A Laser Range Scanner Designed for Minimum Calibration

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

Scanner Designs 2

Quality Tradeoff • Expense vs. accuracy • Different curves possible • Complex calibration is expensive n esig D r ne A Accuracy Scan sign e D r B nne Sca Expense 3

Conventional stripe scanner • Triangulation between camera and laser 4

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

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

Catadioptric layout 7

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

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

Cylindrical lens precision • No precision mounting 10

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

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

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

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

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

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

Depth resolution 440 mm x 550 mm 240 x 240 pixels Expected depth resolution: 1. 8 mm 17

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