A Survay On Scene Detection And Video Mining Techniques

Data mining is an emerging field of interest for research in particular video mining has attracted a lot of research minds because of the vast quantity of video data available in recent years. Video data mining plays an imperative role in competent video data management with the ever- increasing amount of video data. All along with the enormous amount of data, a quick method to mining semantic patterns is needed. The foremost problem on video data mining is the appearance of the object in a visual media. The quandary is complicated because the object can come into view in different customs in different parts of the video due to dissimilar imaging conditions, lightening conditions, back grounds and occlusions. It is extremely difficult to extract and combine the multiple types of audio and visual information that are incorporated in a video. The research conducted in. the field of visual data mining is not enough. All these variations make visual data mining more challenging compared to other data mining. The study is based on current trends in video retrieval. The survey includes the shot detection, key frame extraction, clustering, indexing, and video object retrieval based on similarity measures for color, shape, region, motion, texture, refinement and relevance feedback. In addition this paper also provides marginal idea for future research in this field. Asst Prof at Karpagam University, Coimbatore. Professor and Director in School of Computer Science and Applications, Karpagam University, Coimbatore. Email: skaarthi@gmail.com