Research On Video Classification Based On Association Rules Of Action Semantics

With the rapid development of social computerization and powerful data collection and storage tools,data has exploded.Video data,as the most important carrier for recording human social activities,produces PB level video data every day.In the fields of video retrieval,video on demand,intelligent monitoring and so on,the manual processing of video information can not meet the needs of the people to find the information needed for the mass video search,and the video retrieval and detection containing rich semantics is an important hand for people to obtain video information,in order to allow people to be able to be high in a short time.It can obtain content information of interest in video,and can provide corresponding auxiliary decision for video users,and data mining based on video content emerges as the times require.Video data is unstructured data.Compared with other data,the semantic gap in video increases the complexity of video mining and retrieval.Therefore,the research of semantic association rules mining method has attracted the attention of researchers.Based on the research of video semantic association rules mining,this thesis studies the association rules mining algorithm based on the basic action concept,and studies it for the mining of the basic conceptual correlation of the corresponding complex video scenes or events,and the association rules as a complex video scene or event detection.Classification criteria are used to detect and classify complex video scenes or events.The main contents of this thesis are as follows:(1)Aiming at the problem of fast detection and classification of complex video scenes or events,a video complex action scene detection method based on basic action semantic concept association is proposed.This method uses Apriori algorithm to mining association rules for all the semantic concepts in the corresponding video scene.The association rules of the mining action semantic concept are used to detect and classify the video complex action scene,and the video complexity of the test video is considered in accordance with the related rules of the corresponding scene class.The detection and classification of the scene.The experimental results show that the proposed method can effectively excavate the correlation between the actions in video,and quickly and effectively realize the detection and classification of the concept of video complex action scene,and improve the accuracy of the detection and classification of the video complex action scene.(2)The mining algorithm of the transmission system can not take into account the appearance of the semantic concept of video action with time series,which will affect the more detailed analysis of semantic concept correlation,and proposes a mining algorithm based on time series weighted motion semantic association rules.The algorithm uses video temporal window to mine the association rules mining of video action semantics,introduces the weight factor of the concept rule of video Action Semantic in the timing window,and corrections the support and confidence of the rules,and forms a video basic action semantic mining algorithm that can reflect the time sequence.By studying and designing a video classification criterion based on temporal weighted action semantic concept association rules,the detection and classification of video complex action scenes can be realized.The experiment shows that the mining of video action semantic association rules based on time series weighting can get a more reasonable correlation,and can effectively improve the accuracy of detection and classification of video complex action scenes.(3)Using the proposed method,combined with the object-oriented design idea,using the C# and Matlab hybrid programming technology,a video detection classification prototype system based on the semantic association rules of video action is developed.The system consists of three modules: video data preprocessing,sequential association rules mining and video scene detection.The system runs to show the availability of the proposed method.