Abstract
An image processing algorithm to automatically count asbestos fibers was developed. To accurately detect target objects in an image, a method detecting edges of the objects, instead of existing thresholding methods, was used. Recognition of the edges was achieved by difference of luminance between the objects and the surrounding background in an image. A commercially available image processing library was used to implement this algorithm. Applying this algorithm to asbestos samples (n=18) whose concentration ranges were from several to several thousands fibers per liter, it was confirmed the counting results were consistent with those by the existing visual counting method. Since this implementation enables to automatically process numerous images and count many fibers at a time, the statistical error on counting can be efficiently eliminated.
