YOLO-Blueberry: A Lightweight and Accurate Multi-Scale Detection Model for Blueberry Maturity Recognition

Blueberry ripeness detection faces several persistent challenges, including large model parameter sizes, high visual similarity between fruits and background, and occlusion of target objects. To address these issues, this paper proposes a lightweight detection framework named YOLO-Blueberry. Firstly, we replace the Spatial Pyramid Pooling-Fast (SPPF) layer in YOLOv8n with a novel Grouped Large Kernel Reparameterization (GLKRep) module. This replacement significantly reduces computational complexity while enhancing contextual semantic perception, thereby improving the model’s ability to detect targets in visually complex orchard environments. Secondly, to accommodate substantial variations in fruit size and inconsistent image acquisition distances, we design a Unify module in the model’s neck network. This module aggregates multi-scale feature maps from the backbone and employs a dynamic receptive field selection mechanism to adaptively optimize multi-scale feature fusion, enhancing robustness to scale variation. Thirdly, during the training phase, we introduce VariFocal Loss in place of the original classification loss used in YOLOv8. By incorporating an IoU-aware classification score mechanism, this loss function improves the model’s discrimination between positive and negative samples, thereby reducing both false positives and false negatives. Experimental results demonstrate that YOLO-Blueberry achieves a mean Average Precision (mAP@0.5–0.95) of 97.5%, a precision of 97.8%, and a recall of 95.5%, with only 2.6 million parameters and 7.2 GFLOPs in computational cost. These results outperform several mainstream lightweight object detection models and validate the model’s suitability for deployment in real-world orchard scenarios. This study not only provides a key technical foundation for intelligent blueberry harvesting systems but also offers a viable reference approach for the automated detection of other small berry fruits in natural environments.