Research On Recognition System Of Road Traffic Signs Based On Deep Learning

Due to the substantial increase in the number of motor vehicles in recent years,it has caused many traffic safety problems,and road traffic safety has attracted widespread attention.Road traffic sign recognition is the basis of smart car environment perception and necessary conditions for realizing other functions in driving assistance system.It plays a vital role in car safety.And this is of great significance for realizing automatic driving of vehicles,perfecting intelligent transportation systems,and promoting the construction of smart cities.Road traffic sign recognition is divided into two parts:traffic sign recognition and lane line detection,and traffic sign recognition is'divided into two steps:detection and recognition.This paper uses the YOLOv3 deep learning target detection model method and redesigns the loss function in the model to achieve more rapid and accurate traffic sign detection.By adding data with typical interference characteristics to the standard GTSRB data set,the data set is augmented.And th e algorithm can accurately identify road traffic signs in the presence of environmental interference problems such as motion blur,occlusion,and abnormal lighting.So the anti-interference ability of traffic sign recognition algorithm has improved.This paper applies DarkNet-53 deep learning image recognition method,redesigns the classification module structure in the model and uses the Focal Loss loss function with class adaptation to train the model.In the end,higher recognition accuracy and faster recognition speed were obtained,Use this method can realized accurate traffic sign recognition.To solve the problem of long execution time and low accuracy of the lane detection algorithm using basic Hough transform,this paper improves the performance of Hough transform by combining the probability theory and perspective transformation method,and achieved fast and accurate lane detection.In addition,writing software according to the framework of the road traffic sign recognition system to test algorithm,and get a complete road traffic sign recognition functions.Simulation results show that the algorithm in this paper improves the accuracy of road traffic sign recognition and the real-time performance,so this paper has certain theoretical research significance and engineering application value.