A Deep Learning-Based Method and System for Detecting Surface Defects in Building Materials

By employing deep learning methods, image acquisition, and neural network technology, surface defects of building materials at construction sites can be accurately identified. This solves the problem of low efficiency in manual identification, improves identification efficiency and accuracy, and generates targeted usage and maintenance measures, thereby reducing material costs.

CN122289174APending Publication Date: 2026-06-26NANTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN ยท China
Patent Type
Applications(China)
Current Assignee / Owner
NANTONG UNIV
Filing Date
2026-03-25
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Identifying surface defects of building materials on construction sites requires a large amount of manpower and is prone to error, making it difficult to efficiently and accurately identify and handle surface defects.

Method used

A deep learning-based method for detecting surface defects in building materials is adopted. This method involves image acquisition, BM3D algorithm for noise reduction, grayscale and gradient threshold segmentation, CNN convolutional neural network model pruning, Faster-RCNN region convolutional neural network training, and Bayesian classification algorithm to identify and classify surface defects in building materials.

Benefits of technology

It enables precise identification of surface defects in building materials, improves identification efficiency and accuracy, reduces the need for manual labor, generates targeted usage and maintenance measures, increases the utilization rate of building materials, and reduces material costs.

โœฆ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122289174A_ABST
    Figure CN122289174A_ABST
Patent Text Reader

Abstract

This invention relates to the field of building materials, specifically a method and system for detecting surface defects in building materials based on deep learning. The deep learning neural network model enables accurate identification of surface defects in building materials, improving work efficiency at construction sites. Image data of the building material surface is acquired through an image acquisition device, and the BM3D algorithm is used to denoise the image data, resulting in denoised surface image data. An initial Faster-RCNN region convolutional neural network model is established using a CNN convolutional neural network. The segmented building material surface image data is input into the target Faster-RCNN region convolutional neural network model for training, extracting the average gray values โ€‹โ€‹of image edges and defect areas. A Bayesian classification algorithm is used to classify the surface defect data, resulting in a building material surface defect database. Repair measures are generated for building materials with moderate to moderate surface defects, and application measures are generated for building materials with minor surface defects.
Need to check novelty before this filing date? Find Prior Art