| Co-Investigator(Kenkyū-buntansha) |
FUKUI Yutaka Tottori University, Faculty of Engineering, Professor, 工学部, 教授 (40032023)
YABUKI Noboru Tsuyama National College of Technology, Department of Electrical and Computer Engineering, Associate Professor, 情報工学科, 助教授 (50200572)
|
|---|
| Research Abstract |
The purpose of this research is development of how to detect a road sign using color information. The results of this research are as follows.
1. In various environments (fine, cloudy, and rain weather), distributions of color were investigated. Data of each environment was packed and the distribution function for every color was created so that it could be adapted for various environments. Furthermore, the distribution function of color with the equation so that the distribution function can be created also using few collection data. Since the color distribution can be made large or it can narrow, it becomes easy to perform a detection experiment. The color extraction stabilized conventionally was attained.
2. In the method using the neural network, the method of detecting the color of each pixel in a image was developed. The object road sign was the highest speed sign, and performed the detection experiment. The rate of detection in the case of fine weather was higher. In the case of t
… More
he rainy weather and the backlight with bad conditions, many incorrect detection were seen. Then, in the case of the backlight, the dark region was detected and was made bright. And it carried out performing color detection etc.. Although this method still was not enough, the result was obtained.
3. Program creation and an experiment of the sign detection using Active Net with the image energy function were performed. In the conventional method, although it was difficult to extract an object when an object sign did not exist near the center of an image, it was enabled to also extract the object which exists in the circumference of an image by changing the form of Active Net. Furthermore, in order to perform stable extraction, a network structure with an equal area was proposed. The program which detects a road sign was developed using this network structure. In the experiment, the capturing rate of object which is outside improved, and the good result was obtained.
4. In the method of discriminating the kind of detected road sign, it succeeded by the preparatory experiment with a certain assumption. However, the experiment in the case of using an actual image was not completed enough. Less
|
