The Design And Implementation Of Intelligent Video Surveillance System Based On Embedded Linux

With social security becoming increasingly serious, more and more security equipment is needed. Video monitoring is an important part of safety and protection system, as it can detect potential invasion object quickly and accurately in the form of image. At the same time, the rapid development of computer and embedded technology provides a good condition for the realization of long distance, high intelligent video surveillance equipment. Therefore,intelligent video surveillance system has become a focus of today’s research.This article mainly discusses the application of embedded intelligent video monitoring system in the family security. Using Davinci DM6446 processor as the core, monitoring system can realize real-time detection for detecting area, including the case that someone enters, or something is broken and so on. The concrete research content is as follows:First of all, this paper puts forward the overall design of the system and studies the key technology of embedded video monitoring system, including the development process of the embedded system, DaVinci technology and DaVinci processor. The design and implementation of hardware platform is to be completed, which using Dual-core DM6446 as the core processor, including the design of video processing module, audio codec module,power supply module, reset monitoring circuit, data storage module, network module and communication interface module. It completes the schematic diagram of the hardware platform design and PCB layout. With connection hardware platform to PC, the software development platform basing on Linux operating system has been constructed.Secondly, this paper has a reasearch on moving target detection algorithm. in order to reduce the amount of calculation, video images are pretreated by graying. Then, this paper analyzes the principles, the advantages and disadvantages of three kinds of traditional moving target detection algorithm, inlcuding the interframe difference method, background difference method, and optical flow method. Based on these, the combination of three-frame difference method and edge detection method is choosed to be the finale detection algorithm, after this,the detected results are morphological processing, thus detecting intrusion accurately and effectively avoiding the effect of noise like light. The algorithm judges the segmented moving targets based on centroid features,and deals with images respectively according to the two characteristics of the center of mass of dispersion and centroid trajectory,then compares the two treatment effect,thus getting the most suitable algorithm for this paper. Algorithm has been verified on matlab simulation platform.Finally, The algorithm in C language generates the library files on CCS3.3 software, it achieves encapsulation and transplantation based on the Codec Engine mechanism, and realizes associative function of the system on the embedded Linux platform, namely this algorithms can realize real-time monitoring of the scene, and produce warnings when someone enters the monitoring scope. It optimizes the algorithm, and improves the processing speed and reliability of the monitoring system.