Automated Quality Assessment of Stone Aggregates Based on Laser Imaging and a Neural Network

Abstract

An automated quality assessment technique is proposed for rapidly detecting excessive size variations during the production of stone aggregates. The system uses a laser profiler to scan collections of aggregate particles and obtain three-dimensional data points on the particle surfaces. For computational efficiency, the resulting data are converted into digital images. Wavelet transforms are then applied to the images to extract features indicative of the material gradation. These wavelet-based features are used as inputs to an artificial neural network, which is trained to classify the aggregate sample. Taken together, these components form a neural network-based classification system that can determine whether or not an aggregate product is in compliance with a given specification. Verification tests show that this approach could potentially help to determine, in an accurate and fast (real-time) manner, when adjustments or repairs to the production equipment are needed.