Research Of The Vehicular Infrared Night Vision Video Processing

Traffic accidents occur quite frequently during night due to driver’s poor sight which ismainly caused by too close exposure of vehicular light and too weak of light. The frequentoccurrences of traffic accidents threaten people’s lives and property. The thermal imagingtechnique is based on the temperature difference between the target and its background. It can findtargets at night or in poor light circumstances and has many other advantages, such as withoutbeing interfered by glare or flash, and realizing far-distance, all-weather observation. Infraredimaging is based on the thermal imaging technique so that video enhancement, pedestriandetection and video colorization technology based on the vehicular infrared night vision video cangreatly improve driving safety at night and on special weather. It has very high practical value andimportant practical significance in reducing the incidence of traffic accidents.Aiming at the vehicular infrared night vision video, we study pedestrian detection and videocolorization technology at night in this paper. The main research contents are as follows:1.The paper firstly outlines the principles of infrared imaging technology and the feature ofinfrared video image, which can be targeted for subsequent processing.2. This paper proposes related image enhancement algorithms based on analysis of infraredimages and infrared video features which are for the implementation of pedestrian detection andvideo colorization technology on the hardware platform.3.Conduct research on common threshold segmentation algorithms used in infraredpedestrians images. According to the distribution of infrared image’s histogram, this paperproposes a vehicular infrared video pedestrian detection algorithm based on histogram dynamicthreshold. The algorithm has little computational cost, real-time property, good noise immunityand robustness.4.In this paper, due to the single scene characteristic of vehicular infrared night vision video,a method for vehicular infrared night vision video colorization is proposed which is based on image segmentation algorithm of watershed and colorization algorithm of prior color knowledge.The Experiment results show that the proposed algorithm of automatic colorizing vehicularinfrared video is effective and the proposed method will benefit the target identification for humaneyes and has a good real-time property.5.In this paper, after realizing pedestrian detection and colorization on the vehicular infrarednight vision video, an overall design framework of the vehicular infrared night vision videoprocessing system is proposed and its hardware implementation is finished. Finally, the abovealgorithms are implemented on our vehicular infrared night vision video system and achieve goodresults.