Thermal Fault Detection of Power Equipments with Low-Rank Tensor Decomposition and Guided Filter
Zhihong Huang, Zuoyuan Jun · 2023
This work introduces a thermal fault detection method with low-rank tensor decomposition and guided filter (LRTD-GF). This method contains two main steps. First, based on the sparsity of the thermal fault and low rankness of the image background, thermal fault detection is modeled as a low-rank tensor decomposition, and an initial detection result can be obtained. Then, a guided filter technology produces the initial detection result, utilizing the spatial correlation between pixels in the spatial domain to improve the detection performance. Experiment results demonstrate that the LRTD-GF method has obvious advantages in detection accuracy, compared with other widely used anomaly detectors.