Video Recognition Technology Based On Knowledge Of Features

In the field of computer research, with the development of digital image processing technology, the surveillance video is developed toward the intelligent and smart direction. Although the surveillance video system has improved through three generations of evolution, most current monitoring systems still follow the traditional way, which need people to monitor and identification. With the continuous expansion of surveillance cameras application range, a substantial amount of data is increasing in the surveillance video, while the traditional way is not only greatly consuming time and human resources, but also more easily affect the results of the monitoring and identification accuracy due to reasons like human fatigue, etc.In this paper, based on the environment of popularity of current monitoring and preset features, recognizing the targets which match the features in the videos. This paper compares the common moving target detection algorithm, and proposed better algorithm for surveillance video for monitoring and identification, which is based on the codebook recognition algorithm. This paper also collected three different moving target detection algorithm and combined with the contents of the color image feature extraction and matching technology to develop out a video surveillance video identification system, which is suitable for a variety of environment. This identification system can be used in three different target detection algorithm based on the user preset target characteristics of surveillance video to identify and target tracking marks which matching these criteria in the video. According to a sample video test, the results show that the codebook-based moving object detection algorithm is applicable to most surveillance video detection, which also achieves the best effect and high accuracy in the object recognition. The frame difference algorithm could maximum operating efficiency, accuracy stability, adaptability, while the Gaussian mixture model-based algorithm has a more stable background to identify the effect of changes in the surveillance video.This paper focuses on the theoretical basis of the video recognition technology and give a detail thorough review. After that we also developed a feature-based surveillance video identification system which can be applied. Based on the system testing, we found that the system is functioning properly and achieved a highly recognition accuracy and operational efficiency of the algorithm, which could also be applied to different monitoring video detections.