End-to-end neural network based fabric defect detection method

Abstract

The invention provides an end-to-end neural network based fabric defect detection method. The end-to-end neural network based fabric defect detection method comprises the following steps that initialparameters of a SSD neural network model are set, a fabric defect image in a fabric defect database is input the set SSD neural network model to train, and a fabric detection model of deep learning isobtained; the fabric image to be detected is input into the fabric detection model trained in the step one, feature extraction is carried out on the fabric image, and a plurality of candidate frameswhich may be defect targets are selected and taken out; and based on a set judgement threshold value, the candidate frames in the step two are identified to obtain final defect targets, defect targetframes are selected by using intersection and comparison thresholds of the candidate frames where the defect targets are located, and location coordinate information of defects is stored and the defect target frames are output. The end-to-end neural network based fabric defect detection method has good adaptability and detection performance to both plain fabric and mode fabric, and expands the application range; the detection rate is fast, and the problem of slow manual detection speed is effectively solved; and the model is easy to train, and the operation is simple.

Description

technical field [0001] The invention relates to the technical field of fabric defect detection in textile image processing, in particular to a fabric defect detection method based on an end-to-end neural network, which detects and locates defects in a fabric defect image. Background technique [0002] my country is a big country of textiles, and the textile industry occupies an important position in the social economy. Among them, the quality of fabrics is the key issue, and the detection of fabric defects is an important link in the quality control of textiles. At present, the vast majority of industrial production lines still use manual defect detection. The results of traditional manual inspection are greatly affected by human subjectivity, and the speed is slow and the efficiency is low, making it difficult to guarantee the accuracy and real-time detection. With the continuous progress and development of machine vision, image processing technology and deep learning algor...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the technical problems that the existing fabric defect detection method has poor fabric defect detection effect and large amount of calculation, the present invention proposes a fabric defect detection method based on an end-to-end neural network, and improves the existing SSD deep neural network. Fabric defect detection, which can automatically identify defects and determine the defect location information for the identified defect marks, which is suitable for multi-dimensional and complex textured fabric images

Images