Research on the Application of Computer Vision and Deep Learning in Green Building Material Quality Inspection
Year: 2025
Type: conference-paper
Abstract: The accuracy of green building materials quality inspection directly impacts construction safety. In response to quality control requirements for green building materials such as recycled concrete, eco-friendly insulation materials, and energy-efficient glass, an intelligent inspection system based on computer vision and deep learning has been designed. The system uses multi-channel image acquisition equipment to capture material surface features with a spatial resolution of 0.1mm/pixel. The deep neural network, based on an improved ResNet-50 architecture, achieves defect identification and classification with a detection accuracy of 97.2% on a dataset of 120,000 images. Through model optimization, the system achieves a processing speed of 25fps at the edge, a 5-fold improvement over traditional methods, with a false detection rate below 0.5%. Engineering practice demonstrates that during six months of application in green building materials production lines, the system inspects 8,000 products daily, effectively identifying 32,000 defective items and preventing 12 quality incidents. This has significantly enhanced quality control levels and provides reliable technical support for construction safety.
Authors: Wenjing Li, Jia Liu, Hejia Li, Xiaoqian Zhou
Institution: Chongqing College of Architecture and Technology
Language: English
Subfield: Industrial and Manufacturing Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4414307494