Research On Related Technology Of Stereo Vision In Measuring System Of Turnouts For Rails

Stereo matching algorithm in stereo vision is a hot topic at present.The main purpose of stereo matching algorithm is to find the corresponding points from different viewpoint images.It is a process of estimating pixel visual difference by establishing a minimization of energy cost function.In this paper,the stereo matching technology in stereo vision is mainly used to measure the dimensions of turnout rail parts,and the stereo matching algorithm is improved in the system of measuring the dimensions of turnout rail parts based on machine vision.The experimental results have been improved in matching accuracy and matching speed,so that it can be better applied to the rail technology field and promote the development of railway industry.First,the related theoretical research of stereo vision is expounded,in which the measurement principle and mathematical model of stereo vision are introduced in detail,and the constraints related to stereo matching and the judgment criteria of similarity are also included.And the evaluation criteria of the algorithm are discussed in detail according to the concept of stereo matching.Secondly,aiming at the problems of low robustness,high time complexity and slow running speed of the perceptual hash algorithm,the discrete cosine transform(DCT)is analyzed as the global feature and the local feature of the ORB algorithm is analyzed,and the perceptual hash algorithm is improved.When the height,width,brightness and color of the images change,the algorithm can quickly recognize the image features,and measure the similarity using the hamming distance.Thirdly,analysis the robustness of stereo matching algorithm under nonparametric Census transform under illumination,and the deficiency of Census transform is pointed out.The inverse color of the transform window is used to calculate the neighborhood mean value of the center pixel to replace the gray value of the center pixel,and then the SAD method and hamming distance are used as similarity measure for stereo matching.The accuracy of matching is enhanced by right and left consistency detection and subpixel interpolation.Finally,the experiment results are applied to the turnout rail size measurementsystem and make a concisely introduction about it.The drill hole of the rail is measured in the laboratory environment,and compared with the previous system data,then the advantages and disadvantages of the algorithm are verified.