Research On Depth Information Recovery Technology Of Underwater Target Based On Binocular Stereo Vision

With the development of human beings,the resources on land have become less and less,and it is more and more difficult to meet the needs of human rapid development.As the 71% of the area is the "water polo" of the ocean,the earth has sufficient marine resources.The rapid development of science and technology enables human beings to rely on scientific and technological means to explore and extract marine resources,effectively reducing the impact of the complexity and danger of the marine environment on human beings.In recent years,underwater robots have developed rapidly as an important tool for human exploration of marine resources,and binocular stereo vision systems that coexist with underwater robots have also been developed.Underwater depth information recovery can provide position information for underwater robot navigation,obstacle avoidance and autonomous behavior decision-making.This article studies the use of binocular stereo vision system to restore the depth information of underwater target objects.It is hoped that underwater robots can observe underwater scenery like human eyes and judge the distance of objects.Using the difference in the positions of the two cameras,the disparity between the same pixels on the image obtained by shooting the same object is used to obtain the three-dimensional depth information of the target object.Due to the need to obtain the internal and external parameters of the camera,first,Zhang Zhengyou's checkerboard calibration method was used to perform underwater calibration experiments,and the internal and external parameter matrix and distortion coefficient were obtained.By comparing the X direction value in the translation vector in the external parameter matrix with the actual double The relative error of the comparison of the eye camera is 0.69%,which achieves a high calibration accuracy;secondly,the image pre-processing is performed according to the characteristics of the underwater image,including the implementation of distortion reduction,epipolar correction and three defogging algorithms.The characteristics of the three defogging algorithms are obtained.The advantages of the defogging algorithm based on the dark channel prior are obtained.Then,for the characteristics of the weak texture of the underwater image background,the image segmentation algorithm is used to separate the target object from the background.The image segmentation algorithm uses the Grab Cut algorithm combined with adaptive threshold to segment the image to achieve a better segmentation effect.The feature point matching algorithm is used to verify the reduction of the mismatch rate of the image after segmentation.Finally,the stereo matching is studied Step,the principle of SAD matching cost is analyzed,and the solution expression of the three-dimensional coordinates of the object is solved according to the binocular vision system model in the parallel state and the non-parallel state.Program to realize the BM algorithm to obtain the disparity map of the target object,use the disparity value to program to find the depth of the target object,and compare the actual distance between the measured object and the camera,and use the least square method to obtain the relative error of about 4.5%.It has certain industrial application value.