Processing Mobile Scanner Data - Calibration to Improve Positioning

1. Processing Mobile Scanner Data- Calibration to Improve Positioning

4. Lynx (Optech)

5. Airborne Positioning Fairly uniform satellite visibility Fairly uniform positional accuracy

6. Mobile Positioning Satellite visibility varies Positional accuracy varies

7. Sources of Positional Inaccuracy System calibration Scanner – IMU misalignment angles Other parameters Range measurement (point-to-point noise)? Trajectory solution xyz (GPS)? Trajectory solution hrp (IMU)? Same sources as with airborne data sets

8. Magnitude of Error Sources System calibration Relatively small in project data sets if system has been carefully calibrated Range measurement Small – 1 cm level Trajectory xyz ? ? ? ? Trajectory hrp ? ? ? ?

9. Example Trajectory RMSApplanix smrmsg_xxxx.out Xy Z Heading Roll/pitch

10. Magnitude of Error Sources System calibration Relatively small in project data sets if system has been carefully calibrated Range measurement Small – 1 cm level Trajectory xyz Dominant error source Trajectory hrp Small when short ranges involved

11. Gradual Variation Normally error in trajectory xyz does not change rapidly Trajectory drifts off in one direction and stays for a number of seconds Good potential for fixing

12. General Project Strategy Collect some control measurements Execute LIDAR / image survey Drive every place twice Check system calibration Find bad positioning Display trajectory colored by RMS values (smrmsg_xxx.out) Compare drive passes together Collect control measurements at bad locations Adjust xyz of drive passes together and to control measurements with a fluctuating correction Correction curve which changes over time Remove less accurate data Long range measurement if short range exists

13. Tie lines Matching based on intensity or xyz features in LIDAR Matching based on linear features in images

14. Tie line types Ground point – point feature on ground, seen by multiple passes Xy point – xy point feature, multiple lines Known point – known xyz point on ground Ground line – linear feature on ground, multiple lines Section line – xyz line on terrain slope, roof or wall, multiple lines Known line – known xyz point on ground, one or multiple lines, line runs thru known point

15. Ground point tie line type Point feature on ground Seen multiple times 1. User enters approximate xy position Software find passes which see the location 2. User enters xy position of each observation Software computes z from fitted plane equation

16. Xy point tie line type Xy point feature Example: building corner Seen multiple times 1. User enters approximate xy position Software find passes which see the location 2. User enters xy position of each observation

17. Known point tie line type Point feature on ground Known xyz Seen multiple times 1. User enters known xyz position Software find passes which see the location 2. User enters xy position of each observation Software computes z from fitted plane equation

18. Ground line type Linear feature on ground Seen multiple times Manual Can do anything user sees 1. User enters approx start and end xy 2. User enters start and end xy in each pass Software computes z from plane equation Auto line search Requires bright line on darker background 1. User enters approx start and end xy

19. Section line type Xyz line on building wall or ground Seen multiple times 1. User enters section right and left point Software finds passes 2. User enters start and end xyz in each pass

20. Known line type Known xyz point on ground Seen once or multiple times Line runs thru known point Manual Can do anything user sees 1. User enters known xyz 2. User enters approx start and end xy 3. User enters start and end xy in each pass Auto line search Requires bright line on darker background 1. User enters known xyz 2. User enters approx start and end xy

21. Tie lines Line start and end points do not have to match Correction is perpendicular to each line Correction pulls lines 'together'

23. Heading error Mismatch visible only at longer distance from scanner Mismatch visible in walls -- solvable only with right pattern Mismatch visible in point features (poles, building corners)?

24. Roll error Mismatch visible only at longer distance from scanner Mismatch visible as elevation difference in ground Mismatch visible as leaning building walls

25. Pitch error Mismatch visible only at longer distance from scanner Mismatch visible as elevation difference in ground Mismatch visible as leaning building walls

26. Calibration workflow Drive suitable site with buildings: Good visibility to walls Clean, vertical walls Right driving pattern Search tie lines on building walls automatically Solve HRP misalignment angles

27. Lynx calibration pattern 1 Intersection with at least two buildings

28. Lynx calibration pattern 2 Large building with open area (parking lot)?

29. Usable Objects

30. Terrasolid European Users EventVienna 17 – 20 February 2009