will provide the improvedservices at lower cost," Secretary of Commerce John T. Conner said. "But even moreimportant, it will improve the scientific capabilities of the Commerce Department."
"Our studies range over such fields asmeteorology, hydrology, climatology, seismology, geodesy, geomagnetism. oceanography,hydrography, aeronomy, tropospheric electromagnetic propagation and telecomm unica-
tions," he said. "If we are to understand ourenvironment and use it to the fullest advantage, a thorough knowledge of these fields andhow they interact with one another is essential. This will be expedited if the scientists ilreunder one organizational rooL"
The Secretary said the proposal for the newagency grew out of suggestions made by\'Veather Bureau Chief Dr. Robert M. \\'hite,. BS Director Dr. Allen V. Astin, and C&GSDirector r\dmiral H. Arnold Karo, plusstudies conducted under the dircction of Dr..r. Herbert Hollomon, Assistant Secretary ofCommercc for Scicnce and Technology.
The Commerce Departmen t prod uces nautical and aeronautical charts, and forecasts ofweather and ionospheric conditions (important for radio communications). The Department also issues warnings of natural hazards-hurricanes. tornadoes, floods, earthquakes,seismic sea waves, and solar disturbances. Inaddition, it provides the scientific communitywi th a wide variety of information on theenvironment.