嵌入式视频监控系统以体积小、成本低和便于安装等特点被广泛应用于众多领域，其研究工作受到了国内外诸多学者和生产商的关注。 首先，在分析嵌入式视频监控系统需求基础上，给出系统设计方案。该系统包括服务器端和客户端两部分。服务器端自下而上包括4层：①硬件层，②操作系统层，③实现系统所需依赖的各种运行库，④应用层。其中应用层从视频采集开始，经过视频数据预处理、编码压缩、视频内容分析、视频传输等模块处理，将数据发送到客户端。客户端接收、解码、呈现视频数据，接收并进一步处理视频内容分析结果。 其次，实现了一套一体化嵌入式视频监控系统。该系统将采集到的原始视频数据，通过内置于CPU的DSP，编码为H.264码流；再利用RTP、RTCP和RTSP等协议与客户端协商，将封装后的视频数据发送到客户端。 再次，在嵌入式视频监控系统中，设计并实现了一种用于分析视频内容的车牌定位方法。该方法对原始图像分别在HSV和RGB颜色空间做直方图均衡化、二值化等预处理得到两幅二值图像，然后在HSV颜色空间，根据字符跳变规律定位含有车牌的候选区域，接着在RGB颜色空间，逐个提取出候选区域并确定车牌边界，从而提取车牌位置，再采用像素投影法分割出车牌中的字符。 最后，测试了系统功能和性能。测试表明，本系统能正确采集、编码、传输视频数据；车牌定位方法能在较复杂背景条件下准确定位、提取车牌区域，较准确地分割出车牌中的字符。本系统具有很强的灵活性和较高的性能，在车辆管理领域中具有实用价值和参考意义。

Being small in size, low in cost and easy to install, embedded video surveillance system is widely used in many fields. Accordingly, scholars and manufacturers have paid attention to it in the domestic and overseas. Firstly, on the base of requirement analysis, a design is given. The system is composed of two parts server and client. From the bottom to top, the server includes 4 layers: a) hardware layer, b) operating system layer, c) a variety of runtime libraries necessary to the application, d) application layer. The application captures video, encodes as H.264, analyses video content and transmits the video and the results of video content analysis to the client. On the other hand, the client receives, decodes, presents video and processes the results of analysis. Secondly, a video surveillance system is implemented in an integrated manner. It captures video data, via DSP built-in CPU encodes the data as H.264 stream. Utilizing protocols such as RTP, RTCP and RTSP, video data are packaged and sent the client. Thirdly, in the embedded video surveillance system, a method is designed and implemented for analyzing and locating vehicle license plate. The original image is preprocessed (such as histogram equalization and binarization) in RGB and HSV color space to get two binary images, and then the candidate regions that may contain the plate are extracted from the binary image of HSV space. Next, the candidate regions are positioned one by one, and the plate boundaries are determined. As a result, the plate is located. Further the more, the characters are segmented by projecting the pixels in the plate image. Finally, the system is tested. The results show that it can correctly capture, encode and transmit video data. The license plate location method can locate and extract the license region and accurately segment the characters in the image with complex background. The system has high flexibility and performance, and it is applicable in vehicle management.