Improvement And Implementation Of Moving Detection And Tracking Algorithm In Indoor Monitoring

Moving object detection,recognition and tracking based on intelligent video surveillance are hotspots in the field of computer vision,and they are increasingly used in modern indoor security systems.With these technologies,we can quickly gain the interested objects in the surveillance area,identify foreground targets,and track the formation of the trajectory of the target for the follow-up target behavior analysis and understanding to lay a good foundation.In this paper,we choose indoor surveillance environment as the research scene,single-target human body as the research object,and single-target human detection,recognition and tracking as the main content of this study.The aim of this paper is to enhance the robustness of the indoor intelligent monitoring system by comparing the existing moving target detection and tracking algorithms and improving some existing shortcomings of the existing algorithms to avoid common disturbances in the monitoring process.The main work of this paper is as follows:Firstly,in the moving target detection stage,the background subtraction algorithm based on GMM algorithm is proposed to overcome the shortcomings of the background subtraction algorithm,which is sensitive to the change of the light.Using the GMM algorithm stability and insensitive to the slow change of light,create a background model for static background images.In addition,aiming at the poor adaptability of GMM algorithm to illumination mutation,it is by defining the area occupied by the foreground target and the number of frames of light mutation to detect whether the room light mutations.Experiments show that the improved moving target detection algorithm not only can detect the foreground object completely,but also can solve the error of the traditional background subtraction algorithm caused by slow change of light and mutation,which greatly improves the accuracy of moving target detection.Secondly,in the stage of human target recognition,this paper proposes a SVM classifier algorithm based on HOG feature to classify the foreground objects in indoor surveillance environment,and classifier training is done by using the positive and negative samples provided by the common dataset INRIA.Finally,the simulation results show the higher success rate of the proposed algorithm.Finally,in the stage of a single-target body tracking,for the shortcomings of traditional Camshift moving target tracking algorithm,such as poor anti-occlusion and large sensitivity of target scale change,this paper proposes an improved Camshift target tracking algorithm.The problem of target occlusion is dealt with by the method of target block tracking,and the target occlusion rate can be determined by the target matching ratio.In addition,for the problem of the tracking error introduced by the excessive scale change,it is by combining the geometrical features of the target with the color characteristics of the target to describe the target more fully and improve the recognition rate of the target.The experimental results show that the improved tracking algorithm of moving target can solve the tracking error of traditional Camshift tracking algorithm with poor anti-occlusion and large scale change well,and improves the robustness of the moving target tracking.