Research On Detection And Recognition Algorithm Of Traffic Sign Based On Machine Vision

With the extensive development of electronic technology and computer technology which has promoted revolutionary change in the automotive industry.Nowadays,words have become a hot topic in modern society,such as smart cars and driverless cars.The traffic sign detection and identification technology is one of the key function of driver assistant technology in smart cars to ensure safe driving of drivers.The detection and identification of traffic signs refer to the detection and recognition of the traffic signs in the current driving road,and then provide feedback the sign information to the driver to ensure driving safety in a timely manner.By studying the prohibition signs,warning signs and indication signs on driving lanes.We put forward a stable,efficient traffic sign detection and recognition algorithm.The main contents of the study are as follows:(1)For the images of traffic signs collected on roads,there will be varying degrees of distortion due to environmental and shooting angle.The appearance of distortion will cause certain influence on the detection and recognition process of the mark,and even lead to errors of detection and recognition,so it is necessary to deal with them.This paper converts the dark image with weak light to the HSV color space and the V component.The histogram equalization of V component is adopted to improve the dim image.For the distorted image,an affine transformation based algorithm is used to correct the contour of the traffic sign area in the shape detection stage,then making it easier to detect.These methods solve the common problems of image distortion.(2)The algorithm in this paper is divided into two parts: detection and identification.In the detection stage,this paper mainly detects two main features of traffic signs,color and shape.For the color information,the HSI model color segmentation is performed on the red,yellow and blue colors of the image to obtain a binary image.For the shape information,it is determined that each of the connected regions in the binary image satisfy the conditions of a circle,a rectangle,and a triangle.There are some other ways to make the transition between color detection and shape detection.(3)In the recognition stage,this paper uses the recognition method combining support vector machine and similarity measuring.The entire algorithm divides all traffic signs into five sub-categories by color and shape.The detection stage can already determine the area of interest in the road image.All categories directly use the SVM to classify all the traffic signs in the specified subclass.After the results are obtained,the image of the region of interest is similarity measured to the template,and the resulting similarity is used to make a final determination using the threshold.This greatly improves the recognition accuracy and efficiency of traffic signs.(4)The traffic sign detection and recognition algorithm is verified experimentally.In this paper,the GTSRB traffic sign dataset and self-collected traffic sign datasets are combined to simulate test methods.The dataset images are divided into five categories to detection and identification,according to their colors and shapes.The final test and recognition results are statistically obtained,and the final recognition rate,up to 96%.The experimental simulation shows that this algorithm has high accuracy,good stability and strong robustness.