抄録
The paper describes a novel approach of feature extraction for the fast object recognition based on Model Based Matching method (MBM method) in computer vision researches. The proposed approach is aimed to achieve the recognition in video-rated processing. In general the MBM method is known and utilized as the robust method against the variation of the luminance of target objects and the brightness of their circumstance. On the conventional process in the MBM method, however, it spends more than hundreds [msec] from taking the image of the objects to detecting their positions and orientations because of the large amount of calculation. In our approach, by introducing the Model made of central positions of circles only, quite novel process are structured, which is composed of just two steps to extract the positions. The first is to obtain the differential image by applying the filtering mask feasibly designed in our method. The second is to extract the central positions of circles from the differential image. The validity of our method is confirmed in the experimental results.
