| ABSTRACT:Nowadays, with the rapid development of rail transportation, urban rail transit construction has attained great achievements. As the urban rail transit system is becoming increasingly complex, the high density of passengers is turning greater challenges for the safety management systems. However, it is used the ticketing system to monitor the passenger volume on platform, and there are no specific devices to monitor railway passenger volume in metro vehicle. It is not satisfy the installation requirement of existing video traffic monitoring system because of the limit of visual range. This thesis concentrates on the research of target detection and tracking algorithm to solve the visual range limit and occlusion. The main work and achievements cover the following:It is used the camera by vertical acquisition to detect the passenger in railway cabins. Through the comparison among frame difference, optical flow and background differencing, it focuses on the target detection method based on Gaussian background model, and researches on background modeling and update method, to reduce the influence by illumination and jitter.A method based on the normal vector, described by spherical coordinate parameters(θ,φ)is proposed to detect heads. Canny operator is used to extract the target head contour. Furthermore, it is employed the least-square method for ellipse fitting to extract more complete heads and OTSU for adaptive threshold to improve the segmentation accuracy.Researching on the fusion of Camshift and Kalman to track exactly after target occlusion by using Kalman prediction to calculate the optimal position and update the observed value. Besides, four linear statistical model is set up to count passengers while getting on or off by the method of path analysis.Finally, a passenger statistics test platform is set in laboratory to test the algorithm and statistics method of passenger volume, and verify the feasibility of system function realization. |
PDF Full Text Request