Summary: | 碩士 === 元智大學 === 資訊工程學系 === 91 === Two-step strategy is adopted in this study to recognize road sign. Features of road sign are extracted provided that the sign in the image has been detected and its size has been known. These feature are in turn used to match the detected road sign to one of the signs in database so that the goal of road sign recognition can be achieved.
More specifically, in the stage of feature extraction, the RGB components of the road sign are first converted into HSV color space and then guantized into six dominate colors together with gray of white and black. The distribution of the features is used to construct a color histograms. The template matching and Zernike moments for respective color planes are also computed for later recognition.
In the stage of recognition, two-step strategy is adopted. In the first stage, histogram intersection is used to prune irrelevant signs. In the second stage, template matching or distance disimilarity based on Zernike moments are adopted to find the corresponding sign from the small set of candidate signs. Experimental results show that the proposed method is feasible.
|