Product quality recognition and its industrial application based on lightweight machine learning

Abstract

Quality management plays a crucial role in manufacturing industry as it directly impacts product quality and customer satisfaction. Traditional manufacturing industries have begun incorporating deep learning models to manage production quality. However, existing studies often focus on algorithmic models that require high computing power, while practical industrial applications often suffer from a lack of labelled data. This article proposes a novel and lightweight machine-learning-based quality detection model to address this issue. The proposed model tackles quality detection by efficiently detecting the local maxima of the scale space in input images using a Hessian matrix detection method, eliminating the need for continuous computation of multilayer Gaussian difference images. Additionally, image processing techniques are employed to augment the training data, enabling the model to achieve high accuracy even with limited datasets. Experimental results demonstrate that the proposed model outperforms alternative methods in terms of both accuracy and efficiency.