Research Of The Traffic Sign Detection Algorithm Under The Complex Lighting Conditions

With the rapid development of social economy,the number of motor vehicles and drivers has increased rapidly and the problems of road traffic have become the focus of attention.The rapid development of road traffic brings convenience to people and also brings a variety of security risks and some unnecessary economic losses.The complex lighting conditions cause great trouble to the traffic sign detection in the automatic driving.How to weaken the influence of the complex environment on the traffic sign detection has become an urgent problem to be solved.In order to solve the above problems,this paper studies the traffic sign detection algorithm under complex lighting conditions.By reducing the negative influence of light on the traffic sign detection,it meets the requirements of real-time and accuracy of traffic sign detection system.The specific research work is as follows:First of all,this paper analyzes the research status of traffic sign detection.Through the study of the method of traffic sign detection,it is found that the problems of backlight interference,slow computation speed and not enough high detection accuracy are not solved.While the training speed of the extreme learning machine classification method is high,this paper apply it to traffic sign detection to enhance the traffic sign detection effect.Secondly,this paper presents a backlight compensation method based on fuzzy extreme learning machine based on gradient boosting.This is a pre-processing method of traffic signs under complex lighting conditions.Firstly,the color space of the image is converted from RGB to YIQ,and the two backlight factors are extracted by the HIST histogram and Otsu' method respectively to represent the brightness of the image.Secondly,the fuzzy extreme learning machine based on gradient boosting neural network is used to fuse the backlight.Finally,the backlight compensation curve is determined by the fused backlight,and the brightness of the image Y channel is adjusted by the backlight curve.Finally,the I and Q channels are merged and converted to RGB space.The experimental results on the self-mining data set show that the algorithm proposed in this paper has better compensation effect in backlight compensation than traditional algorithms,and can effectively avoid the over-saturation problem in other methods.Finally,in order to further speed up the speed of traffic sign detection,this paper presents a traffic sign detection method based on regional proposal network and kernelbased extreme learning machine.This method uses the region proposal network to extract the candidate regions.And then HOG features are extracted further from the Top N candidate regions,and the trained kernel-based extreme learning machine is used to sort and filter,and finally determines the location of the traffic sign.In this paper,the proposed method is experimented on the public data set of German GTSDB and the data set collected in complex lighting scene.The experimental results proved that our approach has good effect and has the characteristics of fast operation.