Preceding Vehicles Identification And Distance Detecting Based On The Infrared Image

In order to solve the problem of traffic safety, Advanced Driver Assistant System(ADAS) has become a hotspot in the research of the Intelligent Traffic System(ITS). The identification of the vehicles ahead and distance detecting technology are the important parts of the bump-shielded system. When the driving conditions are bad, for example, driving at night or driving in fog, the vehicle detection technology based on visible light is limited; therefore, the infrared thermography technology which was used in the military market is gradually come into the commercial market. This paper studies the front vehicle identification technology based on infrared image and the distance detecting technology based on monocular vision.Most of traditional vehicle identification methods are based on the static background and moving target, such as the frame difference method and the background subtraction method. They have a relatively high amount of calculation, due to the global features are used. Based on the study of the dynamic clustering algorithm and make full use of road environment auxiliary information, this paper presents a kind of the target vehicle ahead identification method based on clustering analysis, and then to detect the target vehicle ahead base on the machine vision distance detecting.This paper firstly studies and simulates the methods of infrared image pre-processing, and then based on the simulation results we give an infrared pre-processing scheme, which is suitable for the target vehicle identification. By the median filtering and grayscale linear increase, the noise of the infrared image is greatly suppressed and the contrast between the target vehicle and the background image is obviously enhanced.After studying the front vehicle infrared imaging features, using the idea of AND operation of high frequency information estimation and environment estimation, we proposed a target vehicle interested region extracting method based on the auxiliary environment information, which can remove a lot of noise outside the target vehicle. Through analyzing interested regional characteristics, we give a front vehicle identification method based on clustering analysis: identify the target vehicle based on significant testing, ISODATA clustering analysis and vehicle infrared detection algorithm of clustering areas. The comparison results show that the proposed algorithm has the characteristics of higher identification accuracy and anti-interference performance than the traditional methods.After the analysis of characteristics of infrared target vehicle tail, a kind of distance feature point extraction method of the infrared target vehicle ahead is given based on the target vehicle ahead of the effective identification. And then, we give a passive ranging method of the target vehicle based on the distance feature points and monocular vision.The experiment results under different road scene show that the front vehicle identification algorithm based on clustering analysis presented in this article has a good performance in simple and light traffic scene compared to the traditional ones.