The EuroSDR Performance Test for Digital Aerial Camera Systems

Transcription

1 Institut für Photogrammetrie Digital Airborne Camera Performance The DGPF test Overview and results Michael Cramer Photogrammetric Week 2009 Stuttgart September 8, 2009 Institut für Photogrammetrie The EuroSDR Performance Test for Digital Aerial Camera Systems EuroSDR network on Digital Camera Calibration and Validation Michael Cramer 51st Photogrammetric Week September 3 7, 2007 Universität Stuttgart

2 Empirical Evaluation Tests EuroSDR (2007) vs. DGPF (2009) EuroSDR test DGPF Acquisition 10/2003, 06/2004, July September 2008 Time 09/2004 Test site Vaihingen/Enz (1x) Vaihingen/Enz Fredrikstad (2x) Sensor ADS40 (1st), DMC, ADS40 (2nd), DMC, systems Ultracam-D Ultracam-X, JAS-150, Quattro DigiCAM, AIC-x1, DLR 3K, RMK-Top, ALS50, AISA, ROSIS, Spectrometer Research Geometry Geometry, Radiometry, DSM topics Generation, Stereoplotting Active 13 groups/institutions 30 groups/institutions participants (European) (German speaking countries) Institut für Photogrammetrie Digital Airborne Camera Performance The DGPF test Overview and results The DGPF test design

3 The DGPF-Test Objectives & Organization User driven independent and objective evaluation of commercially available digital photogrammetric camera systems (focus on airborne and large format) Follow-up of already done national or international evaluation projects (like EuroSDR network Digital Camera Calibration) Evaluation of sensor specific strengths and potential weaknesses, no direct comparison between performance of different sensors! Processing teams focussing on different topics Team 1: Geometric accuracy and resolution ( K. Jacobsen) Team 2: Radiometric accuracy ( M. von Schönermark) Team 3: Automatic ti DSM genereation ( N. Haala) Team 4: Stereoplotting ( V. Spreckels) Test site Vaihingen/Enz Institut t für Photogrammetrie t () GSD 20cm GSD 8cm

4 Reference data Test field layout Ground references Spectrometer measurements in field, sunphotometer BRDF measurements Manual, on-site land use classifications (multiple days) DGPF evaluation test Vaihingen/Enz test t flight data System System provider Flyer Date of flight(s) DMC Intergraph/ZI RWE Power ADS 40, SH52 Leica Geosyst. Leica Geosyst JAS-150 JenaOptronik RWE Power Ultracam-X Vexcel Imaging bsf Swissphoto RMK-Top15 Zeiss RWE Power quattro DigiCAM, icam 4-Head IGI Geoplana AIC-x1, 1-Head Rolleimetric Alpha Luftbild AIC-x4, 4(3)-Head Rolleimetric Vulcan Air DLR 3K-Camera DLR Munich DLR Munich AISA+ hyper-spectral (parallel with DMC) Specim/FH Anhalt RWE Power ROSIS hyper-spectral DLR Munich DLR Munich ALS 50 LiDAR Leica Geosyst. Leica Geosyst

5 Das Bild kann nicht angezeigt werden. Dieser Computer verfügt möglicherweise über zu wenig Arbeitsspeicher, um das Bild zu öffnen, oder das Bild ist beschädigt. Starten Sie den Computer neu, und öffnen Sie dann erneut die Datei. Wenn weiterhin das rote x angezeigt wird, müssen Sie das Bild möglicherweise löschen und dann erneut einfügen. Das Bild kann nicht angezeigt werden. Dieser Computer verfügt möglicherweise über zu wenig Arbeitsspeicher, um das Bild zu öffnen, oder das Bild ist beschädigt. Starten Sie den Computer neu, und öffnen Sie dann erneut die Datei. Wenn weiterhin das rote x angezeigt wird, müssen Das Sie Bild das Bild kann möglicherweise nicht angezeigt löschen werden. und Dieser dann Computer erneut einfügen. verfügt möglicherweise über zu wenig Arbeitsspeicher, um das Bild zu öffnen, oder das Bild ist beschädigt. Starten Sie den Computer neu, und öffnen Sie dann erneut die Datei. Wenn weiterhin das rote x angezeigt wird, müssen Sie das Bild möglicherweise löschen und dann erneut einfügen. GSD 20 0cm (n nom.) Block design DMC & UCX GSD 20cm / GSD 8cm DMC hg 2160m, 60/60 UCX hg 2900m, 70/70 GS m (nom.) D 8cm DMC hg 870m, 60/63 UCX hg 1200m, 80/70 Sunphotometer & WebCam Day of flight 6. August :00h 09:00h 10:00h 11:00h 12:00h 13:00h art.de/dgpf Aerosol optical 534nm DMC ADS40 SH52 quattro DigiCAM Time [UTC]

