Research On Behavior Recognition In Car Wash Centre

With the fast development of the automobile industry, car wash shop put forward higher requirements for video surveillance. The car wash shop owner wanted a specifically intelligent monitoring system. Can carry on the statistics of the number of vehicles the washed. And automatic identification of different types of services, in order to complete the intelligent charging work. In this monitoring system if we can make good recognition for staff characteristic action, Then we can according to the employee's behavior to determine the specific service. So it becomes the key of the whole system of car washing behavior recognition.Due to the wash room light is dark, and the monitor screen resolution is not high. The monitoring image is not clear. So we need to image processing operations. Through gray stretch and other pre-processing operation, highlighting the region of interest and useful information and filtering out the recognition of useless information. Contrast optical flow method, frame difference , background subtraction method. Analyzes the effect of moving foreground object detection. Around the background modeling of background subtraction method, analysis and comparison of three commonly used background modeling: Mean filter background modeling, based on the single Gauss background modeling and based on the mixed Gauss background modeling. Combined with the advantages and disadvantages frame difference and background subtraction method . Using an improved outlook for target detection algorithm. So that the effect has been to enhance the prospects for extraction.Finally, five kinds of simple washing behavior were identified: vehicles entering or leaving,moving around the car , wipe, waxing and rinse. Analysis from the foreground object characteristics and trajectories. Vehicles entering or leaving detect based on the characteristics of the foreground object area. For wipe the body behavior calculating staff external rectangular target area and centroid. Their frequency changes meet certain conditions. In addition use of the detergent product white foam in the process, the detection of foam as a necessary condition for wipe. In this paper, innovative use of morphological methods to detect foam. Frequency change of the circumscribed rectangle and the center characteristics, the detect of foam can distinguish between the behavior of waxing and wiping. Because rinsing use spray gun produces a strong mist, it different from the normal image. Mist will rise the amount of white light contain, making the saturation of the imagefall. Also, when rinsing the car body, get the prospect of moving target grayscale image.Improve binarization threshold significant change in the region will occur. This is also an important basis for rinsing behavior is different from other behaviors.This paper is aimed at a car wash to do washing behavior recognition. It is also applicable to other similar car wash center,car repair factory etc.. In this paper,the behavior of a single car and an employee were studied. Temporarily it does not applied to more worker or more complex environment.