Detection of Stained Asbestos in Building Material Microscopic Images Using Empirical Wavelet Transform

Asbestos is a globally recognized occupational hazard. While techniques have been developed to stain asbestos for detection on building material surfaces using stereomicroscopy, there are currently no automated methods for identifying stained asbestos in such images. In this study, we propose a novel approach for detecting stained asbestos in microscopic images using the Empirical Wavelet Transform (EWT), which adaptively decomposes signals into their constituent frequencies. Our method utilizes EWT to enhance image contrast and extract features, including gray-level histograms, entropy, GLCM statistics, and GLRLM statistics, from stained images converted to the YIQ color space. These features are then employed to classify asbestos and non-asbestos regions using a Kernel Support Vector Machine (SVM). Experimental results demonstrate that our method achieves an accuracy of over 99% in detecting stained asbestos from segmented microscopic images.