6 Institut für Photogrammetrie External geometrical Accuracy analysis Digital Airborne Camera Performance The DGPF test Overview and results Results from geometric performance analysis Impact of Self Calibration GSD 20cm Frame Sensors Block RMK Uni Hann. no SC 14 / 2160m 60 / 70 RMK Uni Hann. 14 / 2160m 12 BLUH par. 60 / 70 DMC Uni Hann. 9 / 2160m no SC 60 / 60 DMC Uni Hann. 9 / 2160m 12par.+2 DMC 60 / 60 UCX Uni Hann. 9 / 2900m no SC 70 / 70 UCX Uni Hann. 9 / 2900m 12par.+32 UC 70 / 70 #GCP / h RMS ChP [m] g GCP / ChP p/q[%] XX YY ZZ 14 / 82 0,11 0,12 0,15 14 / 82 0,12 0,13 0,08 9 / 95 0,05 0,08 0,19 9/ , , ,06 9 / 99 0,08 0,08 0,19 9 / 99 0,08 0,08 0,08 BLUH results from IPI, Uni Hannover RMS from ChP [m]

7 Block Impact of Self Calibration GSD 20cm Frame Sensors Quattro DigiCAM [m] 0,33 3m Height h RMS ChP [m] g GCP / ChP p/q[%] XX YY ZZ 0,27 7m RMK Uni Hann. no SC 2160m 60 / 70 RMK Uni Hann. 2160m 12 BLUH par. 60 / 70 DMC Uni Hann. 2160m no SC 60 / 60 DMC Uni Hann. 2160m 12par. + 2 DMC 60 / 60 UCX Uni Hann. 2900m no SC 70 / 70 UCX Uni Hann. 2900m 12par UC 70 / / 82 0,11 0,12 0,15 14 / 82 0,12 0,13 0,08 9 / 95 0,05 0,08 0,19 9/ , , ,06 9 / 99 0,08 0,08 0,19 9 / 99 0,08 0,08 0,08 RMS from ChP [m] RMS from ChP [m] Absolute Object Point Accuracy GSD 20cm Frame Sensors Block #GCP / h RMS ChP [m] g GCP / ChP p/q[%] X Y Z DMC Uni Hann. 9 / 2160m 60 photos 60 / 60 9 / 95 0,04 0,06 0,06 DMC Uni Stuttg. 4 / 2160m 42 photos 60 / 60 4 / 180 0,04 0,07 0,11 UCX Uni Hann. 9 / 2900m 52 photos 70 / 70 9 / 99 0,08 0,08 0,08 UCX Uni Stuttg. 4 / 2900m 36 photos 70 / 70 4 / , , ,15 IGI Uni Stuttg. 4 / 2500m 132 photos 60 / 70 4 / 161 0,07 0,10 0,27 RMK Uni Hann. 14 / 2160m 47 photos 60 / / 82 0,12 0,13 0,08 self-calibrating AT using Gruen () and BLUH sensor models (Uni Hannover) no GPS/IMU data used by Uni Hannover but by Uni Stuttgart Uni Stuttgart blocks without using photos from cross-strips RMS from ChP [m]

