Detection And Recognition Of Road Traffic Signs Based On Machine Learning

With the rapid economic development,urban traffic congestion is worsening and has a serious impact on the development of cities.Intelligent Transportation System(ITS)develops rapidly.As an important part of ITS,road traffic sign recognition system plays an important role in the areas of assisted driving.As an important research direction of intelligent transportation system,the research of road traffic sign recognition is also deepening.At present,the pursuit of accurate,real-time and effective methods of road traffic sign recognition is also more perfection.The road traffic sign recognition technology is great significance to the development of the entire intelligent transportation system.On the basis of summarizing the research status of road traffic sign recognition technology,this paper also analyzes the difficulties and methods of road traffic sign recognition technology and studies.This paper improves the road traffic sign detection algorithm and classification algorithm.The main work of this paper:(1)Construction of road traffic sign data set.The original data set used in this paper is the official GTSB data set.Data sets include a classification data set and a test data set for classification and target detection experiments.(2)This paper introduces the application of traditional algorithms in road traffic sign recognition,and introduces the image enhancement technology to make the target area more accurately extracted.Then the color space model and the shape detection technology are used to extract the characteristics of road traffic signs and the method of HOG and SVM is used to identify them.(3)A traffic sign detection algorithm based on R-CNN is implemented.Aiming at the problem that the traditional region-based algorithm based on color and shape information is too sensitive to color and shape the changes,this paper selects region neural network(RPN)as region-based recommendation algorithm,which has better robustness and better quality of recommended region.In this paper,the convolution neural network is used as the target detection algorithm,the classification effect is good,positioning and more accurate.In the end,the whole convolution network is introduced and the whole convolution network is used for the experiment,which has a certain effect on the target detection speed.