Building Material Defect Detection and Diagnosis Method Based on Big Data and Deep Learning

Abstract

Building materials in the use of the process will inevitably appear a variety of defects, such as cracks, corrosion, peeling, deformation, etc. How to overcome the shortcomings of traditional detection methods and improve the efficiency and accuracy of defect detection and diagnosis of construction materials is the hot and difficult point of current research. This paper uses big data and deep learning to construct a new method for defect detection of construction materials. This paper constructs a defective image database of construction materials containing 100,000 images, 8 types of construction materials, and 12 types of defects. The article combines deep learning models such as deep neural networks, CNNS, and attention mechanisms to construct a framework for defect detection and diagnosis. This paper verifies the conclusions of the article through empirical analysis, and the model of the article outperforms the existing models in several evaluation indexes. It realizes automatic, fast and accurate detection and recognition of defective images of construction materials, as well as extracting and analyzing information such as the type, location, size and degree of defects, which provides an effective basis for subsequent defect assessment and repair.