Image splicing and semantic segmentation–based omni-directional monitoring for substation equipment

Aiming at the problems of restricted video monitoring angles, significant traces of image splicing, insufficient coverage of equipment detection, and low efficiency of inspection report generation in the unattended operation and maintenance process of intelligent substations, this paper proposes a panoramic equipment monitoring method that integrates image splicing and semantic segmentation. Through the multi-stage fusion strategy, the infrared and visible images are accurately spliced at different scales, the alignment and fusion problems between heterogeneous images are effectively solved, and the immersive images with high visual consistency are generated. The deep learning-based semantic segmentation model is further constructed and optimized, and combined with the migration training mechanism and optimized suppression strategy, to complete the fine-grained recognition of various types of electrical equipment. Detailed descriptions of the multi-stage splicing framework and the semantic segmentation pipeline are provided to clarify their roles in panoramic equipment perception. The actual engineering application results show that the method performs well in terms of image processing accuracy and equipment detection accuracy, and at the same time significantly improves the automation level of remote inspection, which has good promotion prospects and engineering practical value.