Research On Key Technology Of High-precision Extraction Of Structured Light Stripes In Complex Environment

By studying the relationship between wheel and rail instantaneous contact posture under high-speed conditions and finding the relationship between abnormal contact posture and disease,it can help us to put forward maintenance and repair suggestions for wheel and rail disease,thus providing important support for the safe operation of rail transit in China.The key technology of abnormal contact posture measurement is 3D reconstruction based on structured light.The accuracy of the center point of the structured light stripe directly affects the accuracy of three-dimensional reconstruction,but it is affected by the complex environment of the outside world.Therefore,the main research contents of this article are as follows:(1)We studied the coding structure light stripe image enhancement algorithm based on Retinex theory.Aiming at the problems of noise,high exposure and low exposure in images under complex environment,we propose a multi-scale Retinex algorithm based on adaptive gamma function.The algorithm first uses median filtering and HDRCNN to preprocess the captured images to remove salt and pepper noise and enhance the information of high-exposure areas,then uses the multi-scale Retinex theory to obtain the illumination component of the image to define an adaptive gamma function,and finally uses the gamma function to correct image brightness.Experiments show that our algorithm can obtain better quality images,and has higher real-time performance.(2)We studied the sub-pixel extraction algorithm for the center point of color stripes.Aiming at the problems of incomplete and low accuracy of stripe edge detection and not considering the relationship between the position of the center point and the direction of the stripes.we propose a sub-pixel detection algorithm for color stripe center point based on the normal vector.The algorithm first determines the stripe sub-pixel edge points by calculating the second derivative zero point,then locates the color stripe thick center point based on the geometric center method,and finally adaptively extracts the color stripe subpixel center point based on the Hessian matrix.Experiments show that the accuracy of our algorithm is improved by 24% compared with the geometric center method which has the best effect,and the matching rate of the feature points of our algorithm is an average increase of 0.2% compared with the geometric center method,and an average increase of 0.5% compared with the gray gravity method,and the fixed ? value is compared with our algorithm,and the accuracy of our algorithm reaches its best accuracy.These implementations fully verify the high accuracy and strong robustness of our algorithm.(3)We developed a stripe center high-precision extraction demonstration system.Firstly,aiming at the problems of the slow speed of our stripe center extraction algorithm,multi-process parallel is used to accelerate the processing.Experimental results show that the speed of stripe center extraction is increased by 1.2 times.Then based on the research content of this paper,using Py Charm and QT as the experimental platform,the results of image enhancement,edge detection and stripe center extraction are visualized.