| 摘要 |
Inputs of a plurality of images constituting a group of images of items regarded as non-defective items are accepted and stored, and a defect threshold for detecting a defective portion of an inspection object and a determination threshold for making a non-defective/defective determination are set based on the plurality of stored images. A defective item image which is an image of an item determined as a defective item is previously stored and when an input of an image newly acquired by capturing an inspection object is accepted, non-defective item learning processing is performed by use of a plurality of stored images including the image whose input has been accepted, to at least reset the defect threshold. A defective portion is re-detected based on the reset defect threshold, to determine whether or not the stored defective item image is an image of a defective item based on the set determination threshold. |
