Violation Detection Based on Deep Learning in Substation Scenario
This paper presents a deep learning-based algorithm,namely SEC-YOLO,to address the inefficiencies,high costs,and error-prone nature of manual inspections in power grid operations.Based on the YOLOv7 framework,the algo-rithm incorporates the Swin Transformer attention mechanism in the small object detection layer to enhance model sensitivity and improve accuracy in recognizing small targets.Furthermore,it introduces an Explicit Visual Center Block(EVCBlock)in the feature fusion section,enhancing the model's understanding of complex scenes for more accurate detection of violations in substation environments.Experimental results indicate a 5.26%and 3.77%improvement in mAP and recall metrics.
万方数据
维普
