Research On The Method Of Vehicle Behavior Cognitive In The Limited Scene

The traffic growing accidents and the increasing demands for transportation safety have caused the rapid development of the intelligent transportation system. In this aspect, the research on intelligent vehicles has been a hot topic. Recognizing the movement behaviors of the anterior vehicles can provide real-time traffic information for the following intelligent vehicles. This can effectively avoid the happening of traffic accident. So for unmanned driving and auxiliary driving, the movement behavior recognition is always an attractive topic.In this thesis, to track vehicles reliably and identify lane lines accurately, we improve some common algorithms of vehicle behavior analysis. The main contributions are:1. An improved vehicle tracking algorithm is proposed. In the proposed method, the symmetry of vehicle rear is introduced for effectively eliminating the irrelevant information in traffic environment. Moreover, a pyramid-based optical flow tracking method is used for tracking the fast moving objects reliably.2. An improved Hough transform method is used for lane line detection. Here, we first establish some dynamic region of interest by setting some parameters, then the improved Hough transform is used to detecting the lane lines accurately in real time under the parameter constrains of the current frame to the next frame.3. The complexity of the used traffic scenes is limited and modeled to achieve the vehicle behavior recognition. Based on the limited traffic scenes, the vehicle behaviors, such as driveway keeping, line driving and lane change, are recognized.The extensive experimental results show that the proposed method can effectively recognize the vehicle behaviors in limited traffic scene.