Recognizing and locating of objects using binocular vision system

摘要

This paper examines how to recognize the dynamic object and find its position using a real-time binocular visual image system, which describes contours of the object using edge detection algorithm. The system finds the object of image with shape matched method, which procedures include color segmentation of input image, finding the candidate area of matching color histogram, binary image morphologic management, convex hull process, and finding the object and its contours. Image recognition technique is used to find the features of the object on the image plane. Intrinsic and extrmsic parameters of the features are captured using calibration board to form a fundamental matrix for analysis that increases the accuracy of corresponding points. Finally, it performs the three-dimensional reconstruction to combine and associate the feature messages with camera parameters. We can hence obtain three-dimensional coordinate information of the object with reference to the camera system. Experimental results demonstrated that the high achievable accuracy and stability of position coordinates reconstructed from the features and three dimensional coordinates of the object with the camera system's real-time binocular visual subsystem that provides three-dimensional position coordinate for the robotic arms of an intelligent robot.