The Design And Development Of Image Recognition System On Railway Passenger Transportation Hub

Image recognition has a very important practical value and broad prospects for development in medical studies, traffic monitoring, traffic statistics, astronomical observation and other fields. Image recognition is currently a research hot spot. It includes a moving object detection, object classification and extraction, target tracking and counting, involving many core issues about image processing and pattern recognition. This paper studied the Railway passenger image recognition problem based on the project of "passenger safety state identification technology on comprehensive passenger transportation hub distribution network of the high-speed Railway".In this paper, in order to avoid the overlap or coverage problem as far as possible, it simulated a simple image recognition system. After getting the images and making them be gray, firstly, using the background modeling algorithm to establish the initial background, and to achieve the background update. This method can effectively inhibit the impact of illumination change; Secondly, to obtain the background difference image by the subtraction of the current frame image and background image; again, taking the efficient median filtering, morphological processing and other methods to handle the background difference image to clear the noise and some small invalid areas and extract the real passenger goals; Finally, count the passenger number by analyzing the connectivity of the image.Background modeling is a difficult problem in this system. On-site monitoring environment is unstable, and the illumination change can affect the environment, and it is very difficult to get the pure Background images because of the continuing passenger flow. So how to achieve the extraction and update the background image in the system is a difficulty. Aiming this problem, this paper proposes the average pixel value of multi-frame images as extracted background pixel, and proposes a new effective method of background updating.