Fragmentation Assessment Using a New Image Processing Technique Based on Adaptive Neuro Fuzzy Inference System (ANFIS)

Computational techniques in determining particle size distributions after blasting is getting wide acceptance. A well known approach to extract this kind of information from digital images is edge detection. The number of edge points on a given image increase proportional to the increase of the particles on the image. By selecting a conveniently smooth threshold, counting the signed edge pixels may be a suitable metric further used to quantify the number of particles in the image. Here we propose an image processing algorithm together with Adaptive Neuro Fuzzy Inference System (ANFIS) to efficiently count the rock particles with various sizes on a given surface. To determine a suitable approximation for a mapping between the number of edge pixels and the number of particles, ANFIS is a suitable tool. As a trainable classifier ANFIS proved itself as being better efficient approach than other well known multilayer perceptron classifiers. ANFIS is trained by using convenient training data pairs previously experienced from empirically counted sample images. When given a new image containing the particle information of selected surface, the hybrid neuro fuzzy image processing approach provides satisfying performance on counting the correct number of particles up to the given scale of sizes. In this paper an experimantal study for assessment of particle size using ANFIS is presented.