Research On Traffic Sign Detection And Recognition Based On Computer Vision

With the development of the social economy, automobile gradually spread, the road bears more and more pressure,so the Intelligent Traffic System emerges as the time requires. As an important part of the ITS, Traffic Sign Recognition system (referred to as TSR) is also of concern.Traffic signs are mainly remind the driver current road traffic information, road potentially dangerous,thereby improving driving safety,reducing accidents.Traffic sign recognition based on computer vision contains two parts called detection and recognition of traffic signs in natural scenes, mainly using the image acquisition device to get the images for detecting and recognizing.In order to improve traffic sign detection and recognition speed and accuracy in natural scene, The paper refers lots of relevant literatures and mainly completes the following works through a large number of experiments.1. Traffic sign detection based on color features.Convert traffic sign image from RGB to HSI color,determine the color threshold of traffic sign,including red,blue and yellow. Then use the region growing method to find the area of traffic signs included.2. Traffic sign detection based on RANSAC.According to their shape,the traffic signs are divided into circle,triangle and rectangle.Detect corners in foreground area obtained after color segmentation,using multi-RANSAC algorithm to locate lines and circles.Calculating line’s angle and the difference between line and line,to determine shape of signs,such as triangle and rectangle.3. Traffic sign recognition based on sparse representation.Extract the results of traffic sign detection in the original image as training samples for building over-complete dictionary.We viewed a test sample as the liner combination of training samples,use recognition based on sparse representation classify test samples into respective categories.Trough experimental verification about29kinds of common traffic signs, the recognition rate of91.6%correct.The experiment results show the proposed algorithm can achieve the robust recognition when illumination changes.