Research On Foreign Matter Intrusion Detection System For Railway Based On Single View Technology

The traffic safety is a permanent theme of railway transportation. However, thereliability and automation of the current foreign matter intrusion detection methods is poor inour country, so that it cannot guarantee traffic safety fully. The single view method ismentioned to apply on the foreign matter intrusion detection system. The foreign matter isdetected by image processing technology, and the feasibility of the scheme is validated bysimulation. The main contents of this thesis are organized as follows:(1) In the whole systematic frame designing part, distributed network system architectureis introduced, which improves systematic dependability and made management easier. Byanalyzing the division principles of monitoring points and all functions which the monitoringsystem need to achieve, the monitoring points are arranged best with great applicability.(2) In the process of system design, considering the digital images are generallycontaminated in the process of transmitting, producing, and quantifying, the image is filteredby an improved switching median filter. Segmented statistical means and the principle ofswitching is used to the filtering algorithm, so that this method can not only suppress manykinds of image noise but also protect the image detail effectively.(3) In order to overcome the weaknesses of ineffective edge detection by the existingedge detection algorithm, the improved ant colony algorithm is used to detect the edge ofrailway in the thesis. Searching heat factor is added in path selection of ant colony to increasethe optimization of the algorithm. By after considering the continuity and integrity of therailway edge, heuristic function is redesigned combined with gray gradient values, the phaseangle of the pixel and the directional of ants moving, so that the precision of the edgedetection is improved. Then the pheromone updating rule is redefined and the localpheromone update formula is dynamically adjusted to increase the convergence rate of thealgorithm. The experimental results reveal that the improved algorithm can detect the edge ofrailway clearly and completely, and can provide guarantee for the next step.(4) Design installation location of the video camera according to the requirement of therailway clearance and the height of the obstacle. So that adopting single view technology canmonitor the height of the obstacle. The relation of the obstacle and the video camera is classedas the same kind due to their movement relationship, so that the monitoring process iscondensed.(5) The preliminary identification for obstacles according to the average change of grayvalue, design the precise guidelines aimed at rail lines and texture feature of in-orbit. By amulti-step monitor process, it guarantees the accuracy of monitoring. The final experimentalresults reveal that the algorithm has the higher recognition rate, and show that the scheme is available in effect.