Binocular Stereo Measurement Based On Structured Light

Based on structured light binocular stereo measurement technology is a branch of computer visual technology, which is widely used in virtual realization of 3D digital modeling, detection technology of virtual realize, military simulation, robot guidance, topography measurement, orthopedic medical, identification of criminal investigation and other multiple fields. How to quickly and accurately obtain the surface information of an object in research subject is one of the most important and difficult.This paper build a binocular stereo vision measurement system based on structured light. We use two camera models specifications exactly the same shooting calibration board in different positions, using thecalibration method of Zhengyou Zhangis used to calculate the camera internal, external parameters. Active encoding structured light combination selection of gray code and line shift method, 22 sets of structure light pictures through the projector were projected to the measured surface, and then these pictures which were collected at the same time were decoded. About camera calibration results using fundamental matrix and epipolar equation, the solution of epipolar equation and the decoding data which is storaged solution in a custom data structure were combined for the measured world coordinate of each point on the surface. The the 3D point cloud information was finally formed, and reconstruction of measured object 3D images and the corresponding measurement information were showed in the paper.The only code matching algorithm in the right and left sides of the image matching corresponding point is very low, this paper presents the line matching algorithm. This algorithm uses code value matching and curve search method to solve the problem that the code value is disturbed and can not be matched accurately, which greatly improves the matching rate and 3D reconstruction accuracy.At last, the paper reconstructs the measured three-dimensional image of the object surface, through with standard gauge blocks compared to the error of 0.05 mm. From the pixel compared two pixel error.