由於石棉纖維具有防火性、耐高溫、絕緣、耐磨損、耐酸鹼、耐腐蝕、耐高張力、柔軟度、可撓性等特性，其用途非常廣泛，因此在民國60、70 年代廣泛應用於建築、電器、汽車、家庭用品等。隨著醫學衛生科技進步，陸續有研究報告指出，吸入石棉粉塵會導致肺部纖維化，還可能誘發惡性腫瘤，世界衛生組織(WHO) 所屬國際癌症研究所(IARC) 已將石棉列為一級致癌物；我國環保署於民國78年5月將石棉公告列管為毒性化學物質。但早期慣用的石棉浪板為數眾多且多數皆還在使用，這些年久的石棉浪板若遇到風化或破裂，可能導致石棉纖維飄出，使民眾曝露在石棉危害風險環境，因此找出這些潛在的石棉浪板位置刻不容緩。
本研究主要目的為利用容易取得的影像資料結合AI電腦影像自動辨識之能力，期能簡單快速、低價正確地找到石棉屋頂之分布範圍。分別使用三種監督式機器學習方法－SVM、ML、RF，針對兩種影像資料－純RGB航照、與航照結合衛星影像，進行影像自動辨識與分類，探討在高、中、低三種不同石棉密度分布區，各種分析方法之適用性與分辨能力。
研究結果發現利用機器學習方法確實能簡單、快速、正確地找出石棉之分布，甚至有一些演算法僅用簡單之RGB航照影像就能達到預期之準確率與一致性，如在中石棉密度區結果顯示RF演算法的Accuracy為0.93、Precision為0.8、Kappa為0.88，屬於「幾乎完全一致」。研究顯示RF對石棉影像分類能力最佳，SVM也有很好之結果，但ML則效益較差，這一套結合影像資料與機器學習作石棉屋頂自動辨識之過程值得鼓勵與推廣。

The objective of this study is to combine the remote sensing image data with the automatic image recognition using AI technology to accurately find out the distribution of asbestos roofs. In this study, we use three supervised machine learning methods, such as Support Vector Machine (SVM), Maximum Likelihood (ML), and Random Forest (RF) to make automatically the asbestos roof classification. Two kinds of image data, such as the RGB aerial photography and the satellite imagery are carried out for identification. In addition, three different asbestos density distribution areas of high, medium and low are applied in this proposed procedure to discuss their influence.
Results of this study show that the use of machine learning methods can indeed be simple, fast and correct to find out the distribution of asbestos roofs. Even some algorithms can achieve the expected accuracy, precision, and agreement only using the simple RGB aerial images. In this study, the best result is using the RF algorithm in the middle asbestos density region showing that Accuracy is 0.93, Precision is 0.8, and Kappa is 0.88, which is almost perfect agreement. Results have shown that RF has the best ability to classify asbestos roof, and SVM also has good results, but ML is less effective. Finally, this proposed processes that combine remote sensing image and machine learning method for automatically identifying asbestos roofs are worth encouraging and promoting.

1. Agner, J.C., Newhouse, M.L., Corrin, B. (1988), Correlation between fiber content of the lung and disease in east London asbestos factory worker. Br. J. Ind. Med. ; 45: pp. 305-308.
2. Bassania, C., Cavallia, R.M., Cavalcante, F., Cuomo, V., Palombo, A., Pascucci, S., Pignatti, S. (2007), Deterioration status of asbestos-cement roofing sheets assessed by analyzing hyperspectral data, Remote Sensing of Environment, V. 109, Issue 3, P. 361-378.
3. Bergman, R., Griss, M., Staelin, C., (2002), A personal email assistant. Technical Report HPL-2002-236. HP Laboratories, Palo Alto, CA.
4. Boccaccini, D.N., Leonelli, C., Rivasi, M.R., Romagnoli, M., Veronesi, P., Pellacani, G.C., Boccaccini, A.R., (2007), Recycling of microwave inertised asbestos containing waste in refractory materials. J. Eur. Ceram. Soc. V. 27, P. 1855 - 1858.
5. Bonifazi, G., Capobianco, G., Serranti, S. (2018), Asbestos containing materials detection and classification by the use of hyperspectral imaging, Journal of Hazardous Materials, V. 344, P. 981-993.
