摘要: |
In recent years, the rapid increase in a traffic accident is a social problem with change of traffic environment. In order to prevent these traffic accidents, several pieces of research have been proposed. One of them has the environmental recognition ahead of vehicles. These researches set a sensor in front of the vehicles, and feed back the information to a driver in a certain form. This paper proposes a technique for extraction and recognition of a road sign from color image sequence. We use the images that are taken by the video camera set on the passenger seat. The color and shape of the road signs are the key of extraction processing. The colors used for the road signs are three colors of red, blue, and yellow. The specific color binary image of these 3 color is created first. However, the captured image will be greatly influenced according to the relative position of a camera and the sun, the weather, etc., and brightness will be distorted. In order to solve these problems, the ingredient of a color is converted into HSL. A specific color binary image is created based on H ingredient that cannot be easily influenced by environment. It becomes possible by using this color system to perform color extraction that cannot be easily influenced by the external environmental factor. Next process is shape extraction. Shape is extracted by taking matching prepared template to these specification color binary image. Tracking processing is performed until it reaches the size, which can be recognized in the extracted domain after the following frame. Initial extraction is performed from when the sign is reflected small on an image, a motion of the object, size, and brightness are learned and presumed, and efficient tracking processing is performed. |