Object detection and recognition method based on binocular

Abstract

This paper proposes a method based on 3D version to recognize free-form objects in the complicated environment, which could be used in field of robot picking. The point cloud data is generated by binocular camera system. The depth map is employed to recover 3D point cloud of the scenario with the calibration method from the binocular camera. Segmentation algorithm is used to detect the object. Recognition is performed by using software libraries integrated with custom-developed segmentation algorithm and model database created by the same binocular camera system. Experiments are designed to verify the performance of the method by randomly placing different types of experimental objects in manipulator workspace. The preliminary results demonstrate the excellent ability of the system to perform object recognition and picking.