A Traffic Sign Recognition Method Based On Template Matching

With the increasing number of cars and the intricate of city street, highway, IntelligentTransportation System (ITS) emerged. ITS is a combined system which can take advantage ofthe current information transmission technology, sensing technology, the state detectiontechnology and integrated control technology. In recent years, with image processing andpattern recognition techniques are introduced into ITS continuously, this makes theintelligence level of ITS improved significantly.The detection and recognition of traffic signboard is an important research direction inITS. By a video capture device and a traffic signboard detection and recognition algorithminstalled in the driving car, it can be timely to give the driver the information of current roadcondition, guideline drivers traveling specification effectively, avoid driver distraction andreduces the probability of traffic accidents. Therefore, the detection and recognition of trafficsignboard using visual image information has aroused widespread concern. At the same time,this also has important theoretical study significance and practical valueThis article focuses on solving the detection and identification of traffic signboard basedon the Template matching. The main work includes the following points:First, in accordance with the characteristics of the different color spaces and trafficsignboard in external environment, we choose two color spaces: HSV color space andCIE-LAB color space. They are not sensitive to light changes. In these color spaces, weachieve a coarse segmentation of traffic signboard.Second, image preprocessing techniques (denoising, small region removing, region repairand connected regions extraction) will be used in the coarse segmentation result. Then ashape-degree measure will be used to detect traffic signboard.Third, through the analysis of some commonly used statistical image features and thecorresponding similarity measure methods combined with the actual environment, the HUinvariant moments feature descriptors will be used to extract image features and the Hausdorffdistance and correlation coefficient will be used to recognition.Finally, through method research results on the detection and recognition of traffic signboards, this article designs and implements a traffic signboards detection and recognition systembased on the video sequence. The system uses the HSV color space and shape-degree measureto detect traffic signboards, and uses the HU invariant moments feature descriptors based onHausdorff distance to recognition. Our system achieves good detection and recognitionresults.