Traffic Flow Analysis Based On Video Object Segmentation And Tracking Algorithm

Intelligent Transportation Systems (ITS) are the trend of future transportation system. Video based traffic flow analysis methods which are proposed in recent years have many advantages over traditional methods, and become an important part of ITS. Through monitoring the traffic flow by video camera, the system can numerate the vehicles on the road in real time. Other subsystems of ITS can make use of this information for further processing.Video based traffic flow analysis system mainly consists of two parts: vehicle detection and vehicle tracking. The detection part aims to segment vehicles from the video sequence and the tracking part tracks each vehicle and counts.In this paper, based on the research of key technologies, we propose a new solution to implement the traffic flow analysis system. Our work mainly contains:1. Summarize the key technologies which are widely used in video based traffic flow analysis system, and make contrast of them.2. Propose an advanced multiple video object segmentation algorithm in vehicle detection.According to the street scene environment, we propose a novel multiple video object segmentation algorithm which incorporates object recognition information in vehicle detection step. The algorithm consists of training section and segmentation section. At the training section, it uses codebook method to train the vehicle images: First, it uses Sift method to extract the features of vehicle image samples. Secondly, it segments the images into small patches to cluster. Third, it uses a new way to activate the codebook which can largely reduce the numbers of vehicle sample images. At the segmentation section, it uses an Implicit Shape Model to combine the recognition knowledge and the segmentation knowledge together.3. We propose a new solution to the traffic flow analysis system and implement it with a tracking algorithm.In vehicle tracking step, we search the shortest frame distances among objects of neighboring frames which provided by the object segmentation step, and use color features as assistant information to track. At last, we combine the segmentation algorithm and tracking algorithm to implement a prototype of traffic flow analysis system.