A Laser Range Scanner Designed for Minimum Calibration Complexity James Davis, Xing Chen Stanford Computer Graphics Laboratory 3D Digital Imaging and Modeling 3DIM 2001

2 Scanner Designs

3 Scanner Design A Quality Tradeoff Expense vs. accuracy Expense vs. accuracy Different curves possible Different curves possible Complex calibration is expensive Complex calibration is expensive Expense Accuracy Scanner Design B

4 Conventional stripe scanner Triangulation between camera and laser Triangulation between camera and laser

5 Complexities of traditional design Actuated components Actuated components Cylindrical lens precision Cylindrical lens precision Custom calibration procedure Custom calibration procedure

6 Our design Triangulation between two cameras Triangulation between two cameras No actuated components No actuated components

7 Catadioptric layout

8 Our scanner Simple components Simple components Camcorder, four mirrors, rigid mounting

9 Stripe processing Locate corresponding points Locate corresponding points Use epipolar constraint Discard ambiguous data

10 Cylindrical lens precision No precision mounting No precision mounting

11 Scanner calibration Camera model and pose [Heikkila, Silven 97] Camera model and pose [Heikkila, Silven 97] Well-studied easy calibration Well-studied easy calibration

12 Filter image Filter image Video signal noise Sub-pixel detection Sub-pixel detection Local Gaussian approximation Peak detection Maximum intensityLocal GaussianGaussian with filtering

13 Mesh coverage Video sequence defines mesh Video sequence defines mesh Stripe spacing related to laser velocity Stripe spacing related to laser velocity

14 Constructing a mesh Fill stripe sampling gaps Fill stripe sampling gaps Detect depth discontinuities Detect depth discontinuities

15 Aligning scans Iterative closest point (ICP) [Besl, McKay - Chen, Medioni 92] Iterative closest point (ICP) [Besl, McKay - Chen, Medioni 92] Global alignment [Pulli 99] Global alignment [Pulli 99]

16 Merging scans Volumetric merging (VRIP) [Curless, Levoy 96] Volumetric merging (VRIP) [Curless, Levoy 96] Hole filling Hole filling

17 Depth resolution 440 mm x 550 mm 240x240 pixels Expected depth resolution: 1.8 mm

18 Conclusion Minimal calibration complexity Minimal calibration complexity No actuated components No precision lens placement Well-studied easy calibration model