Monthly Technical Report November 1998 SRI Project Number: ECU-1515 Contract Number: NMA202-97-C-1004 DARPA Order Number: E645 Prepared by: Aaron J. Heller, Sr. Computer Scientist Prepared for: Michele Motsko, Physical Scientist National Imagery and Mapping Agency Reston, VA SUMMARY OF ACTIVITIES: 1. Preparation for November demos. 2. Testing Dynamic Snakes algorithm on NTM 3. SEDRIS export 4. Reading DXF attributes DETAILED DESCRIPTION: 1. Preparation for November demos. (Heller,Connolly) Preparation for the November demos dominated our activities for November. The bulk of the work was running extractions over large parts of the NTC DPPDB data. We have not yet evaluated these results against the DTOP reference. That work will be performed in December. We did, however, evaluate the automatic result against the edited result, to determine the magnitude of the editing job. The following summarizes the extraction experiments: TIMING STATISTICS, AUTOMATIC RESULTS Full frame NTC, light roads: Low level (1/4 resolution): Elapsed real time: 61500.0 sec. = 17 hrs. 5 min. High level (602LF/RF full resolution): Elapsed real time: 137024.54 sec. = 1 day 14 hrs. 3 min. 44.54 sec. Total run time: 92140.22 sec. = 1 day 1 hr. 35 min. 40.22 sec. User run time: 91385.74 sec. = 1 day 1 hr. 23 min. 5.74 sec. System run time: 754.48 sec. Full frame NTC, dark roads: Low level (1/4 resolution): Elapsed real time: 17100 sec. = 4 hrs. 45 min. High level (602LF/RF full): ESTIMATED Elapsed real time: 39855.55 sec. = 11 hrs. 4 min. 15 sec. Cantonment Area, Dark Roads (to be used in the road editor): Low-level on 1/4 resolution, standard parameter settings: Elapsed real time: 6360.4 sec. = 1 hr. 46 min. 0.40 sec. Total run time: 2506.13 sec. = 41 min. 46.13 sec. User run time: 2487.21 sec. = 41 min. 27.21 sec. System run time: 18.92 sec. High-level, 2 images: Elapsed real time: 15465.11 sec. = 4 hrs. 17 min. 45.11 sec. Total run time: 5149.11 sec. = 1 hr. 25 min. 49.11 sec. User run time: 5076.45 sec. = 1 hr. 24 min. 36.45 sec. System run time: 72.66 sec. * * * * * * * TIMING STATISTICS, EDITED RESULTS Road editing on Cantonment Area, Dark Roads: Phase Time Required Feature Set -------------------------------------------------------------- Initial deletions: 6 minutes "Deletions I" Editor Deletions: ~6 minutes "Deletions II" Editor Additions: 120 minutes "Additions II" ------------ Total elapsed time: 132 minutes A significant portion of this time (approximately 30-40 minutes) was spent waiting for MBO functions, but was not deducted from the running time. Therefore, we estimate the actual operator interaction time was on the order of 90-100 minutes. * * * * * * * SIZE STATISTICS, EDITED RESULTS Arc length for automatic results: 66.60 km for false positives: 13.28 km for missed roads: 11.20 km for final result: 64.52 km With respect to the edited final result, the automatic extraction achieved 83% completeness and 80% correctness. Arc length for full frame light roads: 1,105,114 m = 1105km for full frame dark roads: 133,141 m = 133 km 2. Testing of Dynamic Snakes algorithm on NTM (Connolly) The latest version of the dynamic snakes algorithm was moved to our classified facility for testing on DPPDB, mosaic, and other NTM imagery. Two modifications were made to improve the responsiveness of the algorithm: - The computation of the cost function is limited to the current visible region of the image. This is based on the observation that the operator does not expect to be able to extract part of a road that are not visible in the current window. - If, while propagating the costs to a new part of the image, the mouse is moved back over an area where the lowest cost path is known, the cost computation is stopped and the algorithm goes back into tracking mode. This prevents long delays resulting from accidental movement of the mouse over regions with no roads. The algorithm was found to be effective on all the imagery in the test. 3. SEDRIS export (Iverson) Work was completed on an initial version of a SEDRIS export capability. The SEDRIS libraries were compiled and linked into the RCDE/BOS system. The DEPTH and CHECKER programs were modified to be interactively invoked from within RCDE/BOS. Foreign function interface declarations were created for the top-level SEDRIS entry points and callbacks in the RCDE/BOS system. Currently, the interface can export RCDE/BOS planar-faceted objects (e.g., buildings) and ribbon-curve objects (e.g., roads, fences). The user selects the site and feature sets to be translated using RCDE menu facilities and initiates the instantiation of a SEDRIS data model. This can be textually reviewed with DEPTH and CHECKER and a SEDRIS transmittal produced. The transmittal can be graphically inspected using the SEDRIS model viewer. 