Construction of Asbestos Slate Deep-Learning Training-Data Model Based on Drone Images

Our plain-English summary of Construction of Asbestos Slate Deep-Learning Training-Data Model Based on Drone Images, by S.C. Baek, K.H. Lee, I.H. Kim, et al (Sensors, 2023).

The study

• Authors: S.C. Baek, K.H. Lee, I.H. Kim, D.M. Seo, K. Park
• Journal: Sensors
• Published: 2023;23(19)
• Cited by: 2 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

The detection of asbestos roof slate by drone is necessary to avoid the safety risks and costs associated with visual inspection. Moreover, the use of deep-learning models increases the speed as well as reduces the cost of analyzing the images provided by the drone. In this study, we developed a comprehensive learning model using supervised and unsupervised classification techniques for the accurate classification of roof slate. We ensured the accuracy of our model using a low altitude of 100 m, which led to a ground sampling distance of 3 cm/pixel. Furthermore, we ensured that the model was comprehensive by including images captured under a variety of light and meteorological conditions and from a variety of angles. After applying the two classification methods to develop the learning dataset and employing the as-developed model for classification, 12 images were misclassified out of 475. Visual inspection and an adjustment of the classification system were performed, and the model was updated to precisely classify all 475 images. These results show that supervised and unsupervised classification can be used together to improve the accuracy of a deep-learning model for the detection of asbestos roof slate.

In plain English

This was a technical study in drone imaging and computer image classification, not a study of people or health effects. The authors aimed to build a training dataset that a deep-learning computer model could use to recognize asbestos slate roofs in aerial photographs. They flew an Inspire2 drone with a Zenmuse X5s camera at 100 m over an area with many asbestos slate roofs, giving a ground resolution of 3 cm per pixel, and collected 475 aerial images of asbestos slate. Using ArcGIS software, they first had the pixels in the images sorted automatically into 250 classes by their spectral (colour) features, an unsupervised method. They then regrouped these classes by the colour values of pixels between the slate's valley troughs and used the resulting patterns to train a supervised classifier. To check the results, they visited the area and visually examined the roofs. This found 12 errors among the 475 images. After the classification settings were adjusted, all 475 were classified correctly. The authors concluded that an asbestos slate training-data model was established. They noted that damaged slates hindered detection and that more diverse data are needed.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Methods-development study: construction and field verification of a deep-learning training dataset from drone images, using unsupervised (Iso Cluster) and supervised (maximum-likelihood) image classification
• Who was studied: 475 drone aerial images of asbestos slate from an area with numerous buildings with asbestos slate roofs. The article's text does not name the location or the dates; the location is shown only in a figure. No people were studied.
• Exposure: Asbestos slate roofing material, as seen in aerial images. The article does not name a fiber type, and no human exposure was measured.
• Main results:
  • The drone images were taken at an altitude of 100 m to meet a minimum ground sampling distance of 3 cm/pixel. The gap between slate valley troughs was 6 cm.
  • Unsupervised classification grouped the pixels of the images into 250 Classes by their spectral features. These Classes were then reclassified by the colour (RGB) values of the pixels between the slate valley troughs.
  • Field verification was done by visiting the area and visually examining roof materials.
  • Field verification found 12 errors among the 475 asbestos slate images. The authors attributed these to Class misclassification.
  • After the Class parameters were adjusted and the images reclassified, all 475 asbestos slate images were classified correctly.
• The authors' conclusion: The authors concluded that an asbestos slate deep-learning training-data model was established. They expected it to require minimal time and effort to apply compared with traditional inspection methods.
• Limitations the authors state:
  • Damage to asbestos slates hindered the model's ability to detect asbestos slate comprehensively, so more diverse and advanced learning data are required.
  • Accuracy and reliability could be increased if physical inspection or expert evaluation were incorporated.

Related studies

• Asbestos Detection with Fluorescence Microscopy Images and Deep Learning — C. Cai, T. Nishimura, et al. (2021)
• Deep learning-based classification of mesothelioma improves prediction of patient outcome — P. Courtiol, C. Maussion, et al. (2019)
• Deep-learning based classification distinguishes sarcomatoid malignant mesotheliomas from benign spindle cell mesothelial proliferations — J.R. Naso, A.B. Levine, et al. (2021)
• Deep Learning for the Automatic Quantification of Pleural Plaques in Asbestos-Exposed Subjects — I. Benlala, B.D. De Senneville, et al. (2022)
• The use of a next-generation sequencing-derived machine-learning risk-prediction model (OncoCast-MPM) for malignant pleural mesothelioma: a retrospective study — M.G. Zauderer, A. Martin, et al. (2021)
• Mesothelioma Risk among Construction Workers According to Job Title: Data from the Italian Mesothelioma Register — L. Vimercati, D. Cavone, et al. (2023)

Browse more research in Sensors, or return to medical research.