System for Identifying Target Objects to Perform Manipulative Operations by Unmanned Underwater Vehicles

The paper proposes a system for identifying target objects to perform manipulative operations by unmanned underwater vehicles (UUV) equipped with multi-link manipulators (MM). This system processes point clouds received from technical vision systems (TVS) mounted on UUV, and provides high-precision identification of the location and shape of target objects relative to the UUV. Then, the specified trajectories of the MM working tool are translated on the determined surface of the target object, taking into account the possible silting, fouling and deformation of this object. The developed system is implemented in the C++ programming language on the NVIDIA Jetson TX2 on-board computer. The results of recognition of underwater objects using a Zed2 stereo camera placed in a sealed box, calibrated in an aquatic environment using the OpenCV library method, are presented. The experimental data obtained were used to simulate the process of manipulation operation performed by the manipulator of the UUV in the software packages CoppeliaSim and Matlab/Simulink. The research results confirmed the operability and high efficiency of the proposed system.