Research On The Detection Algorithm Of Road Traffic Signs

With the rapid development of road traffic in China,many traffic accidents have occurred frequently,and assisted driving systems and unmanned driving systems have emerged.The detection and identification of traffic signs as an important part of the intelligent traffic system has become a key research topic in the current traffic field.In summary,the detection and identification system of traffic signs consists of two parts: the division of traffic signs and the identification of classified traffic signs.This article mainly completes the following aspects:(1)Because the road conditions in the real environment are complex and changeable,the traffic sign images obtained must undergo pre-processing steps to be detected and identified.In this paper,Gamma correction and histogram equalization method are used to reduce the effect of light intensity change to achieve the purpose of enhancing the image.At the same time,the median filter is used to eliminate the noise in the traffic sign image,and wiener filter is used to eliminate the motion blur generated during image acquisition.Finally,the characteristics of thethree different color model spaces of RGB,HSV,and HSI are compared with each other.(2)Aiming at the problem of detection accuracy in road traffic sign detection system,a traffic sign detection algorithm based on color feature and Support Vector Machine(SVM,Support Vector Machine)fusion is proposed.The detection method first uses HSV three color channels to implement threshold segmentation and eliminates most of the extraneous regions by using morphological opening operations.Then fill the processed image and use the connected area marker to separate different objects to get the region of interest.Next,the Legendre moments and wavelet moments of the rough detection mark image are extracted.Finally,a serial feature fusion technique is used to obtain the combined optimization features.The SVM classifier is trained using the combinatorial moment feature to accurately detect the rough detection mark and determine its shape.From the simulation results,it can be seen that the fine detection of traffic signs based on color features and SVM fusion effectively improves the accuracy of image detection.(3)For the classification and identification of traffic signs,the CNN(CNN,Convolutional Neural Networks)is used to classify traffic signs.First of all,the two-stage preprocessing of size classification and mean removal is performed on the classified images.Then we construct the CNN model,which is the VGG-16 model,including a 3×3 convolution kernel with convolution kernels,a pooling layer with maximum pooling,a Dropout layer,and a fully connected layer.The Re LU activation function was used to classify using the Softmax classifier.Finally,the GTSRB traffic sign image is selected as a data set and trained and tested based on the tensorflow deep learning framework.Simulation results show that the algorithm can improve the recognition accuracy of traffic signs.