Research On Automatic Detection And Forewarning Of Foreign Objects Intrusion In Railway Track

With the rapid development of high-speed railways,great progress has been made in the operation technology of high-speed trains,and the running speed has been greatly improved,which is convenient for people to travel,and at the same time puts forward higher requirements for the safe operation of trains.The study found that one of the most important factors affecting the safe operation of trains is the intrusion of foreign objects on the railway track,which not only endangers the safety of people’s lives and properties,but also causes immeasurable losses to national property.Therefore,it is of great significance to accurately detect intrusive foreign objects in real time and issue early warning signals in time to ensure the safe operation of trains.Based on the analysis of the existing foreign object detection methods on railway tracks,this paper focuses on the detection method of railway track foreign object invasion based on video image processing,and realizes the automatic detection and early warning of railway track foreign object invasion.The main research contents of the paper are as follows:(1)Research on the environmental optimization method of railway track foreign object detection.Firstly,the Canny operator and Hough transform are used to detect and extract the edge of the railway track.Secondly,the dangerous area of the railway track is established with the edge of the railway track as the reference line,which reduces the amount of information processing and improves the real-time detection of foreign objects.Finally,through the analysis of shadow features,a shadow detection and removal method based on region matching without training is proposed,which realizes the detection and removal of shadows,reduces the interference factors in the detection of foreign objects,and improves the accuracy of detection.Shadow detection is realized by YCb Cr color space information,HSI color space information and similarity between regions,and it can be detected directly without training,which better ensures the consistency of similar regions.Shadow removal is mainly in the HSV color space.The detected shadow area is used to detect the corresponding non-shadow area,and then combined with the histogram matching algorithm to remove the shadow,the minimum removal operation is realized,and the impact on other features of the shadow area is avoided.The experimental results show that the environmental optimization of railway track foreign object detection is realized.(2)Research on the debounce method of railway track foreign object intrusion image.Aiming at the influence of surveillance camera shake on the detection accuracy of foreign objects on railway tracks,an image de-shaking algorithm is designed.Firstly,image interpolation algorithm and threshold function are used to detect the jitter of the input railway video image.Secondly,the affine transformation and median filtering algorithm are introduced to de-jitter the images in the video sequence with jitter.Finally,the method of frame-by-frame iterative filling is used to fill the black borders of the image after debounce,and the railway image video frames without jitter and black borders are obtained.(3)Research on the detection method of foreign object intrusion of railway track.Based on image de-jitter processing,a Gaussian mixture model that adaptively selects the number of Gaussian distributions and learning rate is designed.Using the optimized Gaussian mixture model,the background modeling of complex railway video images and the automatic detection of foreign object intrusion on railway tracks are carried out.Through the simulation experiments of railway video images in different scenarios,it is verified that the method can achieve better detection accuracy.(4)Research on the automatic early warning method of foreign object intrusion limit of railway track.First,the morphological algorithm is used to remove the falsely detected pixels detected by the track foreign object;then,the method of the ratio of the track foreign object pixel to the image pixel is used to realize the judgment of the track foreign object on the safety of the train operation,and realize the automatic warning function;Finally,in view of the problem of false target train identification,the number of pixels occupied by the train and the change trend in the motion detection results of three consecutive frames of video are used to realize the identification of false target trains and eliminate the early warning that the train is regarded as a foreign object.Simulation experiments show that this method can well realize the judgment of the danger of dangerous foreign objects and the identification of false target trains.