Research Of Binocular Stereo Vision Matching Algorithm Used In Railway Foreign Body Automatically Recognition

Because of the illegal objects invading railway tracks, railway accidents cause seriousthreat to the transport efficiency and safety of the people. Railway foreign body intrusiondetection plays an important role in ensuring the safety of the railway. With the rapiddevelopment of intelligent video technology, video-based monitoring technology withreal-time, high efficiency and other superior performance has been widely studied andtentatively used.Firstly, the thesis analyzes the structure and imaging model of binocular imaging systemon the whole, then the data processing method focusing on binocular monitoring system ismainly discussed. Secondly, the thesis introduces changes among the coordinates of thecamera, track recognition algorithm, stereo matching algorithm, the invasion of foreign matterdetermination method and other systems processing algorithms. The system can alarm timely,and provide location and size of the foreign body when the foreign matter intrudes monitoringsystem. Lastly, the thesis uses a large number of simulations to verify the effectiveness of allalgorithms. The main contents of the thesis are as follows:(1) Calculating the relationship between three-dimensional world and two-dimensionalimage, and combining with binocular vision calculates the relationship between foreign bodyposition and three-dimensional reconstruction.(2) According to the track’s brightness and continuity feature, combining the edgedetection algorithm and Hough algorithm, the thesis introduces algorithm for railway tracksautomatic recognition. The algorithm not only ensures recognizing accuracy of the straighttrack, curved track and switch section, but also ensures the resistibility to the noise andshadow. So it can be the reference of the foreign body invasion.(3) By comparing the analysis of various stereo matching algorithm performance, scaleinvariant feature transformation (SIFT) algorithm is partilly improved, which has increasedthe computing speed when dealing with railway scene images, and the feature points in thesame match threshold, the amount of information in favor of three-dimensional reconstructionof the objects.(4) By focusing on the delineation of the railway scene image monitoring area, theapplication of background subtraction to determination railway foreign body achievesdynamic monitoring of the railway scene and the alarm and dynamic tracking of objectintruded railway gauge, which also can judge the position and size of the invasive objects.