Defect detection method, system and device

Abstract

The invention provides a defect detection method, a system and a device. The defect detection method comprises following steps of: inputting a to-be-detected textile image; calculating a template size of the textile image; segmenting the textile image according to the template size and acquiring multiple blocks; calculating a correlation coefficient of each block and all blocks and acquiring a correlation coefficient matrix; and calculating a reachable matrix according to the correlation coefficient matrix and determining a defect part in the textile image according to the reachable matrix. According to the invention, detection of textile defects can be achieved.

Description

technical field [0001] The present invention designs a defect detection technology, especially a method, system and device for textile defect detection. Background technique [0002] Textile defect detection is extremely important in the textile production process. Excellent textiles can bring good profits, while defective products are likely to cause economic losses. Existing textile defect detection technologies often use artificial naked eye observation for detection, or automatic detection of textile defects based on computer vision. With the development of machine vision and related technologies, the automatic detection technology of textile defects will be more and more used in modern industrial production. At present, the automatic detection technology of textile defects is mainly divided into two categories: one is to transform the textile image into the frequency domain, and then analyze and detect the defects. The typical algorithm is mainly based on the wavelet t...

AI Technical Summary

Problems solved by technology

At present, the automatic detection technology of textile defects is mainly divided into two categories: one is to transform the textile image into the frequency domain, and then analyze and detect the defects. The typical algorithm is mainly based on the wavelet transform method, but this method is targeted. Strong, can not be applied to a variety of textile defect detection; the second is to use the template matching method, but the current template matching method cannot be self-adaptive, and it needs to invest in manpower for template matching, and it takes a lot of time, so it is not suitable for industrial Production

[0003] According to the above description, the following problems generally exist in the existing textile defect detection: (1) the detection efficiency is often ignored while ensuring the detection accuracy; (2) the detection method involving the conversion of the space domain and the frequency domain is highly targeted, Changes in the type of textiles will cause large deviations in detection; (3) template-based defect detection methods cannot achieve self-adaptation for the selection of templates, and need to manually select the size of the template, which increases the consumption of manpower in industrial production and improves (4) In order to reduce the subtle influence caused by the stretching of textile images, in most cases, a large amount of training is required for flawless textile images, and training for a few types of textile images The time consumed is acceptable, however, training is not a viable method in industry for a large number of textile types

Images