Detection Of Abnormal Behavior In Elevator Cab And Its Design Of Monitoring System

Elevator,as a means of transportation in high-rise buildings,has become more and more popular with the acceleration of urbanization and the shortage of land resources.In the one hand,elevator provides a smooth and rapid experience,in the other hand,its potential safety hazards also need social attention.For example,elevator is relatively closed during operation,which is difficult to be observed and intervened by the outside.Although the latest elevators are basically equipped with monitoring sensors.But the monitoring video is mainly processed manually,which still has many disadvantages such as long-time consumption,large amount of data,single function,poor real-time performance,easy omission and difficult investigation and evidence collection.At the same time,the traditional video surveillance system has little interaction with the outside world and can only passively accept information.The intelligent video monitoring system is efficient and it can realize 24-hours uninterrupted real-time monitoring.In the meantime,the intelligent video monitoring system can analyze video information intelligently and alarm online.First of all,this article describes algorithms which used to extract prospects.Aiming at the extraction of the moving target inside the elevator car,a motion foreground extraction algorithm based on the ViBe algorithm is determined.On the basis of the ViBe algorithm,the cavity is filled and the noise is suppressed by mathematical morphology.Experimental results show that the algorithm adopted in this paper can obtain more complete and accurate motion target.Secondly,for the situation of whether the cab is closed,a detection method of door based on straight line detection of foreground edge is designed.Experimental results show that this method can effectively judge the state of the cab.At the same time,for the problem of classification of abnormal behavior scenarios,a passenger counting method based on darknet model is designed to train the passenger head and implement detection.The experimental results show that the method counts the number of people with an accuracy of 96%.Considering the diversity of abnormal events in the elevator cab,this article divides the abnormal events into three categories.For single-person falls,the aspect ratio of the circumscribed rectangle of the human body is calculated based on the foreground.This article uses the aspect ratio to determine whether there is a fall.In the case of multi-person fighting,a corner point kinetic energy model is established in this paper.The pyramid LK optical flow method is used to calculate the Shi-Tomasi corner optical flow in the motion foreground,and the average kinetic energy in the image is counted to realize the detection of multi-person fighting.Based on the difference of frequency between ViBe algorithm and frame difference method,when there is no object in the foreground of frame difference method and there is an object in the foreground of ViBe algorithm,the detection of abnorband object is realized.The experimental results show that the algorithm used in this paper has a detection accuracy of 90% for violence,more than 95% for abnorbands and falls.The speed of the algorithm is about 26 frames per second.Its accuracy and real-time performance can meet the needs of video surveillance scenes.Finally,the above algorithms are integrated and the monitoring system platform is designed and developed.Considering the development and update of subsequent software and the transplantation of algorithms to the mobile development board,QT5,which can cross platforms,is used for development,and the functions are realized by calling OpenCV visual library.The software platform includes object detection module,people counting module,behavior detection module and communication alarm module.Then the software platform is tested through sample video.The experimental results show that the software platform developed based on Qt5 can realize the functional requirements of the design.