Design And Implementation Of Video Copy Detection System For Internet Application

Content based video copy detection is an important issue in multimedia information processing field. The difficulty of this technology is how to extract features from image quickly and accurately. Due to the great efforts of the academic community, content based video copy detection system has achieved prefect preference in small dataset. However, as the Internet and big data technology developing, large scale based video copy detection became a serious problem. How to search in vase amounts of videos still leave us great challenges. This article use the state of art feature extraction, vector search, and time align technology realize a powerful video copy detection system. The main content of this dissertation is as follows:1. Systematic analysis of the characteristics of the SIFT features, and propose a filtering algorithm based on maximum entropy theory, which can greatly reduce the storage space, and improve the retrieval efficiency.2. Improve the retrieval speed and accuracy by using Cartesian product quantization base algorithm and vector residual information.3. Introduce Hough transformation algorithm, and use it in our post-processing subsystem which make perform excellent in time-boundary align stage.4. Built experimental retrieval system and construct a new dataset for Internet environment. Show the good performance of the system on the large scale dataset.