ROAD SIGNS DETECTION AND RECOGNITION IN IMAGES

Abstract

Object Detection and Recognition in an image is the process of finding and identifying the object in it. It can be anything such as building, bicycle, sign boards or faces, etc. Road signs detection and recognition is one of the major application of object detection application. Major work is done on recognizing road signs and some methods are proposed in this research work. Generally, every object we want to detect has some features which make them unique and road signs also have unique features which is their color. This system is having three major steps: color segmentation, detection and recognition. Two algorithms are used for color segmentation, one is based on HSI color segmentation. Roads signs are generally of red, blue, yellow and white so applying certain thresholding will give the area of interest in the image. But due to varying conditions in the outdoor images such as lighting conditions which will produce bright spots or shadows in images and this segmentation will produce undesired results and false detection. Due to which the color of the sign board is difficult to extract. To overcome this approach our region of interest is extracted using MSER algorithm which overcomes the problem of lightning conditions and shadow in images. Which uses intensity function and outer border segmentation. Recognition step is done using support vector machines classifier which will be trained after extracting the feature vector of region of interest. Using this algorithm results we get has high accuracy and less false positives or less undetected. This show that above algorithm is independent of scaling, disorientation and to partial occlusion.