A defect detection method, system, device and medium for cross-type site merging
By cropping and enhancing historical images of different types of sites, and combining multi-scale learning and site information for secondary judgment of detection results, the problem of low accuracy and poor applicability of deep learning models across different types of sites is solved, and higher defect detection accuracy is achieved.
CN118071730BActive Publication Date: 2026-07-21CHENGDU UNION BIG DATA TECH CO LTD
View PDF 2 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU UNION BIG DATA TECH CO LTD
- Filing Date
- 2024-03-29
- Publication Date
- 2026-07-21
AI Technical Summary
Technical Problem
Deep learning models have low accuracy in identifying defects across different types of sites, poor applicability, and cannot be effectively used across different site types.
Method used
By cropping and enhancing historical images of different types of sites, and combining multi-scale learning and site information to perform secondary judgment of detection results, erroneous results are filtered out, thus improving the accuracy of image judgment.
Benefits of technology
It enhances the ability to detect minor defects and improves the accuracy and applicability of defect detection across different types of sites.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN118071730B_ABST
Abstract
The application provides a defect detection method, system, device and medium for cross-type site merging, and relates to the technical field of defect detection. The method flow is as follows: a historical image is acquired, and feature labeling processing is performed on the historical image to obtain a defect target frame; image cropping processing and image enhancement processing are performed on the defect target frame of the historical image to obtain a defect enhanced image; the historical image and the defect enhanced image are respectively input into a deep learning model for iterative training; a defect type detection is performed on a to-be-detected image, and a filtering correction processing is performed on the defect type detection result based on site type information to obtain a defect detection output result. The historical image of different types of sites is subjected to image cropping and image enhancement, the small defect detection capability is strengthened through multi-scale learning, and the secondary determination of the detection result is performed by combining the site information of different types of sites, the wrong determination result is filtered, and the accuracy of the image determination is improved.
Need to check novelty before this filing date? Find Prior Art