Research On Video Event Detection

Video surveillance has become a hot and difficult topic in the field of computer vision in recent years due to the fact that surveillance video for public security usually contains complicated anomalies.How to effectively generate video description without human supervision has become a recognized hard problem by both industry and academy.The purpose of video event detection is to automatically extract various forms of video event information from video.The related results are of great significance to applications such as intelligent monitoring and human-computer interaction,as well as to promote the combination of theory of computer vision and artificial intelligence with practical applications in some important areas.This article first summarizes the traditional computer vision method of video event detection,and then introduces the recent popular video event monitoring method based on deep learning neural network.The main contents of the thesis include the following three aspects: 1)The structure of long and short term memory unit and the basic theory are expatiated,and the working principle of the encoder-decoder model based on two layers of short and long-term memory neural network is expounded emphatically,as well as its application in video event detection.2)A video event detection algorithm based on long and short term memory neural networks is proposed.The algorithm uses long and short term memory neural network to deal with the advantages of input / output sequence type data,which effectively avoids the dependence of traditional event detection methods on target detection and tracking,and realizes the video event detection problem based on multiple tags.3)I build a concurrent event database,experiment has been done on the above structure in the corresponding database,and make a great deal of evaluation and analysis on the experimental results.At the same time,compared with the traditional support vector machine event detection method,which shows the superiority of the proposed method.