Detection Of Rail Surface Defects Based On Image

Detection of rail surface defects has important significance for the stability and security of the train, Image processing technology has the advantages of high speed, anti-interference, high accuracy, has become a hot research in the field. This paper focuses on the related technology of rail surface defects detection system based on image, the main research work are as follows.The method of fast median filtering method based on mean is proposed. The method according to the characteristics of defect region gray low, Combines median filter and mean filter, and proposes the fast median filtering method based on mean search. It’s can quickly filtering the noise acquisition detecting image included, while reducing the image details loss.An adaptive extraction rail surface region based vertical projection integral method is proposed. In order to improve the speed and accuracy of system, determine the start and end positions of the integral by the average value of vertical projection curvet to reduce the search range. According to the width of the rail calculation definite integral, extract precisely the rail surface region.The top-down visual model based gray gradient is proposed to find quickly the images included defects from a large number of images. The bottom-up visual model based multi-scale Markov random walk is proposed to quickly and accurately extract the defects from images included defect. The method simulation the principle of human visual can extract the defects of rail surface quickly and precisely.The rail surface defects classification method based PLSA model is proposed. Due to the irregular shape and feature unobvious of the rail surface defect, reference PLSA model based probability and design rail surface defects classification model based PLSA. The method across the "semantic gap" between the image low-level features and high-level language of human perception, which has the advantages of accuracy, speed, and is easy to realize.