Using Local Maximum Difference Enhanced Texture and Multifeature Fusion for Evaluating the Aging of Insulating Paper in Microscopic Image

The conventional evaluation of insulating paper aging is inherently invasive and destructive, thereby posing challenges in satisfying the criteria for non-destructive testing. This study presents an innovative method for assessing the aging state of insulating paper by fusing its surface texture and color features, both extracted from microscopic images. To enhance the surface texture representation, the Local Maximum Difference (LMD) algorithm is introduced, and Local Binary Patterns (LBP) are subsequently utilized for the extraction of these enhanced texture features. Furthermore, a color correlogram is employed to comprehensively characterize the evolution of color and its distribution within the insulating paper. Finally, a Support Vector Machine (SVM) is leveraged to implement a data-driven aging assessment framework. The empirical results compellingly substantiate the efficacy of the devised methodology, exhibiting a noteworthy accuracy level of 93.33 % when evaluated against the testing dataset, thereby underlining its robustness and validity.