Image classification of sugar crystal with deep learning

Abstract

Nowadays, artificial intelligent control is essential to replace experienced workers. The correct classification of sugar crystals during the production process is the basis for the control of the sugar crystallization process. Correct Classification of sugar crystals is the basis necessary for automatic control of process. This research uses the principles of deep learning using a neural network to identify the crystallization of sugar from the actual production process of sugar factories in Thailand. Performance was measured and compared with the Fine-tuning VGG16 model. It was accurate to identify sugar crystals between 82% and 92% of four classes sugar crystal images classified by the crystallization conditions. The results of this study also show that this model is more accurate than other models. It can be used as a benchmark for monitoring the crystallization of sugar production processes. It is also the basis of an artificial intelligent control system based on transcribing human expertise