Research On Underwater Pose Measurement Technology Based On Binocular Vision

Humanity has never stopped exploring underwater resources.In order to do this,various types of underwater detection equipment have been developed,such as underwater robots and submarines.Since the visual measurement of underwater targets is the most direct and can get richer information in the short-range underwater measurement,underwater visual measurement has become a research central issue.At the moment,underwater vision measurement is mostly focused on getting the position information of the target,However,there is little research on the posture information of the target.Therefore,in this paper,the technical research of underwater target posture measurement by means of the vision system is of important meaningThe main research contents of this paper are as follows:Firstly,The imaging model of a single medium leads to the problem of large measurement errors in an underwater environment measuring target.According to the law of refraction,to achieve the purpose of improving measurement accuracy by establishing a new refraction imaging model of water,glass and air.Aiming at the problem that the two imaging planes are not parallel due to the inevitable existence of a certain angle when the binocular camera is placed.According to the idea of stereo correction,to achieve the purpose of two imaging planes being parallel by using the method of multiplying the correction parameter matrix and the pixel matrix.Aiming at the problem that the two imaging plane polar lines are not on the same line,resulting in matching errors.The method of multiplying the correction product parameter matrix and the pixel matrix is adopted to achieve the goal that the polar line is on the pixel of the corresponding matching point row.Secondly,the matching algorithm based on Census transform has poor matching accuracy and is easily affected by the center value of the template.A method proposed by people is to add a Gaussian filter function and use a four-bit binary to represent the comparison result,to achieve higher matching accuracy and improve anti-interference ability.In view of the large amount of image data in the binocular vision system,the image data collection is not synchronized,the parallel operation cannot be performed,and the stereo matching cannot achieve real-time performance,people have proposed the use of hardware acceleration,to achieve the purpose of synchronous image acquisition and real-time stereo matching.Next,making a reasonable selection of the hardware system,such as from camera,processor,memory,to water container,etc,according to the characteristics of underwater environment and binocular vision measurement.To complete the construction of an underwater pose measurement experimental platform based on binocular vision and verify the system's automatic acquisition,transmission,previous treatment and display functions.Finally,the improved calibration algorithm and stereo matching algorithm are used,and then the hardware simulation experiment is performed to obtain the result,to prove the effectiveness of the improved algorithm,according to the experimental result data that compare images before and after correction.According to the theory of binocular disparity map and relative pose measurement,experiments prove feasibility of underwater pose measurement.