8 Absolute Object Point Accuracy GSD 8cm Frame Sensors Block DMC Uni Hann. 135 photos 9 / 1200m 60 / 63 DMC Uni Stuttg. 4 / 1200m 110 photos 60 / 63 UCX Uni Hann. 9 / 1200m 215 photos 80 / 70 UCX Uni Stuttg. 4 / 1200m 175 photos 80 / 70 IGI Uni Stuttg. 4 / 1060m 640 photos 80 / 70 RMK Uni Hann. 870m 74 photos 60 / 70 #GCP / h RMS ChP [m] g GCP / ChP p/q[%] X Y Z 9 / 45 0,03 0,03 0,04 4 / 113 0,03 0,04 0,05 9 / 99 0,04 0,07 0,04 4 / , , ,04 4 / 114 0,04 0,07 0,08 14 / 40 0,02 0,04 0,05 self-calibrating AT using Gruen () and BLUH sensor models (Uni Hannover) no GPS/IMU data used at Uni Stuttgart, but introduced in TU Graz processing Uni Stuttgart blocks without using photos from cross-strips RMS from ChP [m] Block ADS Uni Hann. GSD 8cm JAS RAG Herne GSD 8cm JAS RAG Herne GSD 8cm Absolute Object Point Accuracy Line Scanners # GCP / Height h g GCP / ChP RMS ChP [m] GSD 20cm X Y Z 9 / 1900m 9 / 52 0,02 0,02 0,04 0 / 1800m 0 / >50 0,05 0,02 0,11 GSD 8cm 4 / 1800m 4 / >50 0,02 0,02 0,05 JAS RAG Herne GSD 8cm 19 / 1800m 19 / > , , ,04 JAS RAG Herne GSD 20cm 4 / 4600m 4 / >50 0,02 0,02 0,07 JAS RAG Herne GSD 20cm 19 / 4600m 19 / >50 0,02 0,02 0,05 ADS processing by Uni Hannover, JAS-150 done by RAG Herne self-calibrating AT used in both cases RMS from ChP [m]

9 Institut für Photogrammetrie External geometrical Accuracy analysis Digital Airborne Camera Performance The DGPF test Overview and results Some results from other processing teams Sportsfield Analysis of 3D point clouds Sports field (images at GSD 8cm) DSM accuracy DMC Std.Dev.= 5.2cm, Density= 19.7 Pts/m 2 DSM accuracy RMK Std.Dev. = 17.2cm, Density = 0.8 Pts/m 2 Laserscanner (ALS50) Reference Std.Dev.= 1.9cm, Density = 8.3 Pts/m 2 DMC RMK ALS50

10 Automatic DSM generation (buildings) Results from ETH Zürich art.de/dgpf ALS50 reference DSM 25cm grid DMC GSD 8cm 6-folded overlap UCX GSD 8cm 10-folded overlap DSM obtained from SAT-PP Software (ETHZ) Test site industrial building K. Wolff (ETH Zürich) Automatic land-use classification art.de/dgpf Results from DMC (GSD 20cm) S. Klonus (IGF, Universität Osnabrück)

11 Institut für Photogrammetrie External geometrical Accuracy analysis Digital Airborne Camera Performance The DGPF test Overview and results Summary Conclusions (1/2) (geometric) performance of different sensors quite close Abs. Accuracy (RMS) of object point determination realistic control/block configuration ~½ pix GSD results underline the high potential of digital image recording Accuracy potential already close to the accuracy of reference data (i.e. control points, LiDAR data) Self-calibration (to certain extend) is necessary some differences in accuracy are present, but question y p q whether they of relevance for later practical operation or only academic discussion?

12 Conclusions (2/2) Other factors gain in importance (environmental conditions, choice of reference data) and may have larger impact on final accuracy than choice of sensor system itself Sensor and sensor related software chain is tightly coupled and has to be considered All results rely on (one single) DGPF data set only transfer of results to later production environments? Acknowledgments system suppliers, mainly Leica Geosystems, Intergraph / ZI- Imaging, Vexcel Imaging, Jenaoptronic, IGI, Rolleimetric data providers & flying companies DGPF project team leaders, namely, K. Jacobsen, M. v. Schönermark, N. Haala & V. Spreckels. all supporting project partners and participants, represented by H. Krauß, K. Komp, E. Jordan, R. Wahmke, C. Hagedorn, K. Jacobsen, F. Rottensteiner, W. Stößel, S. Holzwarth, F. Kurz, M. v. Schönermark, V. Kaufmann, R. Ladstädter, M. Kähler, M. Breuer, C. Gläßer, A. Jung, S. Baltrusch, E. Wild, A. Grün, M. Baltsavias, A. Busch, F. Rückert, H. Hastedt, R. Schneider, V. Spreckels, T. Luhmann, P. Grussenmeyer, K. Clausen, N. Pfeiffer, T. Kersten, A. Fricker, W. Förstner, M. Gültlinger and others

13 Acknowledgments Werner Schneider Institut für Photogrammetrie Institut für Photogrammetrie External geometrical Accuracy analysis Digital Airborne Camera Performance The DGPF test Overview and results DGPF Project Meeting October 5-6, 2009 Universität Stuttgart see