The Binocular Vision Image Recognition And Position System Research Based On The Features Of Sift Matching Algorithms

Machine vision and image recognition is one of the key technologies to improve the intelligent level of industrial robot. Binocular vision is available in various conditions of obtain the three-dimensional information flexibly. The purpose of binocular vision is realizing 3d scene perception, recognition and understanding. The stereo matching is one of the key technology, and how to select the proper matching primitives and matching algorithm is difficult and hot. With the binocular stereo vision obtain the image information, this paper using three-dimensional feature matching algorithms based on the Scale (Invariant SIFT Transform) to identify targets and determine the position of the workpiece, controlling the machine to grab the goals by hand, and research and development the hardware system of test piece.(1) Using a calibration method of improved Tsai which considering the radial distortion in the MATLAB software environment, and calibrating the cameras under the condition of binocular stereo system. This calibration method is precision and requirements.(2) Using the newest research results in image processing-SIFT, and see SIFT characteristics as matching primitives, treated SIFT feature descriptor as feature descriptor of area-based matching. Then reduced the search space of matching features from 2D to 1D according to epipolar constraint in stereo vision theory. The matching result is finally obtained in the light of the nearest neighbor matching based on Euclidean distance of SIFT descriptors. Experimental results show that this method can well solve the matching problems under the circumstances of the workpiece image is scaling, rotating, translation, shelter, noise and so on, and the processing speed more improved than the original algorithm.(3) Using the template matching principle to identify targets and locate the target with the SIFT matching results, and then calculating the three-dimensional position of workpiece. Sending position of the workpiece to robot control system, and calculating the world coordinates to each robot joints' angle by the inverse kinematics, then drive robot to complete the workpiece grabing.The matching algorithm in this paper can accurately identify targets and robustness by processing lots of images. The test system of robot grab can effectively link the process of identify and locate together, and strengthen the practicability of matching algorithms.