Application Research Of Traffic Sign Recognition Technology In Unmanned Driving

Unmanned driving is one of the important research hotspots in the field of intelligent vehicles,and traffic sign recognition is the key link of unmanned driving.In this paper,the traffic warnings,warnings and instructions are divided into 15 categories.The support vector machine and convolutional neural network are used to study the identification of traffic signs.The traffic sign recognition model is obtained through training and designed and implemented.Traffic sign recognition system.The main work of this paper is as follows:1.Pre-processing,segmentation and positioning of traffic sign images.Firstly,the traffic signs are preprocessed by grayscale,histogram equalization,filtering and morphological processing,and then segmented based on the Lab color space k-means clustering method.Finally,the traffic signs are completed based on the shape characteristics of the three traffic signs.Positioning.2.The traffic sign recognition based on feature serial fusion and support vector machine is realized.The HOG features and Gist features of the traffic signs are extracted separately,and the two features are combined by means of serial fusion.The traffic sign recognition model is obtained by SVM training.The average recognition accuracy of this model is 85.3%.3.Realization of traffic sign recognition based on fine-tuning of convolutional neural networks.The ResNet50 network model is fine-tuned so that its output nodes correspond to 15 traffic signs,and the last layer is retrained to obtain a traffic sign recognition model.The average recognition accuracy of this model is 94.4%.4.Design and implementation of traffic sign recognition system in unmanned driving.The system realizes the functions of preprocessing,segmentation and positioning and identification of traffic signs.This paper gives the results of the system operation,on the basis of which the advantages and disadvantages of the system are analyzed.