Clustering wear particle using computer vision and self-organizing maps

This work presents the implementation of a method for classification of wear particle contaminant present in industrial oil by using image processing and neural networks. It is based on morphological data obtained from a computer vision system and employs Self-Organizing Maps to classify particles' features intro different wear debris groups. The dataset used for training the neural network and further validation of the results was gathered using reports provided by a specialist company in wear particle analysis. The objective is to develop a system feasible for most industries to turn the process of particle classification more autonomous and faster. The results demonstrate that our proposed system could classify particles considering their shape in a reliable and autonomous way.