Research And Design Of Intelligent Monitoring And Early Warning System Based On OpenCV

With the improvement of people's safety requirements,intelligent video surveillance systems,as the main means of security warning,are more and more widely used in daily life.However,the purchase of monitoring equipment often costs a lot,and installation is not convenient.This article aims to develop an intelligent early warning system that monitors specific areas in a short period of time using only a home computer and a USB camera.The software development environment of this system is Visual Studio 2015 and OpenCV 3.4.2,the core code is written by C++,and the software interface is written by C#.The application scenario is to perform anti-theft warning when the target is visible,such as the anti-theft alarm of the electric vehicle downstairs.The design idea is as follows: When there is a moving target,the moving target can be detected in time,and the moving target is judged.If it is a pedestrian,the tracking is performed,and the abnormal behavior is predicted in the tracking process.The main work of this paper is as follows:(1)The theoretical research and experimental exploration of the traditional moving target detection method are carried out.The traditional moving target detection method can not give the accurate position of the target when the multi-target distance is close.This paper introduces the SSD network target detection method based on deep learning.The advantages and disadvantages of the SSD method are explored experimentally.Finally,an improved moving target detection method based on frame difference method and SSD network is proposed.The new method can effectively avoid the shortcomings of the frame difference method and the SSD network alone,and optimize the target recognition at the edge,which effectively improves the accuracy and efficiency of the target detection.(2)The seven tracking methods in the OpenCV extension module are introduced theoretically,and the accuracy and speed of each algorithm are compared through experiments.Because the target tracking encounters illumination changes,scale changes,occlusion,etc.,the tracking fails.The SSD network is added during the tracking process,which not only realizes full-automatic tracking,but also re-detects the target position when the target tracking fails,thereby restarting the tracking method.Finally,the experiment found that the MOSSE tracker is the most suitable tracking method for this system,because the method consumes the least total time.(3)Combining the previous target detection part and tracking part,the design of the overall program of the system and the design of the software interface are completed.Finally,the multi-instance case is tested and the test results of various situations are counted.The measured results show that the system can accurately detect and track.The goal of the movement,and early warning of abnormal behavior,achieved the purpose of the original design.