A steel plate surface defect detection method and system based on an improved YOLOv5s algorithm

By improving the YOLOv5s algorithm and combining it with CBAM, RepDWConv and DRC Block modules, the problems of high false detection rate and difficulty in balancing model efficiency and accuracy in steel plate surface defect detection are solved, realizing efficient and lightweight defect detection and meeting the needs of industrial real-time detection.

CN122289641APending Publication Date: 2026-06-26ZHONGBEI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHONGBEI UNIV
Filing Date
2026-02-10
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing methods for detecting surface defects on steel plates based on the YOLOv5s algorithm suffer from high false detection rates in complex backgrounds and high missed detection rates for small targets. Furthermore, it is difficult to balance model accuracy and efficiency, which fails to meet the real-time detection needs of industry.

Method used

The CBAM module is used for feature calibration, the RepDWConv module for detail enhancement, and the DRC Block module for information aggregation. An optimized link of feature calibration-detail enhancement-information aggregation is constructed to improve the YOLOv5s algorithm.

Benefits of technology

Significantly reduces false detection rate, improves small target detection capability, lightweight model meets industrial real-time detection needs, improves detection accuracy by 5.2%, reduces parameter count by 21.3%, and achieves detection speed of 93.2 frames/second, achieving an excellent balance between accuracy and speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122289641A_ABST
    Figure CN122289641A_ABST
Patent Text Reader

Abstract

This invention provides a method and system for detecting surface defects in steel plates based on an improved YOLOv5s algorithm. The detection method includes: S1, acquiring and preprocessing an image of the steel plate surface; S2, inputting the preprocessed image into a pre-trained improved defect detection model. The improved defect detection model is based on the YOLOv5s network architecture and includes the following improvement steps: S2.1, embedding a CBAM module in the backbone network; S2.2, replacing the standard 3×3 convolutional layers in the backbone and neck networks with RepDWConv modules; S2.3, replacing the feature fusion module in the neck network with a DRC Block module; S3, acquiring the detection information output by the improved defect detection model. This application makes a triple collaborative improvement to the YOLOv5s baseline network, constructing an optimized link of "feature calibration—detail enhancement—information aggregation," effectively solving the technical problems of strong background interference, missed detection of small targets, and difficulty in balancing model efficiency and accuracy in steel plate defect detection.
Need to check novelty before this filing date? Find Prior Art