6. Breiman, L. (2001), Random Forests, Machine Learning, V. 45, P. 5–32.
7. Chen, P.H., Lin, C.J., Schölkopf, B. (2005), A Tutorial on ν-support vector machines. Applied Stochastic Models in Business and Industry. V. 21(2), P. 111 – 136. DOI:10.1002/asmb.537.
8. Cilia, C., Panigada, C., Rossini, M., Candiani, G., Pepe, M., Colombo, R. (2015), Mapping of asbestos cement roofs and their weathering status using hyperspectral aerial images. ISPRS International Journal of Geo-Information, 4(2), P. 928-941.
9. Colangelo, F., Cioffi, R., Lavorgna, M., Verdolotti, L., De Stefano, L., (2011), Treatment and recycling of asbestos-cement containing waste. J. Hazard. Mater. V. 195, P. 391 - 397.
10. European Space Agency (ESA) (2022), Sentinel User Guides, Sentinel Online, https://sentinels.copernicus.eu/web/sentinel/home.
11. Favero-Longo, S.E., Turci, F., Tomatis, M., Castelli, D., Bonfante, P., Hochella, M.F., Piervittori, R., Fubini, B., (2005), Chrysotile asbestos is progressively converted into a non-fibrous amorphous material by the chelating action of lichen metabolites. J. Environ. Monit. V. 7, P. 764 - 766.
12. Fiumi, L., Congedo, L., Meoni, C. (2014), Developing expeditious methodology for mapping asbestos-cement roof coverings over the territory of Lazio Region, Applied Geomatics, V. 6, P. 37–48.
13. Frassy, F., Candiani, G., Rusmini, M., Maianti, P., Marchesi, A, Nodari, F.R., Via, G.D., Albonico, C., Gianinetto, M. (2014), Mapping Asbestos-Cement Roofing with Hyperspectral Remote Sensing over a Large Mountain Region of the Italian Western Alps, Sensors 14(9):15900-15913.
14. IARC (International Agency for Research on Cancer) (1977). Asbestos. In: IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemical to Humans. Overall Evaluations of Carcinogenicity: An Updating of IARC Monographs, Volumes 1 to 42, Supplement 7. World Health Organization, Lyon, France; P. 106-116.
15. IARC (International Agency for Research on Cancer) (1987). IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemical to Man. Asbestos, Volume 14. World Health Organization, Lyon, France; P. 42-106.
16. Iwaszko, J., Zawada, A., Przerada, I., & Lubas, M. (2018), Structural and microstructural aspects of asbestos-cement waste vitrification. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, V. 195, P. 95-102. doi:https://doi.org/10.1016/j.saa.2018.01.053
17. Kirk, O. (1978), Encyclopedia of Chemical Technology, 3rd., Volumes 1-26. New York, NY: John Wiley and Sons, P. 3 (78) 278.
18. Krówczyńska, M., Wilk, E. (2019), Environmental and Occupational Exposure to Asbestos as a Result of Consumption and Use in Poland. Int. J. Environ. Res. Public Health, V. 16, P. 2611.
19. Krówczyńska, M., Wilk, E., Pabjanek, P, Kycko, M. (2017), Hyperspectral Discrimination of Asbestos‑Cement Roofing, Geomatics and Environmental Eng., V. 11, No. 1, pp. 47-65.
20. Krówczyńska, M., Wilk, E., Pabjanek, P, Zagajewski, B., Meuleman, K. (2016), Mapping asbestos-cement roofing with the use of APEX hyperspectral airborne imagery: Karpacz area, Poland – a case study, Miscellanea Geographica – Regional Studies on Development, Vol. 20, No. 1, P. 41-46.
21. Kusiorowski, R., Zaremba, T., Piotrowski, J., Adamek, J., (2012), Thermal decomposition of different types of asbestos. J. Therm. Anal. Calorim. V. 109, P. 693 - 704.
22. Kusiorowski, R., Zaremba, T., Piotrowski, J., Podwórny, J., (2015), Utilisation of cement-asbestos wastes by thermal treatment and the potential possibility use of obtained product for the clinker bricks manufacture. J. Mater. Sci. V. 50, P. 6757 - 6767.
23. Landis, J.R., Koch, G.G. (1977), The measurement of observer agreement for categorical data, Biometrics, V. 33, No. 1, P. 159–174.
