Dsp-based Remote Video Monitoring System

Remote surveillance system is a set of equipment which is designed to transmit real-time image data captured in a fixed camera on a certain scene to the remote host for human intervention or monitoring, in the meantime, it should be able to separate the target from the background, and provide information such as target coordinates and target trajectory. In conventional remote surveillance systems, when the camera captures video, the data is encoded into MPEG form in the embedded DSP system, and then the encoded data is transmitted to the host(PC, in general). After that, it is decoded from MPEG, for the use of target segmentation, identification, and display of the image. Although it can reach a very high compression ratio by using this method and it'll also get a good visual effect, to be easily displayed and inquired. However, the MPEG encoded video stream is not easy to be extracted and segmented automatically. When the host is not PC but embedded system or even portable equipment, as its restrict of system resources and operation capacity, it is impossible to achieve the goal of decoding and the target extraction. While the DSP terminal system, for its fixed position, is provided with more resources in order to improve the operation capacity.Firstly, a modified algorithm base on background subtraction is raised in this paper, together with the regional selection algorithm, image cutting and image fusing algorithm as follow-up algorithms, to realize the acquisition of both the focus and the trajectory of the target; in the meantime, by using regional segment transmission as well as LZW compression algorithm, it is easy to realize the compressed video stream transmission, and to achieve the savings of bandwidth and improvement of transmission efficiency, which will have the comparable data compression rate as the MPEG encoded ones. Secondly, simulation experiment to test the feasibility of the algorithm is done on PC. Thirdly, the experiment platform is also designed to transplant the algorithm into the DSP, with which experiment is carried out in the environment which is similar to the community.As the ability, time, and the experimental conditions are limited, analysis is inadequate in this paper. There are also many defects in the experimental system and the defects are also discussed in the relevant sections for further research and improvement.