A Multiple Defect Detection Method Based on Image Block Variance-Weighted Eigenvalues

Abstract

The present invention is a multi-defect detection method based on image block variance-weighting eigenvalues. The method of the present invention determines defects by using an image block to extract a sub-image, and weighting covariance to calculate a weighting eigenvalue. The image block aims at removing sub-images with relatively small variance values, by comparing image variance with block sub-image variance, to obtain sub-images including potential defects as an eigenvalue of the next step. Then a weighting covariance matrix is established by using principal components analysis, and defect positions are determined by calculating weighting eigenvalues. According to the detection method of the present invention, multi-defect of a strip can be simultaneously on-line detected, a plurality of surface defects in an image can be once detected, and the detection and localization of the multi-defect strip target can be accurately achieved. The detection method has features of high detection speed, good real-time, and a high detection accuracy rate.

Description

technical field [0001] Aiming at multiple defects on the surface of strip objects, the present invention proposes a multi-defect detection method based on image block variance-weighted eigenvalue (IPV-WEV). The invention is suitable for on-line detection of multiple defects on the surface of strip objects or similar objects, and belongs to the intersecting fields of machine vision and packaging engineering. . Background technique [0002] The packaging of strip objects, such as: ham sausage, industrial explosives, etc., is the last process of product production, and the quality of packaging directly affects the quality of products. Due to many reasons, there will be defects on the packaging surface of strip objects during the packaging process, and there are many surface crack defects in the image. Once these objects with defective packaging quality are missed and enter the user link, it will bring serious economic losses and negative impacts to users and enterprises. The...

AI Technical Summary

Problems solved by technology

However, there are the following problems: 1) Manual inspection of products is difficult to meet the needs of production efficiency; 2) Inspection work requires a large number of workers, which greatly increases production costs; 3) Manual inspection is labor-intensive, prone to fatigue, inconsistent inspection standards, and easy false detection

[0004] The surface defect of the strip object shows that there is no abnormality in the outline of the object, but there are cracks on the surface of the object. The reason for the surface defect is that the heat seal of the side end is not firm or the friction during the movement

In the actual production process, the number of strip-shaped objects with such defects is relatively small, but surface defects are still an important reason affecting product quality, and online detection of such defects is relatively difficult

The reason is that the surface defects of strip objects are irregular and their positions are randomly distributed, which cannot be predicted in advance, and there are texts on the surface, and the positions of the texts are also uncertain; such defects only account for a small part of the total target detection area, usually no more than 5% ;During the packaging process, the strip-shaped object presents the characteristics of fast movement

Therefore, the traditional method of detecting all target areas is not suitable for the rapid detection requirements of the packaging process of strip objects

Images