Depth Camera-based Algorithm for Visual Recognition and Automated Polishing of Ship Components

Intelligent polishing equipment comprising industrial robots and sensors is increasingly used to polish ship components, which has traditionally been done manually The core component of such equipment is the visual recognition system, which is responsible for workpiece recognition and positioning, and polishing point recognition and extraction. Current systems are mostly based on CCD or CMOS cameras with 2D image processing. However, recognition efficacy is limited by the auxiliary lighting conditions and image resolution, and depth information cannot be obtained. An algorithm module is designed for object recognition in point cloud from a depth camera in a device used to polish small ship components. The module's architecture and the functions of the visual recognition system are described. The module's functions are based on the OpenCV and printer control language (PCL) libraries and are implemented in the C++ programming language. An algorithm test platform was built to verify the algorithm module its functions, including camera pose estimation and coordinate transformation, point cloud segmentation, centroid extraction, and edge extraction. The limitations of on-site auxiliary light conditions and image resolution were overcome. A workpiece recognition and positioning success rate of 98% was achieved, and the polishing point detection and recognition accuracy are sufficient for real applications. This allows accurate and comprehensive visual data to be provided to the automated polishing device, ensuring smooth operation and effective workpiece handling and polishing.