Research On Identification Technology Of Important Traffic Signs In Haze Weather

Traffic signs,as a kind of pavement facilities which used corresponding symbols,words and different background colors to guide,indicate,limit and warn road traffic participants.Traffic signs are an important part of the traffic system.Once the driver or the corresponding automatic/auxiliary driving system ignores the corresponding signs due to the complexity of the actual traffic scene or they are influenced by bad weather-especially the important signs,it will lead to the violation of the corresponding traffic laws or the traffic accidents,resulting in casualties and property losses.Therefore,important traffic sign recognition is an important part of intelligent transportation system which is a hot research field in the current,it not only has a more important academic significance,but also has a higher social and economic value.However,due to the complexity of the specific environment and the increasingly serious haze in recent years,it puts forward higher requirements for traffic sign recognition and adaptability for the actual application environment.In this paper,through the research and improvement of the previous algorithms,aiming at the haze weather which is one of a specific weather environment.The first step is to quickly defog the collected image.Then the traffic signs are segmented in HSV color space.Finally,the corresponding algorithm process of traffic sign recognition in fog is carried out by convolution neural network,which improves the accuracy and recognition speed of traffic sign recognition in fog.The main works of this paper are as follows:The existing algorithm can not locate,segment and recognize traffic signs in the haze weather environment by a directly and effectively way.This paper think that the first step is photo fogging.At the same time,the original intention of most defog algorithms is to reconstruct the real image,so it requires a higher overall image restoration effect,and the algorithm takes a long time.In this paper,we focus on the specific object of traffic sign,and the improved convolution neural network is used in the recognition part.The tolerance of input data is much higher than the traditional method.And the restoration of other objects in the image is not concerned in this paper,so the paper put forward an improved defogging algorithm.It has less computation and provides images that meet the requirements of subsequent processing.At the same time,the location and segmentation of traffic signs are optimized,and the results are taken as the input data of the subsequent convolution neural network.An improved convolution neural network is proposed.This design optimizes the network.According to the characteristics of classification and recognition tasks,the ascending dimension convolution are used in the network.It deepens the network structure without increasing a large number of parameters.At the same time,it puts forward the optimization method for the whole positioning and recognition process,which improves the accuracy of positioning segmentation and network recognition classification.At the same time,the actual ability of the network is tested by using the data collected in the actual road of China,and the control experiment of the fog group and the non fog group is designed.Experimental results show that this method has higher recognition accuracy.