Research On Recognition Of Road Traffic Sign Based On In-car Camera

With the development of intelligent vehicle,advanced driver assistance system(ADAS)has become an important part of intelligent transportation system(ITS),which helps the driver to perceive the danger in advance trough perceivng and understanding the road environment.In advanced driver assistance system,recognition of road traffic signs is one of the most basic and challenging tasks.The research of recognizing the traffic sign mainly focuses on the roadside traffic sign currently,this thesis focuses on how to use the method of computer vision to recognize the road traffic sign including the following works:(1)A recognition method of road traffic sign is proposed based on the gradient direction histogram(HOG)and support vector machines(SVM).The method extracts the region of interest(ROI)based on the position relationship between lane and the road traffic sign,then the gradient direction histogram is used as the feature extraction method,and the support vector machines classification model is used to identify the type of traffic sign.The method of lane detection adopts the constraint LSD algorithm,which is based on the LSD line detection with the introduction of the lane angle,brightness,and length constraints.(2)A recognition model of road traffic sign based on deep learning is constructed.In this model,the AlexNet network is used as the basic network to extract the feature of the image,combining with regression theory and anchor mechanism to predict the location and classification in the multiple scale features of each position of the whole image and adapting non-maxima suppression algorithm to eliminate redundant detection boxes.(3)In order to test the performance of the recognition algorithm for road traffic sign based on deep learning,the method based on the gradient direction histogram and support vector machines are utilized to collect partial samples.The two algorithms are carried out the contrast experiments on the same data set,which proves that the two algorithms can be used to identify 7 kinds of common road traffic signs.The experimental results also show that the recognition algorithm of road traffic sign based on deep learning has a higher correct recognition rate and robustness,which verifies the superiority of this algorithm.