To track on-site variations in asbestos roofing and monitor its subsequent material flow, this project utilized SPOT satellite imagery from 2022–2023 to compare with the results of 2021, identifying changes in asbestos roofing across Taiwan during the two periods. Additionally, the project integrated Google Street View data and UAV imagery provided by the National Land Surveying and Mapping Center for verification. Geographic locations of asbestos roofing building footprints were corrected, and building attribute data was updated accordingly.According to the 2020–2021 Baseline Survey and Spatial Distribution Methodology for Outdoor Building Sides Containing Asbestos Materials, Taiwan had an estimated total of 240,000 asbestos roofs, with a combined weight of 540,000 metric tons. Based on the satellite imagery variation monitoring conducted from 2022–2023, approximately 28,000 roofs, with a total weight of 57,000 metric tons, were found to have undergone changes. As of now, the total number of asbestos roofs in Taiwan is approximately 229,000, with a combined weight of 493,000 metric tons. These findings indicate a decreasing trend in both the number and weight of asbestos roofs nationwide, with the most significant changes observed in central and southern counties. The variation data has been integrated into the alert page of the Spatial Distribution Management System for Outdoor Asbestos Materials, providing a reference for local Environmental Protection Bureaus to conduct on site inspections and manage identified variation locations. To expand the database of images related to asbestos-containing building sides, this project developed a prototype of an "Asbestos Containing Building Sides Automatic Recognition App," which includes features for capturing images of asbestos-containing building sides, AI-based interpretation, and geolocation. The AI interpretation module incorporates the YOLOv9 model, which was trained, tested, and validated using images of asbestos-containing building sides. The final model achieved an accuracy rate of 70%. Additionally, to explore the potential applications of hyperspectral imaging technology in the monitoring of toxins and chemical substances, this project reviewed international literature on hyperspectral applications, which can be categorized into two main areas: (1) the use of hyperspectral imaging for identifying environmental pollution or conducting long-term monitoring of environmental hotspots and (2) chemical substance identification and the development of optical spectral databases. For chemical substance identification, preliminary tests were conducted using indoor hyperspectral instruments to capture images of PFOA and PFOS. The test results for characteristic wavenumbers revealed significant features at 1016 cm-1 and 1020 cm-1. Based on these findings, a spectral database of chemical substances was established, and spectral testing was conducted, laying the foundation for hyperspectral applications in the analysis of toxic substances.

Asbestos Roof Variation, AI Recognition, Hyperspectral