Research And Application Of Embedded Human Flow Statistics System Based On Cloud Computing

With the continuous enlargement of city scale and the development of transportation network,the number of floating population in cities increases rapidly,and there are more and more criminal incidents such as mass incidents and violent incidents.Video surveillance systems are widely used in urban construction of security monitoring network,which is based on computer,network transmission and image processing technologies,and all.Some video processing technology,such as pedestrian flow detection,pedestrian behavior analysis,the use of video surveillance gradually to automation,intelligent,out of the traditional video surveillance only human naked eye observation unreliable defects.However,with the development of embedded pedestrian flow detection system,the traditional embedded pedestrian flow detection system can not meet the requirements of multi-video simultaneous processing,and its storage resource is very limited.In view of the above,this paper designed a embedded human flow detection system based on cloud computing.The main contents of this paper include:Firstly,the related technologies and algorithms of the number-counting system,which are commonly used at home and abroad,are studied,and four commonly used background modeling algorithms are compared and analyzed.Secondly,the improved Adaboost algorithm is adopted.Aiming at the problem that the output value of Adaboost weak classifier is only determined by a single interval,the output value is determined by multi-interval,which makes the output more stable and improves the stability of the algorithm.Thirdly,the improved Adaboost algorithm is used to detect the pedestrian head,and a practical solution is given for the problems encountered in the head detection process,such as monitoring distance,pedestrian occlusion,multi-scale detection and so on.The effectiveness of the system is verified by the internationally accepted video sequences,and experimental evaluation indexes and experimental results are given.Fourth,a sample bank consisting of 8160 head samples and 14682 background samples was constructed.The library is composed of multiple real scenes,including clothing,hair different head,as well as cloudy,rainy and other background samples,such as pedestrian head detection algorithm for validating validated experimental data for further research work Laid a good foundation.Fifthly,aiming at the shortcomings of embedded computing and limited storage resources,this paper uses the advanced cloud server and cloud storage service of Ali cloud to construct a embedded pedestrian flow detection system based on cloud computing,and completes the process from video capture to upload To the cloud server,the cloud server real-time video processing,and through the Web page shows real-time traffic statistics results,historical statistics and real-time monitoring screen.