基于多尺度特征融合的厂站接线图的图元检测方法OA

To address the problems of missed detections and high false detection rates of small electrical symbols in high-resolution substation wiring diagrams,an improved YOLOv10-based electrical symbol detection method is proposed.First,a lightweight adaptive feature extraction module is introduced into the backbone network of YOLOv10 to enhance the multi-scale feature representation capability for small targets.Second,a weighted bidirectional feature pyramid structure is incorporated into the neck network to achieve efficient cross-layer feature fusion.Finally,the original detection head is replaced with an adaptive spatial feature fusion detection head,which suppresses conflicting feature information introduced during bidirectional feature fusion and further enhances spatial feature integration,thereby improving the detection accuracy of small targets.Experimental results show that the improved model achieves a precision of 95.9％and an mAP of 93.8％,representing improvements of 12.8 percentage points and 17.3 percentage points compared with the original YOLOv10s,respectively.Compared with existing mainstream object detection algorithms,the proposed method demonstrates superior detection performance and can provide an efficient and reliable technical solution for automatic information extraction and graph-model conversion of substation wiring diagrams.