Supercharge Your Innovation With Domain-Expert AI Agents!

Semi-supervised industrial defect detection method and system based on feature comparison

A defect detection and feature comparison technology, applied in the field of deep learning and computer vision, can solve problems such as confusion, unavoidable semantics, and limited precision, and achieve the effects of improving precision, avoiding confusion, and strong detection capabilities

Pending Publication Date: 2022-05-13
SHANGHAI JIAO TONG UNIV +1
View PDF2 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But the invention is not based on feature comparison, semantic confusion cannot be avoided, and the accuracy of defect detection is limited

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Semi-supervised industrial defect detection method and system based on feature comparison
  • Semi-supervised industrial defect detection method and system based on feature comparison
  • Semi-supervised industrial defect detection method and system based on feature comparison

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0080] According to the present invention provides a semi-supervised industrial defect detection method based on feature comparison, such as Figure 1- Figure 3 shown, including:

[0081] Step S1: Collect the pictures of the products to be tested and label the pictures randomly;

[0082] Step S2: Classify the products to be tested, divided into labeled input and unlabeled input;

[0083] Step S3: For labeled input, train the student network with pictures and corresponding labels;

[0084] For unlabeled input, the input into the teacher network produces the corresponding pseudo-labels and representations;

[0085] Step S4: Filter for pseudo-labels to distinguish between reliable pixels and unreliable pixels;

[0086] Step S5: Supervise the feeding of reliable pixels into the student network, and for the unreliable pixels, the characteristic optimization of the student network based on contrast learning is based on their feature encoding information.

[0087] Specifically, in the ste...

Embodiment 2

[0112] Example 2 is a preferred example of Example 1, to illustrate the present invention more specifically.

[0113] Those skilled in the art may provide a semi-supervised industrial defect detection method based on feature comparison provided by the present invention, to be understood as a specific embodiment of a semi-supervised industrial defect detection system based on feature comparison, i.e., the semi-supervised industrial defect detection system based on feature comparison may be implemented by performing the step flow of the semi-supervised industrial defect detection method based on feature comparison.

[0114] According to the present invention provides a semi-supervised industrial defect detection system based on feature comparison, comprising:

[0115] Module M1: Collect the pictures of the products to be tested, and randomly label the pictures;

[0116] Module M2: Classify the products to be tested, divided into labeled input and unlabeled input;

[0117] Module M3:...

Embodiment 3

[0146] Example 3 is a preferred example of Example 1, to illustrate the present invention more specifically.

[0147] 1. A semi-supervised industrial defect detection system and method based on feature comparison, including:

[0148] Step A: Collect the pictures of the products to be tested, and mark the pixels of some pictures to form an input sample with some supervision information.

[0149] Step B: According to whether the product under test has a picture, the input sample is divided into "labeled input" and "unlabeled input".

[0150] Step C: For "Label Input", directly use these pictures and corresponding labels to train the student network.

[0151] Step D: For "Unlabeled Input", enter it into the teacher network, generating the corresponding pseudo-labels and representations.

[0152] Step E: Using the entropy value, the pseudo-labels are filtered at the pixel level to distinguish between high-confidence and low-confidence pseudo-labels.

[0153] Step F: For high-confidence ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention provides a semi-supervised industrial defect detection method and system based on feature comparison, and the method comprises the steps: S1, collecting a picture of a to-be-detected product, and randomly labeling the picture with a label; s2, classifying the to-be-tested products, wherein the to-be-tested products are classified into label input and label-free input; s3, for the input with the labels, training the student network by using the pictures and the corresponding labels; for no-label input, inputting into a teacher network to generate a corresponding pseudo label and representation; s4, screening the pseudo labels, and separating out reliable pixels and unreliable pixels; s5, the reliable pixels are sent to the student network for supervision; for unreliable pixels, feature optimization based on comparative learning is carried out on the student network according to feature coding information of the unreliable pixels. According to the method, the effective neural network model is trained for defect detection under the condition of lack of labeling of the industrial defect detection data, so that the workload of manual labeling is reduced, and the cost is greatly reduced.

Description

Technical field [0001] The present invention relates to the field of computer vision, deep learning technology, specifically, to a semi-supervised industrial defect detection method and system based on feature contrast. Background [0002] In industrial production, almost all products need to be quality inspected, and most of the quality inspection process is completed by quality inspectors with naked eye vision to detect product defects (hereinafter referred to as visual inspection). Due to the diversity of products and defects, the workload and difficulty of the quality inspectors are greatly increased, resulting in a decrease in the efficiency of manual visual inspection and the fatigue and mistakes of the quality inspectors, which leads to missed inspections and misinspectives, which improves the time cost of the production line and may affect the quality of the listed products. Therefore, the use of automated detection technology has a very important value. [0003] Early a...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06V10/764G06V10/82G06V10/48G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/047G06N3/045G06F18/2414G06F18/2415
Inventor 乐心怡王钰超陈彩莲关新平
Owner SHANGHAI JIAO TONG UNIV
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com