Research On Stereo Matching And Position And Orientation Determination Of UUV Target Image

When Unmanned Underwater Vehicle operates autonomously underwater in close range,its binocular vision system can provide relatively accurate integrity environment information,and guide the UUV to locate the target accurately, which is an actual application requirement of this paper. On the background of UUV's close arrival,it researches how to obtain the target and attitude information by binocular vision in close range. This paper mainly includes:Firstly, this paper introduces the hardware composition (underwater camera, image acquisition card, underwater lights) and installation scheme of binocular vision system in detail. Set up the imaging model of the two cameras, linear camera model and nonlinear distortion of camera model. Then, the principle of Zhang Zhengyou calibration method is introduced and the calibration tests on single camera and binocular vision system are carried out. From the calibration tests, their internal and external parameters and space position parameters are solved, which lay the foundation for the following three-dimensional coordinates of the solution.Secondly, several classical corner extraction methods are introduced, mainly including Harris and FAST algorithm. In order to ensure the images captured by UUV accurately matched in spite of low contrast and environmental noise, an improved corner detection algorithm is proposed. Firstly, FAST algorithm with adaptive threshold is applied to eliminate redundant corners. Secondly, improved Harris operator is used to detect corners in multi-scale and multi-direction. Finally,a circular template is applied for non-maxima suppression.Simulation studies are performed based on the experimentally captured gray images to demonstrate the effectiveness, real-time of the proposed algorithm.Then, according to the matching technology of binocular vision system, matching constraints,four key technologies and the selection of matching primitives are introduced.The matching algorithm is proposed to improve the image matching algorithm based on point matching. After corner detection, the corner matching based on Euclidean distance method is completed, and the matching points are eliminated by the most relevant and sub correlation ratio method. Combine with feature vector, the corner matching method is completed based on Euclidean distance method, and by the most relevant and sub correlation ratio method, the false matching points are finally eliminated. The improved algorithm is used on underwater images to verify the effectiveness of the proposed algorithm.Finally, the method to estimate pose and position of the operational objective are researched. The non-parallel binocular vision model is used to solve the three-dimensional coordinates of the feature points. The vector relationship between the world coordinate system and the new right coordinate system is established. According to the transformation relationship between the two coordinate systems, the real goal in the world coordinate system is obtained. Then the UUV pool experiment is carried out, the images obtained are dealt with corner extraction, images matching, the calculation of 3D coordinate and estimation of pose and position, which have been introduced in the previous chapters. Compared with the target position in advance, the algorithm has been proved to be desirable in the range of error allowed.