24. Leonelli, C., Veronesi, P., Boccaccini, D.N., Rivasi, M.R., Barbieri, L., Andreola, F., Lancellotti, I., Rabitti, D., Pellacani, G.C., (2006), Microwave thermal inertisation of asbestos containing waste and its recycling in traditional ceramics. J. Hazard. Mater. V. 135, P. 149 - 155.
25. Martin, D.R., Fowlkes, C.C., Malik, J., (2002), Learning to detect nature image boundaries using brightness and texture. In Advances in Natural Information Processing Systems, V. 14.
26. NASA (2022), Landsat Science, https://landsat.gsfc.nasa.gov/
27. Nvidia (2016)，Artificial intelligence and deep learning, https://blogs.nvidia.com.tw/2016/07/whats difference artificial intelligence machine learning deep learning ai/
28. Parkes, W.R. (1994), Asbestos-related disorders Occupational lung disease. 3rd ed., P. 411-505.
29. Pawelczyk, A., Bozˇek, F., Grabas, K., Checmanowski, J., (2017), Chemical elimination of the harmful properties of asbestos from military facilities. Waste Manage, V. 61, P. 377 - 385.
30. Plescia, P., Gizzi, D., Bnedetti, S., Camilucci, L., Fanizza, C., De Simone, P., Paglietti, F., (2003), Mechanochemical treatment to recycling asbestos-containing waste. Waste Manage. V. 23, P. 209 - 218.
31. Raczko, E., Krówczyńska, M., Wilk, E., (2022), Asbestos roofing recognition by use of convolutional neural networks and high-resolution aerial imagery. Testing different scenarios, Building and Environment, V. 217, P. 109092.
32. Rahman, M.H., (2022), Confusion matrix for binary and multi-class classifier. Research Gate: https://www.researchgate.net/publication/360888923.
33. Renard, J.B., Duée, C., Bourrat, X., Haas, H., Surcin, J., Couté, B., (2020) Brightness and polarization scattering functions of different natures of asbestos in the visible and near infrared domain, Journal of Quantitative Spectroscopy & Radiative Transfer, V. 253, P. 107159
34. Robinson, A.H., Morrison, J.L., Muehrcke, P.C., Kimerling, A.J., Guptill, S.C. (1995), Elements of Cartography, 6th Edition. New York: John Wiley & Sons.
35. Rozalen, M., Huertas, F.J., (2013), Comparative effect of chrysotile leaching in nitric, sulfuric and oxalic acids at room temperature. Chem. Geol. V. 352, P. 134 - 142.
36. Stanton, M.F. (1981), Relation of particle dimensions to Carcinogenicity in amphibole asbestoses and other fibrous minerals, J. Natl. Cancer Inst. 67: pp. 965-975.
37. Sugama, T., Sabatini, R., Petrakis, L., (1998), Decomposition of chrysotile asbestos by fluorosulfonic acid. Ind. Eng. Chem. Res. V. 37, P. 79 - 88.
38. Turci, F., Tomatis, M., Mantegna, S., Cravotto, G., Fubini, B., (2007), The combination of oxalic acid with power ultrasound fully degrades chrysotile asbestos fibres. J. Environ. Monit. V. 9, P. 1064 - 1066.
39. Wille, G., Lahondère, D., Schmidt, U., Duron, J., Xavier Bourrat, X. (2019), Coupling SEM-EDS and confocal Raman-in-SEM imaging: A new method for identification and 3D morphology of asbestos-like fibers in a mineral matrix, Journal of Hazardous Materials, V. 374, p. 447-458.
40. Witek, J., Kusiorowski, R. (2017), Neutralization of cement-asbestos waste by melting in an arc-resistance furnace, Waste Management, V. 69, P. 336-345.
41. Zholobenko, V., Rutten, F., Zholobenko, A., Holmes, A. (2021), In situ spectroscopic identification of the six types of asbestos, J. Hazard Mater., 5; 403: 123951. doi: 10.1016/j.jhazmat.2020.123951.
