基于 DM6437的智能视频监控系统设计与实现

摘要

针对阵地视频监控系统智能化程度低的问题，提出一种基于 DM6437的智能视频监控系统解决方案。在图像处理算法方面，将 ViBe 算法中采用第一帧图像建模和基于随机策略进行模型更新的思想用于改进传统码本算法的训练和更新阶段，提出一种基于随机码本的运动目标检测算法以提高检测效果。提出在 Mean Shift 跟踪过程中采用 Kalman 滤波器预测目标位置，用于提高目标快速运动和遮挡等复杂情况下的跟踪效果。在系统实现方面，基于 DM6437芯片设计了集图像采集与显示、检测与跟踪、JPEG压缩、网络传输等多种功能于一体的智能视频监控系统。模拟阵地环境下的实验结果表明，该系统能够准确地自动检测和跟踪目标，并且帧率可达6、7帧左右。%To solve the problem of low intellectualisation degree in video monitoring system of positions,in this paper we put forward a DM6437-based solution of intelligent video monitoring system.In terms of image processing algorithm,we proposed a random codebook-based moving targets detection algorithm for improving the detection performance,it is achieved by applying the ideas of adopting the first frame for modelling and updating the model based on random strategy in ViBe algorithm to improving the training and updating stages of traditional codebook algorithm.And then we proposed to use Kalman filter to predict the target position in Mean shift tracking process,so as to improve tracking effects in complicated cases such as rapid target motion and occlusion.In terms of system realisation,we designed an intelligent video monitoring system based on DM6437 chip with various functions,such as video capture and display,targets detection and tracking,JPEG compression and network transmission,integrated.And the result of experiments in simulating the environment of positions showed that the system could detect and track targets automatically and accurately with a rate up to 6 or 7 frames.