The Tracks Extraction And Moving Target Detection In Infrared Images

Developed rail transport has always been treated as an important factor to promoteeconomic development. In recent years, in order to promote the development of China’seconomy better, it has become an inevitable trend of the railway transport developmentto speed train speed up and increase the train numbers, but there must be accompaniedby the security problems in the rail transport. The casualties outside the rail road inrailway accidents are always a great potential hazard to the safety of China’s railwaytransportation. They not only have some impact on the rapid development of China’seconomic lifeline, but also cause some threat to the people’s lives and safety. So itrequires a real-time analysis of the condition of road in front of the train process ofmoving. The real-time tracks extraction and the moving target detection are crucial.The research of this thesis is divided into two parts, one part is the tracks extractingbased on infrared images, and the other part is on the moving target detection along therailroad tracks, such as pedestrian detection. In the experiments, use the infrared cameramounted on the train head to collect the original infrared image, and carry out somedigital image processing on the frame of the video image sequence by OpenCV opensource platform.As the tracks marked the most important areas in video, extracted tracks is thebasis for monitoring the safe operation of trains based on the image processing. First ofall, this paper makes a series of preprocessing operation on the acquired images, viagray, to reduce the amount of image data and improve the processing speed of the image.Secondly, through the use of median filtering and histogram equalization process toreduce image noise and enhance the image contrast. And then, proposing the relevantmethod of chain code to identify and remove the edge line in the image based on usingthe Canny operator to detect the image edge, combined with the cumulative number ofchain code values to determine the direction of the corresponding segment, thus furtherdeleted some interference segment of the image. Finally, using the connected by linesegments to complete the rail contour extraction.The main purpose of moving target detection based on infrared images is to detected the target general outline from the complex background from the collectedinfrared image. In this paper the moving target detection is the detection of movingtargets in the dynamic background, this is the most complex and prevalent mode of aclass of dynamic background research. There is litter related research, and the relatedtheory is not yet mature recently. Two methods have been used in the course of thestudy: one is to achieve the background motion estimation based on the feature pointmatching method; the other one is the displacement of the moving target relative to thecamera is different from the displacement of the dynamic background relative to thecamera, based on this, improved block-matching algorithm for moving target detection,thus eliminating the background moves interference and accurately detect the motion ofthe target.Based on a certain amount of experiments on images acquiring from the videoimages, the experimental results can be obtained: Under the real natural conditions, therelated contour information of the rails itself in the image can be effectively and quicklyextracted. An improved block matching method can accurately complete the detectionof the moving target area.