Design And Implementation Of A Video-based Railway Level Crossing Obstacle Detection Algorithm

The railway level crossing safety problem is one of important factors restrictdevelopment of railway transportation. During the reconstruction of enclosed railwaysystems, level-crossings video surveillance systems can effectively relieve level－crossings safety problems. The existing level-crossings video surveillance systems justsend the videos to the train or the monitoring center through the network. And then atrained person always observes the monitor screen, and analyses the scenes, and judgesthe situation of the front railway level crossing. That is an onerous procedure of visual.To address this issue, by seriously analyzing the causes of the level-crossings accidents,a video-based railway level crossing obstacle detection algorithm is proposed. Thisalgorithm can automatically detect and identify obstacles without observers. The maincontent of the article is as follows.First of all, the video image stabilizing method based on the gray projection andblock matching is proposed. Respectively, in the horizontal and vertical directions, themacroblocks without local motions and simple grayscale distribution are selected by thegray projection method. By using the block matching method, the local motion offsetsof the macroblocks are calculated. The global motion compensation of current image ismade by using the mean of the local motion offset in the direction.Secondly, an obstacle detection method based on partitioned background updatingmodel is presented. Real-time updated background image can be obtained via thebackground updating model. The detection and extraction of moving target can beachieved by background subtraction. Ghost detection is completed by the edge of thematching method. False targets are discarded. By accumulating continuously occurringmoving target pixel corresponds to the score on the scoreboard, the obstacles aredetected and identified.Finally, the proposed algorithm is implemented by programming. And thecorrectness of the methods and the cup utilization rates are tested and evaluated.