<p>The increasing demand for blueberries necessitates efficient and reliable methods for ripeness detection to support automated harvesting and intelligent inspection. Traditional approaches based on manual visual assessment are inefficient and prone to subjective bias, limiting their applicability in standardized production. To address these challenges, this study proposes YOLOv8-BPDown-SynergyNet-BranchFuser (YOLOv8-BPSNB), an improved detection model based on the You Only Look Once (YOLO) framework. The model integrates a Lightweight Feature Distillation (LFD) module to enhance multi-scale feature representation, incorporates a SynergyNet module for feature alignment and fusion, and employs a BranchFuser module to improve high-level feature reconstruction for ripeness discrimination. Experiments conducted on a dataset of 2,400 images demonstrate that YOLOv8-BPSNB improves precision and recall by 3.2% and 4.5%, respectively. The proposed model consistently outperforms existing methods in terms of precision, recall, mAP50, and mAP50–95, achieving an mAP50–95 of 91.9%, which is 9.9% higher than the original YOLOv8. These results indicate that YOLOv8-BPSNB effectively captures discriminative features under complex conditions, making it well-suited for accurate blueberry ripeness detection. Future work will focus on improving inference efficiency and model lightweighting to further support practical deployment.</p>