Traffic Sign Recognition Algorithm Research Based On Deep Forest

In recent years,with the rapid development of economy and technology,cars have become more and more common in People's Daily life,bringing great convenience to people's travel.However,the increasing number of cars and the increasingly complex traffic network also bring a series of traffic congestion and safety problems.Therefore,the intelligent transportation system emerges at the historic moment,and the automatic driving system has been greatly developed as an important part of it.As one of the key technologies of autonomous driving,traffic sign recognition has become a research hotspot in the field of machine vision and pattern recognition,attracting the attention of many experts and scholars and relevant scientific research institutions.However,there are many kinds of traffic signs and complex traffic environments such as changeable weather,uncertain light intensity,similar background or occlusion of signs in natural scenes have brought a series of challenges to traffic sign recognition.On the basis of summarizing a large number of domestic and foreign related researches,this paper focuses on the algorithm research of traffic sign recognition using deep forest and other related knowledge.Traffic sign recognition research work of this paper is divided into two stages traffic sign detection and classification of traffic signs,the main work includes:(1)in aspects of preparation,the basis of traffic sign recognition for public testing data set of samples is less,made domestic traffic image data sets so that subsequent further validation of the algorithm.In addition,in the detection data set,the illumination condition is determined by the histogram,and the image with poor illumination condition is preprocessed by an adaptive equalization algorithm with limited contrast.(2)in the detection stage of traffic signs,a HOG feature and SVM detection method for traffic signs are proposed.This method firstly through statistical traffic signs after the color of the threshold of the image threshold segmentation,in addition to a large number of interference,and then detect the connected area biggest stable extremal regions,completed the traffic sign interested area of crude extract,finally HOG features of traffic signs interested area is extracted,and combining the SVM for binary classification to determine the real regional traffic signs.Experiments show that this method has great advantages in detection and higher accurcy.(3)in the classification stage of traffic signs,a traffic sign classification method based on the deep forest is proposed on the basis of the random forest and deep forest algorithm.In this method,the traffic sign data is divided into multi-instance features by means of multi-granularity scanning,and the features are represented layer by layer through the deep cascade forest.Experimental results show that the algorithm is effective in accuracy and operation time.