Research On Road Pedestrian Detection And Recognition Method Based On Vehicle Infrared Night-vision Goggles

With popularity of vehicles, car leads to traffic accidents easily at night or under bad weather conditions, such as fog. Infrared night vision technology has the characteristics of an obvious target relative to the background, imaging without the interference of light intensity. Its function in the driver assistance system can be used to expand the driver’s driving at night vision. In this paper, vehicular infrared night-vision goggles is mainly presented in the application of driver assistance system, and it is studied from the pedestrian detection, recognition and ranging technology on the vehicular infrared image separately.Firstly, the infrared imaging principle and the structure character of infrared night-vision goggles was used to deduce the interior and exterior parameters of infrared night-vision goggles, Matlab toolbox was used to correct the distortion of infrared night vision image in the paper. Secondly, Image enhancement method was designed for applying to vehicular infrared night vision image. The method established wavelet-genetic grayscale image enhancement algorithm, it used the non-contact infrared night-vision goggles to generate the principle of thermal image. Then, using image matching and compensating the three frame difference method, which established an improved three frame difference to detect the road pedestrian in view of the monocular vision and according to the geometry and movement characteristics of pedestrians, established geometry-movement rate method based on monocular vision to distinguish pedestrians. The method has the characteristics of high pedestrian recognition accuracy. Finally, pedestrian fast ranging method based on monocular vision was deduced according to two target camera imaging principle and driving characteristics of automobile, and was analyzed from the aspects of real-time and practicality. The experimental results show that the fast ranging method is effective and feasible.Wavelet-genetic grayscale image enhancement algorithm was designed in this paper, which can improve the image brightness uniformity, avoid excessive enhancement and effectively suppress noise. Geometry-movement rate method based on monocular vision achieves high pedestrian recognition accuracy. Pedestrian fast ranging method based on monocular vision reaches high real-time ranging accuracy. The simulation of infrared night-vision goggles pedestrians ranging system was planned starting from the practical application, it verifies road pedestrian ranging algorithm and lays a solid foundation for vehicular infrared pedestrian detection instrument product development.