Abstract :
Massive video data analysis has been a major question for discussion in the Internet age. This study proposes an approach to analyze video data parallel based on Spark. Firstly, this paper gives a brief introduction about application of Hadoop and Spark in the massive video data analysis. Then, video data analysis algorithm and parallel implementation method are described in detail, including feature extraction, clustering, bag of features, etc. Finally, performance evaluation experiments of video action detection and near-duplicate video retrieval on a cluster demonstrate the efficiency of approach to analyze massive video data in this paper.
