Label identification method based on improved structure similarity

Abstract

The invention discloses a curved surface similar label identification method based on improved structure similarity. The method comprises the steps of image acquisition, template image processing, image segmentation and stretching, image registration, similarity identification and the like; the image acquisition comprises acquisition of a to-be-detected image and a template image; the template image processing realizes stretching of the template image and tailoring of a label region; the image segmentation and stretching realize segmentation of the label region from the to-be-detected image, so that a to-be-detected label image is obtained; the image registration step is used for registering a template label image and the to-be-detected label image; and the similarity identification step is used for calculating the similarity between the template label image and the to-be-detected label image through the improved structure similarity GDSIM, so that whether labels are the same labels ornot is identified. The label identification method disclosed by the invention is simple in step and is suitable for not only plane labels but also curved surface labels; and for the problem that thelabel is difficult to identify under the condition of non-uniform illumination and slight deformation in the label image, a good identification effect can be achieved.

Description

technical field [0001] The invention relates to the technical field of machine vision and image processing, and is a label identification method with improved structure similarity. Background technique [0002] Labels are used to record key information of products, especially in the field of food packaging. In the current food production line, affected by many uncertain factors such as environment, equipment, and human factors, the labels on food may have various defects, such as Labels are missing, crooked, wrong, etc. These defects are serious product quality problems. Mislabeling of similar labels is a defect that is difficult to identify. Similar labels have similar patterns or most of the text is the same. Therefore, the identification of similar labels is a difficult problem for traditional defect detection methods. [0003] Traditional label identification is mainly based on manual identification. Manual identification is greatly affected by subjective factors, and w...

AI Technical Summary

Problems solved by technology

[0005] The problem to be solved by the present invention is: the main technical problem in the current label identification application is that there is a certain degree of uneven illumination and label deformation in the label image, which not only leads to the loss of precision in image registration, but also the existing label The identification method is not robust enough to uneven illumination and deformation, and the detection accuracy cannot meet the needs of use

Images