Multi-scale presswork defect detecting method based on random forest

Abstract

A multi-scale presswork defect detecting method based on a random forest comprises the steps that a certain number of aligned and qualified presswork images are input to form training samples; after setting of a detection area, generating of defect detection points and multi-scale expressing of the images of the training samples are conducted, the cross channel binary feature of each defect detection point under each scale is collected, then the random forest is trained, and a random forest defect detecting model for all defect detecting points under multiple scales is generated; after the model is available, a test image is given, firstly, multi-scale expression is generated, whether defects exist at the defect detecting points or not is judged through the trained random forest defect detecting model under all scales, multi-scale screening and composition are conducted on the defect detecting points, and finally a detecting result is formed. By means of the method, the detecting speed is increased, and the detecting precision is improved; under the premise that the calculation complexity is not obviously increased, the detecting effect is superior to the detecting effect of other presswork defect detecting algorithms at present.

Description

technical field [0001] The invention belongs to the technical field of image-based industrial product quality detection in computer vision, and in particular relates to a random forest-based multi-scale defect detection method for printed matter. Background technique [0002] Packaging printed matter is the "appearance" of many industrial products, and the quality of its printing is related to the degree of consumer acceptance of the product. In recent years, with the advancement of production technology and the continuous improvement of people's quality of life, new printing technologies have been applied to many printed products. The quality of printed matter is an important part of product quality. Therefore, higher requirements are put forward for the automation level of printed matter quality management in the production process. It is required to use machines to detect printed matter defects instead of manual work, so as to reduce costs and improve efficiency. [0003...

AI Technical Summary

Problems solved by technology

[0014] In order to overcome the problems existing in the above-mentioned prior art, the object of the present invention is to provide a multi-scale printed matter defect detection method based on random forest, which improves the speed and accuracy of detection; under the premise of not significantly increasing the computational complexity, the detection effect is excellent. Compared with other current printing defect detection algorithms

Images