This paper proposes a feasible program of 3D curve surface reconstruction based on the method of binocular stereo vision system which is composed of two industrial CCD cameras. On the basis of bipolar linear constraints, the method gets information modules to match the 3D curve surface feature at sub-pixel level by the sift algorithm so that the reconstruction of 3D stereo structure can be realized. Then, through the combination of the acquired feature vectors with 2D image to affine the curve surface, the feature extraction and the recognition progress can be realized by iteration. Finally, a 3D depth map can be produced. This method can help improve the robustness of the stereo matching and get a better 3D reconstruction in a simple way with low cost.
In the position-based visual servoing control for robot, the hand-eye calibration is very important because it can affect the control precision of the system. According to the robot with eye-to-hand stereovision system, this paper proposes a direct method of hand-eye calibration. The method utilizes the triangle measuring principle to solve the coordinates in the camera coordinate system of scene point. It calculates the estimated coordinates by the hand-eye calibration equation set which indicates the transformational relation from the robot to the camera coordinate system, and then uses the error of actual and estimated coordinates to establish the objective function. Finally the method substitutes the parameters into the function repeatedly until it converged to optimize the result. The related experiment compared the measured coordinates with the actual coordinates, shows the efficiency and the precision of it.