Wheel Hub Overheat Detection Based On Far Infrared Image Analysis

The tires of trucks friction with roads for a long time, which causes excess temperature so that tires would be damaged and have tire punctured, seriously affecting the safe driving. As a result, it is urgent to detect the temperature of tire hub of truck on the highways. Currently, the tools mainly used to detect the temperature of tires are contact sensor and infrared thermometer, whose cost is high but with a low recognition rate. When the temperature of tire hub appears anomalies, truck drivers always tend to ignore the reminder of the unusual information of temperature, and it’s difficult to detect security risks. Therefore, it is significant to use image processing technology to detect the temperature of tire hub, and to warn truck drivers in time so that the temperature of tire hub will be dropped. In this paper, we separate the area of tire hub and other areas, and extract various features of the tire area so that we can discriminate whether tire hub is overheating. This method is less susceptible to outside inte rference, and is able to detect anomalies of vehicle and take measures to reduce casualties and property losses. The main contents of this paper are as follows:Firstly, this paper introduces the background and research significance of detecting the temperature vehicle hub. The process of discriminating whether tire hub is overheating or not is divided into three parts: d ivision of the target area, feature extraction and classification recognition. This paper mainly research on the method of feature extraction. Respectively from global and local aspects, we introduce the basic concept and development status of feature extraction method. Based on above all, we expound the method of multiple features fusion.Secondly, this paper introduces the concepts and steps of HOG feature and LBP feature. According to the relationship between the temperature and the brightness information, the method of scattering brightness difference descriptor is proposed. Extracting the brightness information of the target area discriminates effectively whether tire hub is overheating or not. Then the incremental learning algorithm of support vector machine(SVM) is introduced in detail. The judgment model of tire hub overheating with multiple feature fusion is proposed.Thirdly, frame differential method is used to extract the images of effective vehicle tire. Then we pretreat the extracting image, including tire hub extraction and segmentation of target area. Using principal compo nent analysis, the dimension of HOG features and redundant information can be reduced. More features in series make up the final vector.Finally, the results of experiment show that the SBDD feature which is proposed by this paper is a valid. And the three types of features provide different information about the images, so it is reasonable to improve the classification accuracy.