The Occlusion Remove Research In Video Intelligent Traffic System

With the acceleration of urbanization, traffic congestion became more and more serious. To solve this problem, traffic facility construction is not enough. Enhancing the level of transportation control and management by Intelligent Traffic System (ITS) under the existing condition is a solution. Video ITS system has been used in America, Europe, Japan, Singapore and several cities in China.Object occlusion depressed the accuracy ratio of video surveillance system. A lot of projects with occlusion remove function have been applied by academicians and laboratories. The vehicle occlusions are classified and analyzed in this paper. Corresponding solution projects are brought forward for occlusions under different surveillance conditions and traffic environments.Firstly, the camera calibration algorithm is lucubrated to make sure the translation between image coordinate and real coordinate without camera inner-parameters, which insured the experiments on traffic video and improved the system precision.Blob tracking is a traditional real-time algorithm, which is easy to understand but cannot avoid the occlusion influence. Color histogram based Mean-shift tracking combine with Kalman predictor is good at occlusion remove. The blob score system and life parameter are proposed for occlusion estimation to colligate these two methods in odd-camera system. The experimental results indicate that our new system has satisfying veracity and real-time performance.Multi-camera tracking also can remove the occlusion influence. The calculation speed of three-dimensional reconstruct system is too slow for real-time performance. We use two or three camera shot the same traffic scene from different angle to gain more object information. After logic analysis, the system also can eliminate voluminous occlusion.Key problems in our work are discussed in the last section.