Research On Key Technology Of Video-based Intelligent Transportation Surveillance System

Video-based intelligent transportation surveillance system is an important application in transportation industry, and has profound significance to the realization of automation of scene surveillance and traffic management. The key technologies of video based intelligent transportation surveillance are researched in this thesis, the main content are as follows:For object detection, a deeply discussion is taken, which is about theoretical basis of Gaussian mixture model(GMM) and its application。Based on the presented analysis to the producing process of object foreground's tail of the conventional GMM method in the paper, an online updated GMM moving object detection algorithm which can suppress tail is proposed with the effectiveness proved by experiments. For the nocturnal scene, a frame difference based method is used through searching the highlights block pairs since the high stability of vehicle headlight feature and the low visibility of other features.The essence of the online updated GMM moving object detection algorithm is motion information in the video sequence, the object regions generated by the GMM algorithm will include the vehicle's shadow in sunshine environment, this may cause error when tracking objects. So the shadow characteristics are analyzed in HSI color space. Based on this, according to Local Illumination Model, a shadow suppression algorithm is proposed, in which the intensity is used to find the shadow-like pixels first and then the edge features are used to get a estimation of vehicle position to remove the pseudo shadow pixels. Good effect is obtained in experiments.For object tracking, a dominant color and region block features based object matching algorithm is discussed. Then a object tracking algorithm is proposed based on multi features fusion by using the kalman filter. Some researches about nocturnal tracking are done in the paper as well.A video-based transportation surveillance target tracking software is developed, which can get the vehicle's trajectory. The performance test shows the software system's feasibility.