Aramid paper honeycomb gluing defect detection method

Abstract

According to the aramid paper honeycomb gluing defect detection method provided by the invention, aiming at various defect problems generated in the aramid paper honeycomb gluing production process, adeep network is designed, defect characteristics are automatically extracted, and pixel-level positioning is realized; for the problem that samples are difficult to obtain, a large number of samplesare automatically generated through an image enhancement algorithm, and therefore dependence of machine learning on the number of samples is reduced. The detection scheme provided by the invention solves the problem of simultaneous detection of defects of different sizes and different length-width ratios, is accurate in positioning and high in detection efficiency, and can meet the requirement ofautomatic detection of aramid paper honeycomb gluing defects.

Description

technical field [0001] The invention belongs to the technical field of visual detection, and in particular relates to a method for detecting aramid paper honeycomb glue coating defects. Background technique [0002] During the honeycomb gluing process of aramid paper, there will be six kinds of defects, including black spots, lack of glue on the glue strip, holes, incomplete glue strips, impurities and wire drawing. If these defects cannot be processed in time, it will affect the subsequent honeycomb manufacturing process. At present, the detection of this defect mainly relies on human eye observation. This method is inefficient, labor-intensive, poor in reliability and prone to errors. A vision-based automatic detection method for honeycomb gluing defects of aramid paper. The detection of images by traditional machine learning methods mainly relies on manual feature extraction. However, the defect detection of aramid paper honeycomb has the following characteristics, so th...

AI Technical Summary

Problems solved by technology

[0003] 1. There are many types of defects, and the defect forms are complex and diverse, and there is no unified model

Therefore, the pattern recognition method based on traditional artificial feature extraction cannot effectively adapt to a variety of defect types.

[0004] 2. The defect scale changes greatly and the aspect ratio changes drastically

Therefore, the image analysis and processing methods in general inspection methods cannot adapt to the size of the analysis area, so that they cannot effectively detect defects of different scales and aspect ratios.

[0005] 3. In the paper itself and the actual working environment, there is a lot of noise interference, and the detection based on the visual method is easily affected by problems such as dust / light / imaging

Therefore, the existing defect detection methods based on features or templates cannot effectively avoid the influence of noise, which can easily lead to misidentification and waste of raw materials

[0006] 4. In the actual industrial working environment, it is not easy to obtain samples. It is difficult to obtain a large number of samples of defects of different types and shapes, and sample marking will waste a lot of time

Therefore, detection methods based on sample libraries and big data cannot play a good role in factories

Images