A Fast Structured Light Matching Method in Robot Stereo Vision

Abstract

3D surface reconstruction has been widely used in reverse engineering, remanufacturing and automatic measurement. Binocular stereo imaging technology is the main method to obtain 3D information. Images are acquired from different positions of the object. The 3D coordinates of the point can be calculated according to the point’s coordinates in left and right cameras. The key to ensure the accuracy of 3D reconstruction is high precision and fast matching method of two images. As the image feature was not obvious, structured light is used. For its easy identification and extraction, the method has been more and more widely used. But the former research cannot ensure the matching exactness or matching algorithm takes too much time. In order to improve the precision and efficiency of 3D measurement, a fast and simple structured light matching method is designed based on epipolar geometry. This method can ensure 100% matching accuracy. It does not need other assistant tool except for structured light projector. On the beginning of the measurement, structured light is projected onto the object to be measured. Two images of the object with structured light are captured by two cameras in different positions. Structured lights extracted from the two images are matched by the new algorithm. The 3D coordinate of this structured light is calculated. The position of structured light is changed on the object till completed. From a lot of experiments, the proposed grating matching method is proved and it is a technique with high precision, low costs, easy operation, and an automatically matching method. Furthermore, it can be widely used in most of 3D reconstruction system.