42. Cheng (2021)，最大概似估計(Maximum Likelihood Estimation, MLE)介紹，網址: https://medium.com/qiubingcheng/%E6%9C%80%E5%A4%A7%E6%A6%82%E4%BC%BC%E4%BC%B0%E8%A8%88-maximum-likelihood-estimation-mle-78a281d5f1d
43. Chung (2019)，隨機森林(Random Forest)介紹，網址: https://medium.com/chung-yi/ml%E5%85%A5%E9%96%80-%E5%8D%81%E4%B8%83-%E9%9A%A8%E6%A9%9F%E6%A3%AE%E6%9E%97-random-forest-6afc24871857
44. EasyAI (2020)，人工智慧相關知識，支持向量機－Support Vector Machine | SVM，https://easyai.tech/ai-definition/svm/.
45. 丁亞中、伍肇雄、楊淑蓉 (2002)，高解析度衛星影像分類製作裸露地圖層之研究，地圖，12 期，頁67-80。
46. 中央大學遙測中心 (2019)，美國大地衛星(Landsat)介紹，網址: https://www.csrsr.ncu.edu.tw/rsrs/satellite/Landsat.php
47. 中研院地理資訊中心 (2009)，Google Earth 使用操作講義，37頁。
48. 中研院地理資訊中心 (2015)，Quantum GIS 操作手冊，179頁。
49. 中興測量有限公司 (2022)，數位影像技術：彩色正射影像，http://www.chsurvey.com.tw/page05-01.html.
50. 王秀蘭、包玉海 (1999)，土地利用動態變化研究方法探討，地理科學進展，18 卷，1 期，頁81-87。
51. 台灣化學品製造廠商目錄 (2006)，台灣化工資訊服務社。
52. 李俊賢、蕭汎如、鄭雅文、王榮德 (2016)，石棉的健康危害與台灣現況，台灣職業安全健康連線，網址：http://oshlink.org.tw/about/issue/asbestos/63
53. 汪禧年、郭錦堂 (2011)，利用XRD檢測作業場所粉粹粒子中的石棉分析研究，勞委會勞工安全衛生研究所，IOSH100-A321。
54. 范慶龍 (2021)，監督式機器學習於土地覆蓋分類效益之研究，台灣土地研究，第24卷，1期，頁67-94。
55. 高慶珍 (2004)，遙測影像之符號色彩探討，地圖，14 期，頁145-152。
56. 健康2.0 (2015)，「零保護」？小心石棉致癌危機！，網址：https://health.tvbs.com.tw/review/320978
57. 張建忠、沈哲緯、冀樹勇 (2012)，山坡地社區周緣環境地質災害主題圖製作，台灣建築學報，80 期增刊號，頁1-24。
58. 曹智超 (2008)，建材中石棉含有率測定之探討，碩士論文，中國醫藥大學環境醫學研究所。
59. 維基百科 (2022)，台南市東區介紹，網址: https://zh.m.wikipedia.org/zh-tw/%E6%9D%B1%E5%8D%80_(%E8%87%BA%E5%8D%97%E5%B8%82)
60. 維基百科 (2022)，台南市善化區介紹，網址: https://zh.m.wikipedia.org/zh-tw/%E5%96%84%E5%8C%96%E5%8D%80
61. 維基百科 (2022)，台南市新化區介紹，網址: https://zh.m.wikipedia.org/zh-tw/%E6%96%B0%E5%8C%96%E5%8D%80
62. 維基百科 (2022)，哨兵2號介紹，https://zh.m.wikipedia.org/zh-tw/%E5%93%A8%E5%85%B52%E5%8F%B7。
63. 趙忠明、周天穎、嚴泰來 (2012)，空間資訊技術原理及其應用理論基礎篇，台北：儒林圖書公司。
64. 環保署毒化局 (2019)，石棉之種類、用途、來源及特性，網址：https://topic.epa.gov.tw/asbestos/cp-132-7611-2d1c1-4.html
65. 環保署毒化局 (2019)，毒性化學物質，石綿管制沿革，網址： https://topic.epa.gov.tw/asbestos/cp-135-7614-a4756-4.html
66. 環保署毒化局 (2021)，波形石綿瓦屋頂空間分布推估基線調查計畫，期末成果報告，執行單位：中興測量有限公司。
67. 環保署環境檢驗所 (2008)，我國石棉管制現況與展望，網址：http://www.niea.gov.tw/epaper/epeper_detail.asp?c_id=200
68. 藍福良、石東生 (1982)，石棉暴露危害，工業安全衛生，42期， 頁76-81。