Video-Based Detection System For Postal Station Anomalies Reporting

Due to the rapid expansion of express logistics business and the intensification of industry competition,the service quality and service quality of some express companies can not meet the actual needs of consumers,such as items damaged,lost,“Slow”,poor rating.Attitude,rough treatment,and so on,these problems emerge one after another,seriously damaging the rights and interests of the majority of consumers.According to the survey report of the China Consumer Association,nearly half of the logistics outlets have the problem of violent sorting,of which throwing,smashing,kicking and other illegal delivery of parcels is still very much the phenomenon.Many express parcels exist“Internal injury”,but because the outer packaging intact,there is no way to obtain the corresponding compensation.Based on the above problems,the current solution is through manual monitoring and customer feedback to the salesperson score,according to the score of the salesperson to deal with.However,it is too inefficient to rely only on manual analysis,which requires a large amount of manpower and material resources,and it is difficult for manual workers to monitor a large number of video images in real time for a long time,thus causing low accuracy of alarm and high false alarm rate,the alarm cycle is long,the video data analysis is difficult and so on.At present,there are many related researches at home and abroad,but most of them are confined to image processing or machine learning algorithms,which have some problems such as high false alarm rate,low accuracy rate and poor real-time performance.Many of these systems use only still images,which makes them less efficient at detecting dynamic events such as violence.In order to solve the above problems,this paper proposes a video detection technology-based logistics express violent sorting recognition system.The main work of this paper is as follows:(1)the research background,significance and current situation of violent sorting recognition are analyzed,including human detection,target classification and tracking,action recognition and high-level behavior understanding,etc.,this paper provides the basic ideas for the program(2)in this paper,the video depth image is acquired by Kinect sensor,and the captured image is transformed into joint image and the coordinate information of bone node is obtained.Based on this information,the human joint motion pattern of the violent sorting behavior is analyzed,and the joint motion angle velocity and coordinate change are taken as the target characteristics,and a joint node change sequence model is proposed,which has low complexity,easy to calculate.(3)because the traditional feature extraction methods are limited by the lack of the ability of feature expression and generalization of shallow learning algorithm,based on the STIN network model,this paper uses ROIALIGN operation to extract the different object features of the network output of the basic model,and increases the branches of the characters’ interaction,which effectively improves the generalization ability of the model.(4)based on the PLE sub-module of STRM model and the characteristics of the input data of the model,an improved spatiotemporal enhancement module is designed to aggregate the intra-frame context,for a given feature tag,end-to-end learning is performed using cross-entropy loss(CE),and motion-object features are stitched together by means of average aggregation,for the cases with high spatial-temporal correlation and human behavior,the weight of the value is larger,while for the cases with low correlation,the weight of the value is smaller,the information suppression is realized without significantly increasing the complexity.The system based on video detection technology has many advantages,such as high accuracy,real-time,good stability and so on.It has great application prospect and development space.The significance of this study is to provide a feasible technical scheme for Logistics Express Enterprise Safety Management and Prevention of violence,which has great practical significance.