Hbase Based Surveillance Video Processing, Storage and Retrieval

Abstract

Due to the rapid data growth of video monitoring, how to efficiently storing and querying massive surveillance videos is challenging, such as performance of querying, and fault tolerance for storage. The emerging cloud computing and big data techniques shed lights to intelligent processing for large-scale video data. This paper proposes a HBase based approach for surveillance video processing, storage, and querying. We adopt a distributed storage architecture, cut videos to many small ones and stored them in HDFS, extract video data through Hadoop preprocessing. In our approach, a number o strategies are used, e.g. pre-building regions, multi-thread and row-key optimization, to write data into HBase cluster in parallel. Evaluations show that our method has good performance.