Shear Detection and Key Frame Extraction of Sports Video Based on Machine Learning

Abstract

Due to the continuous progress of video technology, the improvement of network transmission speed, and the continuous emergence of new and effective video compression technologies, people have realized the sharing of various video information. The explosive growth of video data makes it urgent for users to classify, browse and retrieve videos quickly and effectively. Content-based video processing and retrieval, which is dedicated to content analysis, semantic classification, index retrieval and storage management of videos by using IT, has attracted more and more attention of scientific research institutions and researchers. Sports estimation can reflect the temporal correlation between adjacent video frames. This paper adopts a simple and flexible sports estimation algorithm based on block matching. Key frame is a key image frame used to describe a shot. Based on the complexity of the shot content, one or more keyframes can be extracted from a shot. This paper studies several key technologies of semantic-based video content extraction and analysis. The research mainly focuses on how to automatically extract and analyze video content, and realize semi-automatic or automatic analysis and classification of video data to meet the needs of retrieval. Finally, this paper proposes an algorithm to extract key frames according to the content changes, and combines the shot detection and key frame extraction algorithm to form a video browsing platform based on key frames.