The EuroSDR network on Digital Camera Calibration and Validation

Transcription

1 Institut für Photogrammetrie The EuroSDR network on Digital Camera Calibration and Validation Michael Cramer 109th EuroSDR Meetings Lisboa, October 25-27, 06 Phase II Active Participants # Code ICC ICV LM itacyl inpho CSIRO DLR-O DLR-B Anhalt HfT Institutions Institute Cartographic Catalunya, Barcelona, Spain Institute Cartographic Valenciano, Valencia, Spain Lantmatäriet, Gävle, Sweden ITACYL, Valladolid, Spain inpho, Stuttgart, Germany CSIRO Information Sciences, Wembley, Australia DLR, Oberpfaffenhofen, Germany DLR, Berlin, Germany University of Applied Science, Anhalt, Germany University of Applied Science, Stuttgart, Germany

2 Phase II Active Participants # Code UoL IPI ETH UoP UoN Ingr.ZI Vexcel Leica Institutions University of Leon, Spain IPI, University of Hannover, Germany ETH Zürich, Switzerland University of Pavia, Italy University of Nottingham, England Intergraph ZI, Aalen, Germany Vexcel, Graz, Austria Leica Geosystems, Heerbrugg, Switzerland The whole EuroSDR network list includes 60 entries Phase II Data Set ADS UltracamD # Requests Phase II Data request Participant (w/o manufacturer) UoP, DLR-B, DLR-O, ETH, inpho ICC, IPI, Anhalt, inpho, HfT, ETH, UoL, LM UoN, ICV, itacyl, inpho, ETH, CSIRO, IPI light grey: very recent data request

3 Experimental Phase II data # Altitude [m] GSD [m] # strips long/cross % overlap long/cross # Images / Data size [Gb] Additional data ADS Vaihingen/Enz, June 26, 04 low / / / 16.7 GPS/INS high / / / 9.8 GPS/INS Fredrikstad, October 10, 03 low / / high / / UltracamD Fredrikstad, September 16, 04 low / 1 80 / / 30.6 GPS high / / 6.5 GPS Experimental Phase II data and UCD flights Fredrikstad ( + UCD) demanding data sets (image quality affects performance of point measurements) object points ~ control points ~ 25 check points

4 and UCD flights Image quality and point measurements low altitude flight, Oct 10 sun-angle Φ 60deg and UCD flights Image quality and point measurements low altitude flight, Oct 10 sun-angle Φ 60deg

5 Experimental Phase II data ADS flights Vaihingen/Enz (ADS) object points 12 control points >0 check points Phase II Data request & return Phase II Data Set ADS UltracamD # Requests Participant (w/o manufacturer) UoP, DLR-B, DLR-O, ETH, inpho ICC, IPI, Anhalt, inpho, HfT, ETH, UoL, LM UoN, ICV, itacyl, inpho, ETH, CSIRO, IPI light grey: very recent data request Phase II Data Set ADS UltracamD # Results Participant UoP, DLR-B, ETH ICC, IPI, inpho, HfT, Ingr.ZI UoN, itacyl, inpho, CSIRO Status: October 15, 06

6 Remarks on data processing (1/2) typically two different flight heights processed independently only three participants used both heights for common adjustment (UCD (2x) and (2x)) standard and proprietary software packages used Process step Matching and point measurement Bundle adjustment Software Manual, MATCH-AT, LPS, ISAT, Gpro, PhotoMod, others Match-AT, ORIMA, InBlock, BLUH, PhotoMod, ACX-Geotex, IS-PhotoT, others participants typically provided more than one solution, altogether 75 different solutions evaluated and results reported back to participants Remarks on data processing (2/2) Self-calibration was mostly applied for all image blocks but, almost each participant also provided solution w/o use of additional SC 2 participants used modified SC approaches taking the specific image geometry of large format imagery into account Data set UCD ADS Self-calibration parameter set (if applied) 12 Ebner per quadrant, BLUH specific Ebner, Grün, Polynom, BLUH parameters Brown, Grün Brown

