Transmission line foreign object target detection incorporating inverse residual shift convolution

To address weak small-object detection, high redundancy, and significant background interference in foreign object detection for power transmission lines, this paper proposes a lightweight algorithm based on YOLOv11n. An Inverted Residual Shifted Convolution (ISWC) module is designed to enhance the C3k2 structure, improving local receptive fields and feature sensitivity for small-object extraction. An Adaptive Downsampling (ADown) module is introduced to reduce parameters and computation while retaining key features. Additionally, the Inner-WIoU loss function is employed to optimize bounding box regression, enhancing detection accuracy and training stability. Experiments on the Southern Power Grid dataset show that the proposed method achieves a 92.3% mAP, improving by 1.3% over the baseline, with 22.9% and 23.5% reductions in parameters and computation, respectively.