| 摘要 |
PURPOSE:To accurately and stably extract a featured shape and to eliminate variance in feature quantity due to positioning by using the individual threshold values of windows and performing segmentation only in the window. CONSTITUTION:The individual threshold values of many windows on one screen are determined by sampling brightness and the segmentation is performed only in the window according to the threshold values. Namely, the best threshold value is determined by looking at the inside of a window 19 whose threshold value is to be determined so as to determine the individual threshold values of the respective windows 19. Therefore, when threshold values of plural images are determined, the threshold values need not be determined in consideration of whether or not other featured shape parts are normal, and an area check and a position check at the time of teaching are made by performing the segmentation only in the windows by using the individual threshold values to stably and accurately extract the featured shape.
|
