Color Recognition Method based on Circle Extraction and Similarity

To drive and control autonomous guided vehicle by the autonomous system, the landmark can be attached and reported the driving information. This landmark can notify the information by its color. Vision-AGV captured the landmark by the image and then they analyze the meaning of the landmark. To implement this concept, the system should be designed for detecting and recognizing the landmark. Therefore, this paper proposes the method to detect and recognize the color of the landmark. To detect the landmark, circular Hough transform is applied. This algorithm can detect the shape of the circle. However, it detects the circle which is not the landmark. To sort out the landmark, peak signal-to-noise ratio is calculated. This value represents the similarity between two images. So, we compare the circle and the reference for determining whether it is a landmark or not and recognizing the color of the landmark. The circle is extracted by the circular Hough transform, and the reference represents each color such as green, red, yellow. To compare two things by peak signal-to-noise ratio, we determine the color of landmark. As shown in experimental results, the proposed method can detect and recognize the color of landmark in case of the traffic light.