Extraction and Recognition of Traffic Signs in Scene Image.
Youro Minori, Naoki Suganuma, Naofumi FUJIWARA · TRANSACTIONS OF THE JAPAN SOCIETY OF MECHANICAL ENGINEERS Series C · 2001
Information of traffic sign is very important for automatic driving. In this paper, we propose a new recognition technology of the traffic sign in a scene image. At first, a scene image is binarized by the dynamic thresholding. This method is robustness against the variation of brightness. And we extract the central white region of the traffic sign using shape feature of it. Next, we use color feature of the red region, which is periphery of the white region. Color is unsusceptible feature against the variation of brightness. So, we combine shape feature and color feature. Therefore, this new technology can extract the traffic sign in a scene image that have variation of brightness. And this technology can extract the traffic sign that leaned in scene image. Finally, in order to recognize the numerical characters on the traffic sign, we normalize it by means of bilinear interpolation. And we recognize it using 3-layerd neural network.