抄録
An increase in the number of parking lots is very slow though that of cars increases much every year. Thus, efficient management of parking lots is needed. If information of vacant divisions is transmitted to cars waiting at gates of parking lots, traffic jam caused by cars searching parking divisions will be decreased. Therefore several methods for detecting parking cars have been developed. Especially, the method that detects cars hiding white lines drawn on parking lots is highly reliable. The method needs a template image for each camera. Since these images were created by human in the previous research, much cost and time was needed.
In this paper, we present the method automatically generating the template images. Firstly, our method synthesizes an image of parking lots including no cars from several images. Then, the method detects white line segments from the image. The line segments are corrected in consideration of the rule that white lines on the parking lots are parallel and their length is constant. Finally, parking divisions are determined from the line segments and stored in the template image. In the experiment using the template generated by our method, accuracy of detecting cars was about 96%. The template is comparable to the manually created template in accuracy.
