Target Segmentation And Tracking In Traffic Image Sequences

The Intelligent Transportation Surveillance System (ITSS) aims to extract the interested objects in the range of monitoring by means of a series of hardware and software method such as image acquisition, denoising, enhancement, shadows elimination and targets tracking. While images accurate segmentation and real-time vehicles tracking are the most important steps. In the paper, we have implemented a series of experiments that focus on urban roads traffic information extraction. The hardware platform is constructed by a digital camera and a personal computer.In order to establish a high-performance, stable hardware platform of Video Surveillance System, we choose a popular Gigabit Ethernet-based digital camera, after a deeply analyzed the advantages and disadvantages of various surveillance system solutions. Experiments show the superiorities of our solution, which lays a solid foundation for the following experiments.As the beginning stage of the study on traffic image sequences, Image segmentation determines whether the objects can successful be tracked, therefore it is necessary to develop a algorithm that compromise the segmentation accuracy and rapidity. By comparing the traditional optical flow method and frame differential method, this paper adopts an algorithm based on background subtraction of the traffic images. We realize the effectively segmentation of moving objects through a series of steps, includes the initial background establishment, dynamic background updating, background subtraction, threshold self-selection, as well as the morphological processing.Shadow elimination is the key technique in traffic video monitoring system. So it is valuable to construct a shadow elimination algorithm to segment the moving targets accurately. Based on the deep study on prevalent shadow elimination algorithms, an algorithm that adopt generalized RGB color model has been established. Experiments show that the proposed method is easy to be implement and accurate in segmenting the moving targets.Target tracking is another key technique, which aim to track interested targets accurately in the monitoring fields. By studying the algorithms include template matching, active contours model and Kalman filters,a new algorithm has been proposed based on active contours model Kalman filters. The algorithm realized the vehicles tracking among traffic image sequences. Experiments show its accuracy.