Research On Statistics-Based Human-Computer Interaction Video Data Mining System

In recent years, methods and technologies for structured data mining have been established. However, with the development of information technology, more and more data with complex types are produced. Many scholars and organizations are aware that more valuable information to people implicit in these data with complex types, and begin to transfer data mining objects to semi-structured or unstructured data. Video data mining is a very significant part of it.Currently, many scholars are researching on data mining from specific types of video, such as movie video, news video, and sports video, et cetera. And they have made some progress. Our research focused on surveillance video, analysis "mine what" from surveillance video base on the feature of them. Through the analysis of the current status and problems of research on the video data mining, we proposed human-computer interaction data searching methods, which are based on statistical principles. Then human-computer interactive data searching system was built. It could be used for video data searching in the form of human-computer interaction, and information for application and summarized laws or knowledge will be obtained. We made a study of the following aspects:1) Analyze of the background and significance of the project, understand the research status of data searching and video data mining, then analyze and summarize the research status to lay the foundation for the future study.2) Analyze video data and its type. As there are lots of types of data, video data mining methods and requirements are different following the variation of video data types, this study focused on the surveillance video. We analyzed structural features and the characteristics of the content of the surveillance video data to clear on surveillance video data searching requirements, the need to "mine what." By analyzing the characteristics of video and data mining requirements, the surveillance video data mining features and problems were summarized, and then human-computer interactive video data mining methods based on statistical principles were proposed.3) Video surveillance is widely used in all walks of life. Requirements of video data mining are not the same for different application. Due to limited time and resource, this study focused on surveillance video of the manufacturing, apparel retailing, and hotel industries, considering different industry characteristics and needs to determine the data collection methods, the sampling methods, and statistical methods used in the process of human-computer interactive video data mining.4) Requirements analysis and system design on human-computer interactive video data mining systems. This part includes system overall requirements analysis, overall function model establishment, and the detailed design of the four modules.5) Summary of the research results, inadequate analysis of the study, and conduct future research content and direction.