Abstract: Blue honeysuckle is a small fruit crop valued for its nutritional properties. During field growth, it faces significant challenges in image feature extraction due to the clustering of fruits at different maturity levels, partial occlusion by branches and leaves, and highly variable illumination conditions caused by changing weather and sun angles. Previous studies have demonstrated that single-stage object detection algorithms, particularly improved YOLO models incorporating attention mechanisms, optimized feature extraction networks, or enhanced small-target detection capabilities, can achieve high mean average precision in complex field environments for fruit maturity detection tasks, exhibiting strong detection performance and applicability. This indicates that the YOLO framework possesses robust generalization ability. However, existing YOLO models are often highly specific to the growth environment, growth patterns such as clustered growth and dense fruiting, and detection metrics of particular fruits or vegetables. This specificity results in poor transferability across different detection objects, metrics, and requirements, while also struggling to ensure adaptability for deployment on resource-constrained mobile devices essential for real-time field applications.To improve the accuracy and precision of blue honeysuckle fruit maturity identification in field environments, this study proposes a detection model named YOLOv11s-ACM based on the YOLOv11s algorithm. First, to address the problem of low accuracy in blue honeysuckle maturity detection under field conditions, the Attention-based Intrascale Feature Interaction module is introduced into the backbone network. By incorporating a globally content-adaptive attention mechanism that selectively emphasizes informative features within the same scale, this module enhances detection accuracy for blue honeysuckle maturity in the field. Second, the C3K2 convolution in the backbone network is replaced with a C3K2 structure fused with Dynamic Snake Convolution. Through residual connections and dynamic feature aggregation to adaptively capture elongated structures, this modification improves the model's adaptability to occluded and low-illumination scenes, thereby enhancing detection capability for irregularly shaped blue honeysuckle fruits in field environments. Third, the Multi-Separated and Enhancement Attention Module Head is introduced to strengthen the model's capacity for detecting dense targets with multi-scale features under field conditions.Experimental results demonstrate that compared to the baseline YOLOv11s model, the proposed YOLOv11s-ACM achieves a 4.5 percentage point improvement in mean average precision and a 27% reduction in inference time per image. Furthermore, when compared against YOLOv5s, YOLOv7, YOLOv7-Tiny, YOLOv8s, YOLOv11n, YOLOv11s, and YOLOv11m, the YOLOv11s-ACM attains mAP improvements of 6.0, 5.9, 2.6, 5.8, 12.8, 4.5, and 5.5 percentage points respectively, demonstrating significantly enhanced overall performance. Mobile deployment validation confirms that the model's detection accuracy and inference speed satisfy the requirements for rapid and accurate detection of blue honeysuckle fruits under various field conditions, including different lighting scenarios such as direct sunlight and deep shadows, and occlusion levels ranging from light leaf coverage to heavy branch obstruction.Future research will focus on constructing large-scale real-world datasets containing extreme lighting conditions, severe occlusion, and diverse backgrounds to further optimize model adaptability. Additionally, adaptation and optimization of next-generation YOLO models for mobile terminals will be pursued to comprehensively enhance model robustness and lightweight deployment capabilities in complex environments, ultimately promoting the robust application of this technology in practical agricultural production.