Poses And Motion Parameters Analysis Of Rigid Object Based On Binocular Vision

Visual monitoring of motion object is a research focus on robot worldwide at present, and the study, which has scientific significance and practical value at aspects of improving robot motion precision and enhancing robot intelligence. At present, more achievements according to directly monitoring for object point. Whereas, considering some special scenes, such as spark or chip will generate in tool carving, tool maybe enter into body in medical surgery. In more practical application, many similar situations make directly monitoring operation point become difficult. Under this condition, this paper analyzes and compares synthetically the research status of visual monitoring of motion object at home and abroad, and proposes a new approach which indirectly analyzes poses and motion parameters by extracting visual information of target mounted on object.Firstly, this paper describes the mechanism characteristics of binocular monitoring platform, and builds conformable mathematical model for it; This paper also explains camera visual model and its calibrated method.Secondly, this paper designs truncated cone-shape target mounted on tool according to tool characteristics of parallel carving robot. And proposes a X-corner sub pixel extracting approach from target images and builds three dimensional reconstruction model of target points.Thirdly, this paper study least square estimation algorithm of poses of rigid motion object at sampling time by using geometric invariance of object and truncated cone-shape target mounted and motion constraint of rigid displacement. Then this paper builds motion estimation model based on Extended Kalman Filter by using continuous image sequence information.Finally, this paper completes the algorithm above mentioned by Matlab program. And verifies the feasibility of indirectly monitoring poses and motion parameters of motion object by using visual information of truncated cone-shape target mounted.