Aiming at the problems of difficult multi-source heterogeneous data fusion, limited resources of edge terminals, and insufficient recognition accuracy of small-target faults in power inspection and equipment state perception, this paper proposes a multi-source data fusion edge pattern recognition method based on a lightweight deep network. The revised method constructs a cloud-edge-end collaborative recognition architecture, clarifies the original difference-guided local–global feature extraction design, adds a multi-dimensional coordinate collaborative attention mechanism for multimodal and temporal feature competition, and strengthens small-target fault detection through a high-resolution four-head multi-scale structure. The experimental section further reports dataset statistics, hyperparameter settings, state-of-the-art baselines, and robustness comparisons under strong illumination, rainy/foggy, occlusion, and cross-scenario conditions. Experimental results show that the proposed method achieves 99.2% mAP in power defect detection tasks. The compressed model maintains 98.7% mAP while improving inference speed and deployment feasibility on edge platforms such as Jetson Nano and Raspberry Pi 4B.
原文链接
分享
引用