4. Reading DXF attributes (Wolf) GDE delivered the extraction work done on Ft. Benning by the NV Office. These are in DXF format with attributes. We extended our DXF reader to handle the attributes. --------------------------------------------------------------------------- Automatic Population of Geospatial Databases Monthly Report to SRI for November Bob Wilson Vexcel Corporation 9 December 1998 1. MAJOR TECHNICAL ACCOMPLISHMENTS 1.1 Extraction of Buildings from Dense DEMs Chris Lesher continued to work on software to refine and "clean up" building edges extracted from dense DEMs. For test data he has been using DEMs of downtown Denver and Ft. Hood. The software now supports computing straight edges, squaring off corners, and computing local preferred orientations from road/street networks and averaging building edges. I presented some of the preliminary results at the September '98 IPR. 1.2 Image Understanding Workshop Bob attended the IUW in Monterey. This provided an opportunity to meet with SRI and GDE to begin work on a final IFD for APGD. Bob discussed (with both government sponsors abd FREs) possible avenues for continuing Vexcel's current APGD research directions while awaiting DARPA's anticipated IU BAA. 1.3 Aerotriangulation of MIT Photos The original scanned color imagery of MIT was provided on tape to Seth Teller at the IUW. Planes were made to acquire and process a data set with roughly 1000 images. (This work would not be funded by our subcontract with SRI.) 1.4 Incipient Collaboration with UMass on Building Extraction from IFSAR. Bob discussed with Allen Hanson and Ed Reisman possible collaboration between UMass and Vexcel in radargrammetric building extraction. It was agreed that new funding was required in order to support such work. 2. ACCOMPLISHMENTS VIS-A-VIS STATEMENT OF WORK 2.1 Refine the BOS architecture 2.2 Survey automated model extraction techniques 2.3 Develop feature extraction and consistency enforcement algorithms 1.1 2.4 Refine the design of and implement the dynamic database 2.5 Produce, maintain, and distribute data sets and ground truth 1.3 2.6 Develop evaluation metrics and perform evaluations 2.7 Interface to FRE contractors 1.2, 1.3, and 1.4 2.8 Develop and perform demonstrations 1.2 2.9 Transfer technology 2.10 APGD program management 1.2, weekly conference calls, monthly report, and phone conversations. --------------------------------------------------------------------------- GDE Systems APGD Monthly Report December 10, 1998 Technical Section Summary Work during this period has emphasized data set generation and planning for the final demo in January. Detailed Work Description (by SOW item): 1. Architecture Refinement No activity 2. Algorithm Survey No activity. 3. Algorithm Development No activity. 4. Multi-Sensor Registration No activity 5. Persistent Store. No activity 6. Dataset Production & Distribution: Feature extraction work on the main parts of Fort Hood will be completed tomorrow (December 11) at the NVO. Work on the residential areas is on hold pending receipt of further funding TEC headers for new Fort Benning imagery are nearing completion. A few minor discrepancies remain to be straightened out in discussions with Max Roos. 7. Evaluations: No activity 8. Interface to FRE Contractors: Agreement in principal with USC to include the latest version of their algorithm in the final demo has been reached. 9. Demonstration Scenarios: Plans for the building extraction portion of the final APGD/IFD demo have been roughed out, but their implementation will depend on another increment of funding. 10. Technology Transfer No activity 11. Option Years No activity 12. Program Management We are continuing to work according to the priorities agreed on with the prime. We keep in close contact with team members via weekly conference calls and e-mail communications as appropriate.