Research On The Early-Warning-of-Collision System Based On Stereo Vision

As one of the research topic of Intelligent Transportation System(ITS),early-warning-of-collision system has been drawn much attention in recent years. With the development of stereo vision algorithms and the improved speed of hardware, stereo vision is widely used in ITS systems.For decreasing experiment cost and improving future market competition power, this thesis designs and develops a horizontal stereo vision experiment platform through deeply research and analysis of stereo vision algorithms. And then, we propose a scheme including safe headway measurement, preceding vehicle detection and recognition for express way. Firstly, by study and analysis of camera imaging models and camera calibration methods, this thesis proposed a new calibration method based on the traditional calibration method, which obtained internal and external parameters by calibration experiments. Secondly, by the shade character of the front car, the different coordinates of the target car in the binocular images can be detected. Finally, a Secure-Vehicles- Spacing-Measurement(SVSM) program applied to expressway traffic was proposed, the distance between the target vehicle to the camera can be calculated by employing the common perpendicular midpoint method. The results show that the measurement program can measure a wide range of distance and can meet the needs of a SVSM for intelligent vehicles, which is an effective way for measuring the distance to the front car.Following the research of the algorithms, a real-time image processing system based on TMS320DM642 was designed in this thesis. The system has the function of real-time dual channel video acquisition, real-time image processing, video output and so on. The system was applied in previous collision warning system and can carry out image processing in the frame-by-frame way, the processing time of one frame was 0.1s.