Abstract:The purpose is to measure robot location accuracy of target workpiece by monocular camera based on matching three dimensional (3D) computer aided design (CAD) model. A method is proposed of using the geometrical shape of target which can interface with related information of features in image to searching 3D models data. Using this approach, the target objects could be quickly and efficiently isolated from the background and had efficient recognition ability to target with clear outline and light reflective character. Firstly, the 3D CAD model frame was fit continuously to the boundary of target object in the image by iterative algorithm. Then, the object current pose with position and orientation of 3D CAD model was estimated. Finally, the object grab precision measurement was achieved by the robot hand-eye calibration to determine the relative position between the object and the robot. After rotating the object on xy plane with 0°、45°、90°, the average location error was 3.012 mm, 2.856 mm, and 4.983 mm under three different poses. The experimental results show that with this method of location object, the location error will be increased when the inclined angle increases.