Study On Ranging Sonar-optical Vision Target Positioning And UUV Control Technology

As a high-intelligence unmanned system working in complex marine environment, UUV(Unmanned Underwater Vehicle, UUV) must have good situational awareness which is the key skill. The underwater vision system is one of important methods for UUV to perceive the information of the external environment. According to the demands for UUV underwater tasks,such as repairing cable and collecting sample in the seabed,this paper describes a research of UUV in looking for operation target autonomously, mensurating pose and controlling of UUV. It is valuable to promote UUV,s underwater autonomous operating capability in theory and engineering. For the particularity of UUV underwater environment and mission requirements, this paper designs and implements a positioning system basing on ranging sonar and optical vision.Firstly, it proposes a UUV's design scheme of targeting platform based on the ranging sonar and optical vision and completed a monocular vision system calibration.Secondly, this paper describes a research of the methods of underwater operation target recognition. Aiming at the problems of huge underwater image noise and seriously blur gray effect, it proposed the method with smoothing filter and top-bot hat transformation which can enhance the contrast of underwater images and make the details more obvious. For the segmentation process of underwater images, the writer used region growing and modified segmentation method based on prior knowledge of false targets filter, and achieved better result. Choose geometry description with the high anti noisy ability and stable characteristics under different imaging conditions as the characteristics of target region. Through the experiments, it is proven that space characteristic vector which is built by circularity and slenderness can distinguish this designed goal accurately. At last, in order to achieve the recognizing process, the writer achieved target characteristic vector modeling by the Gauss mixture model, obtained the optimization parameter by training samples and improved to reject small probability of the results by adding the threshold judgment.Thirdly, the ranging sonar and optical vision targeting algorithm is designed and the tracking method based on compressed sensing was proposed. For the particularity of the underwater environment and the inaccuracy of underwater laser ranging,the writer proposed a method by using small opening angle ranging sonar to assist monocular vision to complete the target positioning. Furthermore, in order to improve UUV's real-time requirements of underwater target recognition and pose determination in the navigation status, the writer used compression tracking method based on compressed perception technology. That was proofed the feasibility of this method in the tank test.Finally, in order to accomplish underwater operations better, this paper designs a PID controller to control the UUV to make the camera assembled on the UUV ensure the center of the captured image coinciding with the center of the target. The design of the control algorithm was designed on the basis of the current position of the UUV and the obtained three-dimensional coordinates of the center of the object. So the UUV could make the center of image coincide with the center of the target, which could provide a better visual effect for subsequent robotic operations.