Fabric image defect real-time detection method

Abstract

The invention relates to a fabric defect automatic detection technology. The invention particularly relates to a fabric image defect real-time detection method. According to the method, an industrialcamera is adopted to acquire the fabric image and judge whether the fabric image has defects or not in real time, and the fabric image acquired by the industrial camera has a large amount of redundantinformation, so that useful information can be extracted from the redundant information, and whether one fabric image has defects or not can be accurately judged. The average calculation time is about 0.15 s, and the industrial real-time requirement is met. The non-contact judgment means does not interfere with fabric production, the ex-factory qualification rate of products is greatly increased,and a factory is assisted to achieve unmanned operation.

Description

technical field [0001] The invention relates to a fabric defect automatic detection technology, in particular to a fabric image defect real-time detection method. Background technique [0002] In the textile production industry, due to the error of the machinery and equipment for weaving textiles, the deviation of raw materials and the unsatisfactory environment, many defects will occur, including many defects such as yarn dragging, knots, broken filaments and poor correction. Because of the irregular shape and inconspicuous features of these defects, it is difficult to use traditional detection circuits for automatic identification. [0003] There are mainly two ways to detect and identify irregular defects. One is to rely on manual detection of irregular defects. However, due to individual differences, fatigue and other reasons, there are often missed detections. In addition, because different inspectors have different experience in fabric inspection and have different s...

AI Technical Summary

Problems solved by technology

This method not only requires a large number of defect data samples for training, but also requires that the defect features to be detected should not be too different from the sample features, otherwise the trained features will not be able to adapt to new defects, resulting in missed and false detections.

However, many defects of fabrics are not all caused by predictable reasons, resulting in various shapes and different characteristics. Moreover, if an additional layer of printed patterns is added to the fabric, the existing image algorithms are basically difficult to apply industrial practice

Images