Shape aggregates classification using activation function based MLP network

Traditionally, the quality of aggregates has been assessed through mechanical sifting and hand grading methods, which involve manual and time-consuming mechanical, chemical, and physical tests. However, this study aims to develop a more efficient image-based classification system for categorizing aggregates. To achieve this, an artificial neural network was employed to process the captured images and classify their shapes. The aggregate images serve as the input parameter for prediction before undergoing the threshold process. The study found that the Tansig activation function, which is based on a Multilayer Perceptron (MLP) network, performed better than the Purelin activation functions, exhibiting a lower mean square error (MSE) of 1.5237 and a higher regression capacity of 0.9728.