Application Of HOG+SVM In Passenger Crossing Warning Software For Railway Station Platform

With the expansion of urban scale and the rapid development of economic construction,more and more passengers choose to travel by train.In the background of proportion of railway passenger volume in total passenger volume increasing year by year,passenger safety on the platform of railway station has been concerned by public.To ensure the safety of passengers at railway stations,the ministry has set up yellow security lines to alert passengers and arrange staff on the platforms to supervise them.In recent years,intelligent video monitoring technology is a research hotspot at home and abroad.The use of visual intelligence video monitoring on the platform can timely have a feedback of the safety status of passengers,so as to avoid the occurrence of dangerous accidents and reduce the burden of station staff.Based on HOG(1)and SVM(2),the paper carried out the algorithm design and accomplished the design of the early warning software,so as to improve the timeliness and accuracy of the early warning for platform passengers.The main work of the paper is as follows:First of all,the paper establishes the research foundation through extensive literature reading.In the preprocessing of image,the application of filtering,connected domain and other preprocessing technologies in the image are analyzed.In the process of moving target detection,traditional inter-frame difference,background subtraction and background modeling and gaussian mixture modeling are analyzed and compared.In the part of human body detection technology,the HOG feature extraction process and SVM classification principle are analyzed.Then,through the research on passenger crossing warning problem,the solution is divided into three parts: acquiring foreground target,acquiring human body area and crossing behavior detection.In the use of Gaussian Mixture Model,the early warning detection algorithm extracts the moving target and optimizes the moving target to improve the contour segmentation result.Then,the algorithm extracts the feature vectors by gradient histogram and combines the SVM human body training model to obtain the human body region.Finally,through the early warning rules,the region of human body is detected,and the detection result has feedback.The forewarning detection algorithm in the paper establishes the background model by sufficient video data which is from video cache,so as to ensure the accuracy of acquiring the foreground region;the good human body region was obtained by maximal suppression of the detection results;by using multi-feature detection,boundary point deviation detection and hazard area detection are combined to ensure the accuracy of detection results.At last,in the part of the design and implementation of railway station platform passenger trans-boundary early warning software,the overall design of passenger transboundary early warning software is carried out on the basis of analyzing the overall demand,functional demand,database demand and software performance demand of the software.The detailed design is carried out around login module,main interface module,initialization module,early warning module and data analysis module.On the platform of Windows10,the Python development environment is used to implement this early warning software.Through the detection and verification of three groups of video cases,the detection accuracy of the behavior of passengers trans-boundary by this early warning software reaches into 88.4%,which can better meet the application requirements.