A Coarse-to-Fine Matching Method in the Line Laser Scanning System

Abstract

The measurement accuracy of the traditional line laser scanning method is affected by many factors, such as calibration accuracy of the light plane, the detection accuracy of the stripe center, the positioning accuracy of the motion scanning module, etc., resulting in the operation of this method is complex and the precision is not high. In this paper, a line laser binocular stereo vision system is constructed. Because of the binocular stereo vision, it is not necessary to calibrate the light plane and the angle information of the laser rotation has no effect on the measurement accuracy, which avoids the high precision positioning requirement of the motion control module in the traditional line laser scanning method. In order to increase the precision of the stripe center detection, a coarse-to-fine matching algorithm is proposed. First, the extreme pixel of each row is computed in the left image and right image as the rough extraction point for laser stripes center. Second, the foreground and background are extracted and the searching range is limited in a small zone. Third, sub-pixel matching is adopted to refine the laser stripe center correspondence. The proposed method is confirmed in the simulation and experiment. The simulation results show that the RMS of laser stripe center error is less than 0.03 pixel, when the image noise level is less than 7. The three-dimensional measurement experiment shows that when the measurement distance is 1400 mm, the reconstruction precision is as high as 0.314 mm.