An Improved Method of Contour Extraction of Complex Stripe in 3D Laser Scanning

Abstract

The contour extraction of laser strip is the most critical part of the 3D measurement based on line-structured laser scanning, and its accuracy and robustness directly affects the performance of the system. For the problem that the width of the stripe changes with orientation or reflectivity on the surface, an improved self-adaptive extraction methods is proposed based on Steger's algorithm. In this method, a fixed scale is firstly chosen according to the width of laser beam to convolute the image with Gaussian kernels and get the hessian matrix of every point, then calculate the width along normal direction. The contour of the complex stripes is extracted with sub-pixel accuracy by combing the character that the second derivative takes different extremums for different width under fixed scale space. The validity and superiority of the method are tested separately and the results show excellent performance on extracting the precise contour of complex strips with variable width.