Method, program, apparatus, and system for abnormality detection such as for determining whether a plurality of containers to be stacked on a pallet is normal or abnormal

An abnormality detection method, comprising:

• acquiring a determination target image that is an image of a determination target object to be regularly aligned while standing, which is a cylindrical container;
• creating a difference image emphasizing a difference between an image of a normal determination target object regularly aligned and the determination target image;
• calculating a multi-segmentation error of the difference image to determine a maximum value of the multi-segmentation error;
• and
• determining whether the determination target object is normal or abnormal in accordance with the maximum value, wherein
• the creating the difference image includes
• acquiring, by inputting the determination target image into an autoencoder configured such that an image of a normal determination target object is output in response to input of an image of a normal determination target object through learning with an image of a normal determination target object regularly aligned, a reference image as output of the autoencoder;
• and
• creating the difference image by calculating a difference between the determination target image and the reference image.