Asbestos Detection in Building Materials Through Consolidation of Similarities in Color and Shape Features

As the general public has become increasingly aware of problems caused by asbestos, it is a pressing issue to check for the presence of asbestos fibers in buildings and elsewhere. In this paper, we propose a method using computers to detect and count asbestos particles in microscope images. Noting that asbestos particles have distinctive colors and shapes, we have extracted color and shape features to detect them. As detectors, we have used Support Vector Machine (SVM), which has better performance in utility than other classifiers. We do not consider it possible to properly utilize similarities in color and shape features if we input consolidated features of colors and shapes into a kernel function. We have therefore defined weighted summation of kernels applied to colors and shapes respectively. As a result, we have confirmed that the use of weighted summation of kernels can improve the precision of detection.