关键词: YOLOv8n-LF模型, 机收小麦, 含杂率, 破碎率, 检测

Abstract: To address the need for real-time detection of wheat impurity and broken rates during combine harvesting, a detection model (YOLOv8n-LF) was proposed based on an image segmentation method to achieve automatic assessment of wheat impurity and broken rates. First, the LSKA attention mechanism was introduced into the SPPF module to enhance the model’s multi-scale feature extraction capability. Second, the focal adjustment mechanism of Focal Loss was employed to optimize the CIoU loss function, focusing on different regression samples and improving the model’s detection performance. On a self-constructed dataset, the YOLOv8n-LF model maintained lightweight characteristics while demonstrating good segmentation performance, facilitating its deployment on edge devices. It can provide effective technical support for the automatic detection of wheat impurity and broken rates, thereby promoting the advancement of agricultural machinery intelligence.

Key words: YOLOv8n-LF model, machine-harvested wheat, impurity rate, broken rate, detection