针对传统监控系统通过更换硬件设备实现智能化所带来的巨额费用和视频分析算法受到硬件平台的限制，没有发挥其应有的作用等问题。本文提出了一种新的基于视频分析技术的监控系统架构，采用这种架构的监控系统不仅可以作为独立的智能视频分析系统使用，而且可以作为智能视频分析模块加入到原有的监控系统当中，使得可以在支付最少成本的情况下改变传统监控系统过于依靠人力来监视的应用模式，进而使得人工智能技术能够在安防体系中得到广泛推广。 在系统设计部分，首先设计编写系统界面，采用多线程技术设计系统框架和图像处理算法接口；其次捕获摄像机和硬盘录像机输出的视频数据，并且实现预览、抓拍、压缩存储等监控系统基本功能；然后通过帧回调函数将原始视频数据提取出来，将其转换为适合OpenCV处理的图像类型以供算法使用，再利用这个接口嵌入入侵检测、人脸检测、火焰检测、烟雾检测等算法；最后，根据不同算法设计报警规则，进行声光报警，支持人机联动。由于采用多线程技术进行算法处理，所以可以单独或并行运行以上算法检测内容,实现了一机多用。 视频分析方面着重对火焰检测算法进行研究，首先概述传统图像分割方法的不足，然后说明本文使用的灰度积分投影分割法，该方法使用火焰的亮度信息可以有效分割疑似区域；然后对图像进行颜色特征的提取，用颜色模型对图像信息进行匹配，提取出火焰疑似目标；接下来对疑似目标进行面积变化率、分散度和图形相似度等三个特征的计算；最后对这三个特征进行融合判定，判断是否发生火灾。由于系统可以同时运行火焰检测和烟雾检测算法，所以在火灾的判定告警方面取得了较好的效果。 实验方面，首先给出组成监控系统的硬件设备；然后演示监控系统的基本功能；其次对算法实时运行状况进行验证，结果表明系统功能运行正常，各算法均能及时有效的检测出异常并发出报警信号，最后对报警响应模式进行演示。

The traditional monitoring system intelligentize by replacing hardware brings huge costs and video analysis algorithm is limited by hardware platform. To solve this problem, this paper proposes a new monitoring system based on video analysis technology architecture. Using this architecture monitoring system not only can be used as the independent intelligent video analysis system, and can be used as an intelligent video analysis module is added to the original monitoring system, can make the payment under the condition of minimum cost to change the traditional monitoring system too rely on manpower to monitor the application of the model, at the same time make the artificial intelligence technology can be widely spread in the security system. In the system design part, firstly designing system interface, using multithreading technology design system framework and interface of image processing algorithm; secondly capture cameras and output video data of hard disk video recorder, and realize basic functions of monitoring system such as the preview, captured, compressed store; then by frame callback function to pick up the original video data and converts it to suitable for OpenCV image type to processing, using the interface embedded in intrusion detection, face detection, fire detection, smoke detection algorithm; finally, according to the different algorithm design different sound and light alarm rules , realize the man-machine interaction. Because the multi-thread technology is adopted to improve the algorithm, so can separate or run in parallel detection algorithm. In terms of video analysis, focus on the study of flame detection algorithm. Firstly explain the shortage of the traditional image segmentation method, and then illustrate the gray integral projection method, this method used the brightness information can effectively segmentation suspected area; and then to extract the image color feature, using the color model for determining and obtain the suspected targets; then calculated the characteristics of suspected target such as: area change rate; dispersion; similarity; finally, fused the characteristics of suspected target in together, determine whether there is a fire. Due to the system can run the flame detection and smoke detection algorithm at the same time, so the judgement of the fire alarm has achieved good effect. In the experimentation part, firstly introduction hardware equipment of control system; and then demonstrate the basic functions of monitoring system ,secondly detection effect of the proposed algorithm, results show that the algorithm can timely and effective to detect the abnormal and send out alarm signal, finally presentation alarm response pattern.