2023 Volume 79 Issue 5 Article ID: 22-00129
Asbestos had been widely used for building materials in past, and has been banned. In case the buildings are demolished, the ambient concentration of asbestos fibers is monitored by local governments because the scattering of asbestos fibers from building materials is concerned. In order to rapidly make the countermeasure to scattering of asbestos fibers, it is important to rapidly measure the ambient concentration of asbestos fibers. In this study, a new asbestos detection and counting method using a convolutional neural network (CNN) model is proposed as a more rapid measurement method. The method can automatically count the number of asbestos fibers in the image of phase contrast microscopy. The model was fine-tuned by teacher data created by samples obtained at the site of building demolishing, and samples made in laboratory. The model showed good performances in both of detection (F-score is 0.83) and counting (relative error is 11%).
