Traffic Sign Recognition System Based On Shape Features

Nowadays,traffic accidents happening everyday,The number of casualties caused bytraffic accidents are increasing.How to improve road traffic safety has become a pressingproblem.To solve this problem,driver assistance systems are being used,this systems usingonboard cameras to capture video images which include real time road information,getsuseful traffic information for drivers through computer processing and feedback to thedriver timely,to reduce and avoid traffic accidents caused by the drivers’negligence.Trafficsigns recognition is an important part in the system.The road traffic signs includes warning signs、prohibition signs and directionalsigns.They all have specific colors and shapes to distinguish them from other objects, toachieve the purpose of reminding drivers and pedestrians.According to the characteristicsof the traffic signs.This paper presents a traffic sign recognition system based oncolor-shape feature.Firstly, HSV color model is being used to extract the suspicious trafficsign area from the whole image,then using morphological methods to filtering,after that,applying the geometric shape feature descriptors to extract features,finally,relenting theareas which are satisfied both in color and shape.In recognition stage,BoW-SIFT model isbeing used to extract feature points and selects LibSVM library for featuretraining,ultimately achieves the purpose of identification.In this paper, acquires a large number of real-time images to the establish a databasefor the experiments.The results shows that the propose identification method’scomputation is small, and is not sensitive to external illumination changes.Traffic signswithin the driver’s line of sight can be detected accurately and quickly.