Component state detection method and related device for power inspection

Abstract

The application belongs to the technical field of power systems, and discloses a component state detection method for power inspection and related devices, which comprises the following steps: acquiring a target component image collected in a power system inspection process; constructing and optimizing a target detection model according to a SimAM mechanism and an LKDAF module, wherein the target detection model comprises a deep and shallow feature information multi-scale fusion mechanism; inputting the target component image into the target detection model for target state detection to obtain a component state detection graph, wherein the component state detection graph comprises a target detection frame and a state probability. The application can solve the technical problem that the existing detection model is prone to missing shallow detail features, resulting in missed detection, and lacks a targeted feature processing mechanism and a good multi-scale fusion mechanism, thereby leading to poor actual detection effect.