Research On Urban Road Traffic Sign Detection And Recognition Algorithms

In the development of intelligent transportation systems,the detection and recognition of traffic signs is an important part of the research of this system.It can provide real-time road sign information to drivers and driverless cars.However,due to changes in light,Problems such as scale changes and ambiguity of signs have brought great difficulties to the system,so researching a real-time and robust traffic sign detection and recognition system has become a hotspot in the field of intelligent transportation research.Therefore,this article mainly studies the image enhancement of urban roads and the detection and recognition of road signs.The main research contents are as follows:Aiming at the problems that urban road image enhancement algorithms are prone to cause color distortion,less local information and reduced contrast,an urban road image enhancement algorithm with gray scale adjustment and Gamma correction histogram reconstruction is studied.First,use the image current value,average value,maximum value and scale adjustment coefficient to obtain the grayscale adjustment coefficient of the image,and adjust the grayscale of the image;then,the average value of all the peaks of the single-channel histogram is used as the cropped peak Gamma correction is performed on the cumulative distribution function to obtain a single-channel enhanced image;finally,the enhanced image is obtained by using the proportional relationship of the three channel gray values of the image.Experiments show that this algorithm can improve the visual quality of images,with clear details and good contrast.In order to reduce the accuracy of urban road traffic sign detection algorithms due to changes in lighting,low contrast,and sign scale changes,a traffic sign detection algorithm with improved IGist features in the RGB color space was studied.First,obtain an NRGB image in the improved RGB space,and then use the cumulative distribution function of the NRGB image to calculate the threshold to binarize the image to obtain a candidate area for thetraffic sign;then,extract the boundary points of the candidate area in the image and use geometric symmetry to further detect Candidate area;Finally,use IGist feature and support vector machine to get the real candidate area of road traffic sign.Experiments show that the algorithm improves the real-time and accuracy of road traffic sign detection.Aiming at the problem that the change of the scale of urban road traffic signs leads to the reduction of the recognition accuracy of the system,an improved convolutional neural network is studied to realize the recognition of traffic signs.First,the RGB image and grayscale image of the input image are reconstructed to form an RGBG image as the input image.Then,the network is multi-layered convolution and the scale features are fused to serve as the input features of the final classifier.At the same time,use Mish activation function increases the nonlinearity of the system and improves the system's ability to classify images.Finally,through experiments on the data set GTSRB,the recognition accuracy rate reached 98.77% and the average recognition time was 1 ms,which effectively improved the real-time and accuracy rate of the traffic sign recognition system.