The Design And Implementation Of Escalator Intelligent Video Surveillance System Based On TK1

With the extensive use of escalators in public places,issues including energy conservation,smooth passage and security protection have been receiving increasing attention.Besides,under the conditions like passenger flow congestion,big abandoned objects and stranded or falling passengers at the escalator exit,the traffic efficiency and personal safety of passengers are greatly affected.How to monitor the real-time status of escalators and passengers has become a major problem in ensuring the safety of passengers on escalators.The artificial care is costly,while the rapid development of image processing technology,pattern recognition and embedded technology has provided a new idea to solve these problems for escalator manufacturers.This paper aims to design and implement an intelligent video surveillance system of escalator to detect and track the escalator passengers,achieving the target of passenger flow statistics and passenger retention detection and expanding the function of detecting big abandoned objects.According to the characteristics of application scenario,the installation method of vertical erection of camera is adopted to detect human head.The TK1 embedded platform based on ARM + GPU is chosen as the hardware platform of the system.The OpenCV open source visual library is used to proceed the algorithm design,the Qt cross-platform graphical interface library is used to carry out interface design,and the CAN bus and escalator controller are used for communication.The detection algorithm of the system in this paper mainly includes three aspects:1)Passenger head detection: adopt the passenger head detection method based on HOG features and AdaBoost cascade classifier.Collect a large number of head and no-head samples on the escalator scene.Train the classifier based on positive and negative samples,whose recognition rate and the error rate meet the requirement of the test.2)Passenger head tracking: analyze the frequently used moving target tracking method and the tracking characteristics on the escalator scene.Adopt the passenger head tracking method based on neighbor frame matching and Kalman filter.Originally transform the problem of target matching into that of assignment and adopt the Hungarian algorithm to solve the problem of assignment.Complete the passenger head tracking through the Kalman filter,reduce the impact of missed and false detection on the detection result,and proceed the detection of passenger flow statistics and passenger retention.3)Detection of big abandoned objects: analyze the advantages and disadvantages of common background modeling on the escalator scene.Adopt the average background method for background modeling and improve the updating method of Surendra background.Carry out the foreground detection based on background difference method and design a detection algorithm of retained objects based on tracking of shape and center of objects.Finally,this paper adopts class oriented method and multi-threaded architecture to design a software for intelligent video surveillance system of escalator,including the exchange management module,image acquisition and display module,image processing module,communication module as well as audio and video recording module.Carry out tests on the detection result of the entire system and CAN bus communication to prove that the system can complete its expected task.The system studied in this paper is of great significance to realize the intelligentization of escalator.