| 摘要 |
PURPOSE:To contrive to improve a recognition rate by a method wherein a threshold level is modified by a difference between the number of the white points of an input binary image and the number of the white points of a reference image and the binary-coded threshold level is feedback-controlled so that the numbers of the white points become equal with each other. CONSTITUTION:An original image signal from a camera 3 and a threshold level for binary code use from a threshold level setting part 15 set at a teaching value are compared with each other by a comparator 6 and an input binary image signal 7 is inputted. A reference binary image is led out from a reference binary image memory part 8, the signal 7 is inputted in a correlation device 10, a correlation degree is found and when the correlation degree is larger than a correlation limit value, coordinates in an input binary image are registered in a memory part 13. The coordinates of the maximum correlation degree are extracted in a maximum correlation degree candidate point extraction part 14, the number of white points of an image to correspond to the coordinates is counted, the number is compared with a reference number of white points in a reference image number-of-white points memory part 17 and when the number is +, the threshold level is reset a little highly in a threshold level amount-of-change input part 20. In such a way, the binary-coded threshold level is feedback-controlled so that the number of the white points of the input image becomes equal with the number of the white points of the reference image and the threshold level is averaged. Thereby, a recognition rate can be improved. |
