Research And Application Of Intelligent Video Surveillance Resource Scheduling And People Statistics Algorithm

Monitoring system is widely used in real life,traditional monitoring system can only play a simple video and alarm function,with the integration of image processing,computer vision and AI technologies,the concept of intelligent surveillance is becoming more and more popular.Initially,intelligent monitorings mainly adopt cloud computing technology.The cloud computing center has strong computing capabilities,but energy consumption,bandwidth limitations,and privacy security have always been its main problems.The emergence and rapid development of edge computing technology have effectively solved these problems.This article designs an intelligent monitoring system with real-time intelligent video analysis functions,add real-time intelligent video analysis function on the traditional monitoring system,pedestrian detection and feature extraction are mainly performs through deep learning technology,and the people counting function is based on this.Then the result is sent to the platform for display through the network.Different from the cloud computing analysis method,intelligent algorithms are implemented inside the edge system.Embedded system development needs the guidance of the simulation system.This paper first analyzes and compares the development environment of the system simulation platform and the edge platform,On this basis,they were developed differently.The simulation platform design focuses on video processing algorithms and data transmission algorithms.In order to meet the performance and cost requirements of embedded systems in practical applications,this system selects the ZYNQ7010 development board as the edge processing platform.The intelligent processing component performs good,but the resources and processing capabilities of the edge processing platform are limited.After performing performance analysis on it,it is found that the pre-processing process cannot reach real time at all.Therefore,this article analyzes the resources of the ZYNQ7010 development board,and designs a resource scheduling algorithm,which accelerate the video preprocessing process greatly.In order to ensure the real-time performance,also proposes several frame skipping processing schemes to make the video pre-processing process become completely realtime based on the video characteristics.This system has the function of people counting.This paper studies the common methods of it and optimization results of other scholars,combine with the intelligent analysis results,designed a people counting algorithm.The frame skipping processing schemes will reduce the continuity of the video,which will cause errors in the intelligent analysis results,and this will reduce the accuracy of the people counting.To this end,this paper also designs the algorithm of pedestrian tracking and accurate statistics for the unique attributes of the system.The application scenario of this algorithm is not only limited to this system,but can also be applied to the situation of poor video analysis results caused by various problems.This algorithm can compatible with face recognition and pedestrian detection results,can effectively improve the accuracy of people statistics.The system is powerful,its design and deployment of the software and hardware environment are relatively complicated.Therefore,this article also introduces these in detail,and designs a visualization scheme to verify the system's functions effectively.The main point of the performance design lies in the real-time nature and the accuracy of the people counting algorithm.Therefore,this paper selects representative experimental data to test the performance of the system and chooses the implementation plan.The results show that the system can achieve the expected performance indicators after optimization.