Research On Detection Of Abnormal Behavior Of Passengers In Elevator Cars Based On Video

In recent years,with the increase in high-rise buildings,people have become increasingly dependent on elevators.However,due to the narrow and opaque space of the elevator car,it is difficult for the outside world to discover the robberies,fights,infringements,and other human abnormal behaviors that occur in the elevator car in time.Therefore,effective monitoring measures need to be taken to avoid the occurrence of abnormal behavior of elevator passengers as much as possible.The traditional artificial video surveillance system has disadvantages such as fatigue,negligence of monitoring personnel,long alarm response time.The video-based image analysis and processing technology is applied to the monitoring scene in the elevator car.The detection algorithm automatically detects human anomaly and performs an alarm,which is of great significance for timely detection and processing of abnormal elevator behavior.In this paper,according to the particularity of the elevator car environment,the common state of the elevator car and the behavior of the passengers are analyzed,and the violence between the passengers in the car and the behavior of the passenger's doorway are studied.Aiming at the preconditions and characteristics of the two abnormal behaviors,the corresponding detection algorithms are proposed.The paper first analyzes the image characteristics of the car scene,determines whether the elevator car carries passengers,and compares different foreground extraction algorithms in the car environment,and introduces background updating rules to adapt to dynamic background changes and achieve effective goals.Secondly,for the detection of passengers' violent behavior,the optical flow that can embody the motion information of the image feature points was selected for analysis.At the same time,the optical flow at the corner points was calculated to reduce the calculation amount.The pyramid LK optical flow calculation based on Shi-tomasi corner points was used to determine whether there is suspicious violence in a single image,and to give a method of judging violence based on single image suspicious violence.For the passengers' cardias behavior,the body's joints in the car are identified using the twodimensional pose estimation method based on the Location-Focus Field(PAFs),and the presence of cardia behavior is determined by calculating the angle between the limbs.In the end,we mainly use OpenCV to complete the algorithm,and complete a Java webbased abnormal behavior detection and display platform.At the same time,we briefly introduce the details of software design.Finally,the software was tested on abnormal behavior detection software.The results showed that the method proposed in this paper can effectively detect the passenger's violent behavior generate alarm signals in time.