Research On The Key Technolgy Of Consumer Behavior Analysis Based On Omni-Directional Vision

Currently, in order to analyze the behavior of the customers effectively, the acquisition of the information about the behavior of customers is performed by lots of stores in a kind of manual way. The clerks need to follow the customer and make the observation on his/her behavior and keep the related information recorded. The limitation of this method lies in the higher cost of manpower and the results with more or less subjective judgments. To mitigate the sense of the distrust and the disrespect from the customers and lower the cost effectively, a system of consumer behavior analysis based on omni-directional vision is proposed in the paper, and the computer vision is utilized to gathering the information instead of the manual approach. By utilizing the equipment of ODVS, the tracking method of multiple objects, environment customization, relation database and the approaches of behavior analysis, a kind of ODVS-based analysis device on the purchasing behavior of consumers is implemented with low implementation cost, accurate and objective results of the investigation, as well as high level intelligence and automation. The related research work is listed as followings:1. In order to detect the multiple targets quickly and efficiently, this paper suggests a detection algorithm for multiple targets called MHoEI, which is based on the MHI algorithm and MEI algorithm. The MHI algorithm is adopted to detect moving targets and the MEI algorithm is used to detect motionless targets. The MHoEI algorithm confuses the advantages of the MHI algorithm and the MEI algorithm to detect the moving and temporally stationary targets.2. To track multiple targets successfully in complex scenarios, an object-oriented tracking algorithm for multiple targets is recommended in the paper. The proposed algorithm makes use of the multi-feature fusion method to combine several features of the target, and then the multiple targets are matched with the multi-target matching algorithm. And finally, the continuous tracking of the multiple targets are realized.3. In this paper, a system of consumer behavior analysis based on omni-directional vision is presented. Multiple consumers can be effectively followed, and the trajectories of consumers, the spatial location of consumers, the environmental information can be obtained, and then with those information, the important scene information concerned with customers’ movement and behaviors can be inferred.