7 UltracamD high h g 3800m, GSD 0.34m 0,50 0,45 RMS [m] 0,40 0,35 0,30 0,25 0, GSD dx dy dz 0,15 0,10 0,05 0,00 CSIRO UoN UoN UoN Config. A no no yes IESSG In House Orima Orima Orima manual LPS LPS LPS UltracamD low h g 0m, GSD 0.17m 0,30 0,25 RMS [m] 0, 0,15 0,10 GSD dx dy dz 0,05 0,00 CSIRO CSIRO itacyl UoN UoN UoN Config. B Config. C GPS drift no no no no yes IESSG In House In House Match-AT Orima Orima Orima manual manual Match-AT LPS LPS LPS

8 high h g 00m, GSD 0.15m 0,30 0,25 RMS [m] 0, 0,15 0,10 GSD dx dy dz 0,05 0,00 IPI IPI IPI HfT HfT ICC Ingr.ZI Ingr.ZI 19 semiauto no P12 P12+ Gruen Polynom 12/Quad. no Ebner BLUH BLUH BLUH Match-AT PhotoMod ACX IS Photo-T IS Photo-T LPS LPS LPS Match-AT PhotoMod Match-AT ISAT ISAT high h g 00m, GSD 0.15m P12+, 12 per Quadrant, 19 s Residuals from check point analysis

9 0, RMS [m] combined h g 00m + 950m, GSD 0.15m m 0,15 0,10 dx dy dz 0,05 0,00 IPI IPI IPI IPI IPI IPI IPI inpho inpho inpho inpho inpho inpho no P12 P12A A P12+ P12+ A Brown Ebner no Ebner Gruen no BLUH BLUH BLUH BLUH BLUH BLUH BLUH InBlockInBlockInBlock Ma-AT Ma-AT Ma-AT LPS LPS LPS LPS LPS LPS LPS Ma-AT Ma-AT Ma-AT Ma-AT Ma-AT Ma-AT 0, 0,15 RMS [m] ADS low h g 1500m, GSD 0.m non staggered GSD 0,10 dx dy dz 0,05 0,00 UoP UoP UoP UoP UoP UoP ETH ETH ETH ETH ETH ETH 6 GCP 6 GCP 6 GCP no para self no para self Test1 Test2 Test3 Test4 Test5 Test11 Orima Orima Orima Orima Orima Orima ETH ETH ETH ETH ETH ETH GPRO GPRO GPRO GPRO GPRO GPRO ETH ETH ETH ETH ETH ETH

10 Preliminary conclusions non sufficient number of samples per data set to recommend most-optimal sensor related processing approach and UCD data evaluation are influenced by image point identification errors in measurement (in some cases dominating error source), this also influences the comparison between different processing runs (i.e. applied model of SC) self calibration seems to be necessary to improve object space accuracy for all three camera types in all cases, for data standard parameters seem to be sufficient to compensate for the dominating error sources, although the quadrant specific approach shows very reasonable results 2 participants mentioned, that systematic corrections for UCD are more significant compared to ADS results are almost exceeding accuracy expectations (but note the different quality of data set) Open topics reference solution missing so far will be provided by pilot centre soon influence of errors in image measurement could not be separated from influence of different SC approaches pilot centre will ask individual participants to repeat their calculation by using the same set of measured image coordinates future role of self-calibration? significant determination of additional parameters non trivial and complex task, requires certain block configuration (i.e. overlaps, GCP distribution, GPS and/or GPS/IMU) does AT really have to compensate for weaknesses in geometric digital camera design? are the SC parameters also considered in the further processing chain (e.g. via correction grids)? does the individual user pay for this additional effort?

11 Schedule Final report including participants feed-back till December 06 other data sets from different sensors (i.e. medium format)? new proposal? new project covering new aspects like radiometry, resolution, pansharpening, colour? Who has expertise in that? new data sets required for that, besides already distributed data (those data still should be accessible for future requests) other aspects besides European Digital Airborne Camera Certification?!! Support?