The purpose of this study is to provide a method for detecting moving wear particles by integrating spatiotemporal features, aiming to achieve fast and accurate localization and feature extraction of particles of various sizes in complex background images.
Moving particle analysis provides comprehensive information for characterizing the wear condition of mechanical systems. However, the detection accuracy of this promising technique is constrained by complex backgrounds. To address this, a two-step detection methodology is proposed by integrating the temporal and spatial features, involving a Gaussian mixture model-based coarse localization module for initial particle contour estimation and a Distance Regularized Level Set Evolution method-based precise detection strategy for boundary refinement. With this model, misdetections resulting from the background complexity can be corrected. Furthermore, morphological and statistical features are precisely extracted by matching and tracking detected multiview particle images across consecutive frames.
For verification, the proposed model is tested on fifteen samples of lubricant oil obtained from simulation experiments and industrial robot reduction gearboxes. The results demonstrate that the proposed method effectively detected moving particles from lubricant oil, with improvements in the accuracy of particle feature extraction from 70.2% to 89.5% compared to traditional methods.
Compared to traditional methods that rely solely on temporal or spatial information, the proposed method enhances the accuracy of particle localization and feature extraction.
The peer review history for this article is available at: https://publons.com/publon/10.1108/ILT-01-2025-0022/
