中图分类号：TP391 文献标志码：A DOI: 10.16157/j.issn.0258-7998.257687
中文引用格式： 易云飞，史英杰，陈旺，等. 基于多尺度特征增强的改进YOLOv11n桑蚕健康状态实时检测研究[J]. 电子技术应用，2026，52(7)：142-150.
英文引用格式： Yi Yunfei，Shi Yingjie，Chen Wang，et al. Research on real-time detection of silkworm health status based on improved YOLOv11n with multi-scale feature enhancement[J]. Application of Electronic Technique，2026，52(7)：142-150.

Research on real-time detection of silkworm health status based on improved YOLOv11n with multi-scale feature enhancement

Abstract： This paper proposes an improved silkworm health detection algorithm based on YOLOv11n (MSDA-YOLO) to address the issues of delayed disease recognition and insufficient detection accuracy in traditional silkworm farming. The core innovations of the algorithm include: 1) introducing the Multi-scale Grouped Dilated Convolution in the backbone network and combining it with C3k2 to propose the C3k2MS module, which enhances the model’s ability to extract features at different scales; 2) introducing the DySample dynamic upsampler for the first time in silkworm disease detection, which improves the model’s ability to perceive spatial details and maintain geometric structure for small targets; 3) using the Adaptive Spatial Feature Fusion strategy to modify YOLOv11’s detection head, optimizing multi-scale feature fusion and suppressing feature conflicts between scales. Experimental results show that MSDA-YOLO achieves superior performance, with a precision of 87.0%, recall of 77.8%, mAP@0.5 of 87.3%, and mAP@0.5:0.95 of 52.8%, surpassing the baseline YOLOv11n and other mainstream YOLO models, demonstrating its clear advantage in silkworm health detection.

Key words : YOLOv11n；multi-scale grouped dilated convolution；DySample；adaptively spatial feature fusion