Application of Deep Learning Techniques for Thermal Imagery Analysis in Abnormal Identification of Floor Tiles in Heritage Environments

Thermal imaging has gained significant use in recent years, particularly during the epidemic, including its application in architecture for damage detection on archaeological monuments through high-temperature analysis. The non-invasive nature of thermal imaging, along with its ability to visualize temperature levels, allows for problem identification while preserving the building's structure. The integration of artificial intelligence further enhances its potential applications. This study aims to propose an automated inspection system using a convolutional neural network (CNN) for analyzing materials in high-temperature floor blocks. Data for this study are provided by the Bureau of Cultural Heritage, Ministry of Culture (BOCH), Republic of China (Taiwan). In addition, a professor with more than seven years of experience in monument maintenance is responsible for assisting in the identification of overheated data and collecting thermal images of the floor. Analyzing materials at elevated temperatures for restoration purposes offers several benefits: efficient identification of problematic materials, identification of materials suitable for specific environmental conditions through high-temperature analysis, and determining the sequence of temperature variations among different materials to aid in restoration planning. The identification accuracy rate of this study is as high as 92.1%. Compared with the speed of 3600s for professionals to identify 100, this research only needs 3s. The efficiency is increased by about 99.91, which is an amazing improvement. These functions increase the practicality of restoration efforts, improve restoration quality and efficiency, and contribute to academic research on ancient monument preservation.