Discussion on coding rules and labelling specifications of defect image samples of substation equipment

In the image recognition of substation equipment defects based on deep learning, image sample coding rules and labeling specifications are one of the key factors to ensure the quality of sample data. At present, the image of the defect of the substation equipment is scarce, and the sample management specification is not uniform. There is a low level of image recognition for equipment defects in the substation intelligent inspection. Therefore, the paper carries out analysis and thinking on the coding rules and labeling specifications of the substation key equipment defect image samples. The sample making process is standardized through the standardization of the coding of typical substation equipment type, component type, defect type and annotation format, to some extent, which improves the accuracy of sample data quality and substation equipment defect recognition, and improves the level of defect management in intelligent inspections of substation. Finally, the results show that the standardized labeling samples help to improve the recognition accuracy and have wide application value.