Prefabricated part surface defect detection method based on machine vision

By combining fractional-order operators with structural tensor decomposition, the problem of low accuracy in weak-contrast fine edge detection under complex working conditions is solved, achieving high-precision multi-type defect identification and low-cost system deployment.

CN122368056AActive Publication Date: 2026-07-10SHAANXI CONSTR ENG (ANKANG) NEW BUILDING MATERIALS CO LTD
9 Cites 0 Cited by

Patent Information

Application Number
CN202610813686.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-08
Publication Date
2026-07-10
Estimated Expiration
2046-06-08

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve high-precision multi-type defect identification under complex working conditions, especially in the detection of weak contrast fine edges, which has low accuracy. Furthermore, the systems are costly to deploy and lack sufficient anti-interference capabilities.

Method used

A method combining fractional-order operators and structural tensor decomposition is adopted. The high-frequency texture of the background and the edge transition components are separated by fractional-order diffusion operators and structural tensor decomposition matrices. Multi-scale convolutional recombination is performed using cross-scale cascaded residual modules to generate high-dimensional semantic feature vectors. The defect results are output by Mahalanobis projection distance comparison.

Benefits of technology

It improves the accuracy of weak contrast fine edge detection, enhances the system's anti-interference capability, reduces system deployment costs, and achieves high-precision geometric quantization evaluation.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention belongs to the technical field of quality detection of concrete prefabricated parts, and relates to a machine vision-based prefabricated part surface defect detection method, which comprises the following steps of: separating high-frequency texture and edge jump components by utilizing a fractional order diffusion operator and a structure tensor decomposition matrix, and recombining and outputting a recombined image matrix through a cascade residual module; extracting a local binary pattern of the recombined image matrix and a gray level co-occurrence matrix eigenvalue, and splicing to generate a high-dimensional semantic eigenvector; and comparing the high-dimensional feature vector with the Markov projection distance boundary amplitude of the standard center cluster to output a marking result, establishing a spatial adaptive fractional order diffusion mechanism, blocking gray fusion crossing a weak defect edge by using a fractional order weak memory attribute, solving the reverse exclusion restriction between noise suppression and edge fidelity, and improving the accuracy of the marking result. And the stripping precision of the weak-contrast micro-edge is improved.
Need to check novelty before this filing date? Find Prior Art