Research on sports video retrieval algorithm based on semantic feature extraction
- PDF / 2,097,144 Bytes
- 15 Pages / 439.37 x 666.142 pts Page_size
- 56 Downloads / 186 Views
Research on sports video retrieval algorithm based on semantic feature extraction Cuixiang Guo 1 Received: 9 June 2020 / Revised: 29 October 2020 / Accepted: 10 November 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Traditional content-based video retrieval algorithms usually only used the low-level features of video images, and the content description is not enough, which leads to the unsatisfactory Retrieval results. This paper studied how to combine the low-level features and semantic features of video, improved the existing index structure, and designed an efficient sports video retrieval algorithm. The experimental results show that the proposed algorithm can meet the real-time requirements and improve the accuracy and recall rate compared with the existing methods. In addition, the proposed sports video index combined with semantic features is better than the existing index in both query time and query results. Keywords Video retrieval . Semantic feature extraction . Signalling . Vector index . Gaussian mixture model
1 Introduction Video has attracted more and more attention because it is intuitive and easy to watch. When searching video, people usually use keyword search. However, due to the rich meaning of video, the title is not enough to fully describe enough video information, and the title of video naming is often affected by the personal preferences of the publishers, which is not necessarily very accurate overview of video information, so people are eager to use content-based search to search video directly. Sports video is one of the most watched videos in daily life, so it is of great practical significance to establish effective retrieval of sports video. Taking soccer as an example, the user wants to watch the wonderful shots of goals, if the system can immediately
* Cuixiang Guo [email protected]
1
Shandong Polytechnic, Jinan 250104, People’s Republic of China
Multimedia Tools and Applications
return the goal video to the user through a corresponding goal image, then it will greatly improve the user’s experience. The difficulty of video retrieval lies in the accurate extraction of video semantic information and the establishment of efficient high-dimensional index structure. Due to the existence of semantic gap, although human beings may seem intuitive, it is difficult for computers to understand the semantic information described by video. In addition, the feature vector dimension of video description is often very high. It is a great challenge to establish an efficient index structure and provide real-time retrieval service in this high-dimensional vector space. The contribution of our works is: according to the existing domain knowledge of sports video, the corresponding semantic features are designed and calculated. Combining the lowlevel features of images, the existing video index structure is improved, and efficient search and sorting algorithm of video is designed.
2 Related works Content-based video retrieval generally includes three st
Data Loading...
