| In order to solve the problem that USV has the short-distance perception blind zone using the traditional marine environment perception system,the perceptual result is not precise enough,the cost is high,and the installation is inconvenient,This paper mainly attempts to realize the three-dimensional fine perception of the USV to the environment by using stereo vision technology.In this project,I analyzed the stereo vision camera projection model in detail.Based on the theory of stereo error,I build the offshore experimental platform and did the acquisition experiment in the offshore environment outside.The paper deals with the problems encountered in the marine environment,Analyzes and explores the corresponding solution.In view of the adverse effects of stereo image matching in the maritime environment,this paper analyzes the unfavorable influencing factors,and uses the improved image enhancement algorithm to preprocess the image.The paper analyzes and compares the impact of seawater on matching calculation,Based on the local stereo algorithm and the global-based stereo matching algorithm,I proposed an improved scheme of local matching algorithm based on gradient and luminance fusion is proposed.The semantic segmentation of seawater image by U-Net network is proposed to separate the seawater in the image.For the calculated parallax image,this paper performs the parallax subsequent processing of the left and right consistent detection and maps the stereo image to the 3D point cloud space in combination with the stereo structure parameters.The system is combined with the point cloud data to generate the environmental cost map of the USV.Make the environmental observation data enabled to be used directly for subsequent avoidance planning decisions.It can be seen from the final comprehensive experimental results that the proposed stereo vision technology framework has certain practicability and provides a basis for the future intelligent application of USV. |
PDF Full Text Request