Research On Track Obstacle Detection Algorithm Based On Machine Vision

Railway plays a leading role in the field of transportation in China and shoulders the important mission of driving the development of national economy.Because of the line long mileage and the complicated environment of railway,the random foreign matter intrusion,such as landslides caused by natural disasters and employees staying at their posts in the front of the high-speed train,which seriously affects the safety of the train.Nowadays,with the rapid development of high-speed railway,only artificial detection and fixed monitoring points cannot meet the needs of the society which requirements are the most automatable.As the realization of full automatic driving in rail traffic,the research and application of detection technology based on machine vision,how to realize intelligent detection of foreign matter intrusion in front of high-speed trains has become a new trend of railway operation,which has irreplaceable practical significance.Based on the research of railway obstacle detection technology here and abroad,this paper uses On-board Monocular Camera to realize the detection of railway obstacle.By analyzing the image of the acquired video frames,putting emphasis on the segmentation algorithm of railway track obstacle images,and focusing on how to improve the real-time and robustness of the improved algorithm,and the algorithm is applied to comprehensive identification system to detect obstacles,and according to the two key steps detection to complete the judgment of whether there is any obstacle.The main contents of this paper are as follows:(1)Railway track image edge detection.In view of the static obstacle existing in the front of the railway track surface,firstly,it is necessary to preprocess the obtained frame sequence image,so as to improve the real-time performance of the algorithm and reduce the interference of the external noise.Secondly,the edge detection of preprocessed image,the extraction of railway line and the establishment of detection window are used to improve the efficiency and the accuracy of the detection.In the above process,different smoothing and edge detection algorithms are carried out for qualitatively and quantitatively analysis,so as to select the appropriate algorithm to achieve the purpose of building a detection window.(2)Research on track obstacle detection algorithm.A good real-time and robust segmentation algorithm is the key to fast recognition of obstacle detection.The paper proposes to improve the real-time performance of the algorithm and improve the intelligent optimization algorithm to complete the segmentation of the obstacle images.On the one hand,the idea of recursion and restricted range is introduced in the traditional fast binarization algorithm.Firstly,according to the characteristics of the fast binarization algorithm,the recursion formulas of four parameters are deduced,and the complexity of the image is calculated,then fast threshold segmentation of image is finished within the reduced gray level ranges,finally,the improved algorithm is applied to the image segmentation of the track obstacles.On the other hand,a two-dimensional Tsallis entropy decomposition algorithm based on improved particle swarm optimization is proposed.Firstly,the two-dimensional Tsallis entropy algorithm is decomposed into two one-dimensional Tsallis entropy,and in the objective function,the minimum within-cluster scattered degree is introduced.Secondly,the traditional particle swarm optimization algorithm is easy to be precocious and difficult to converge,and the improved algorithm improves the defects of the algorithm though introducing sensitive particle.Finally,the objective function is used as the optimization function of the improved particle swarm optimization algorithm to complete threshold segmentation with the global optimal solution.(3)Detection of static obstacle.At present,the obstacle detection cannot be judged under a single index at the changing environment of the rail track.In this paper,the black and white pixel ratio after segmentation and the integrality of the sleepers are judged by using the gray level symbiotic matrix,both of them are used as the accurate detection.At the same time,the gray feature of the image is used as the preliminary detection,and a comprehensive recognition system is established to judge whether there is an obstacle in the detection window.The improved algorithm can improve the real-time and accuracy of obstacle detection by analyzing the experiments in different scenarios,different obstacles and measuring the algorithm under the objective index.