Semantic-based Wireless Video Surveillance System

摘要

The technology of semantics is a hot research topic in recent years. Especially, it is extensively used for information retrieval and classification. For most of the traditional wireless surveillance is based on store-and-forward model, it is less active and inefficient in data retrieval. In this paper, we propose the wireless active video surveillance based on the semantic technology, and the basic model we used is shown. The wireless transfer model is based on pear-to-pear and the analysis model is objectoriented. In practice, we design the platform based on DSP and FPGA for image processing and storage. Then we recognize the objects by assembling difference image, and classify the objects according to their different features base on semantics ontology library we proposed. The user can retrieve the image by semantics ontology conveniently. In this paper, we describe the principle of ontology construction in detail. Finally, we have made experiment for proposed algorithm and get satisfied results. The rate of wireless transmission can reach 10-15 frame/sec under the condition of EDGE network, furthermore semantic technology greatly improve the efficiency, the accuracy rate of retrieval exceeds 96%.