Abstract: Abstract: Advancing the intelligent transformation of peach orchard management is of great significance for ensuring the sustainable development of the peach industry. As a critical operation, the timing of harvesting directly determines the flavor quality of the fruit, thereby impacting the economic benefits of the industry. In large-scale cultivation models, the uneven spatial distribution of factors such as light and nutrients leads to variations in fruit maturity among different trees and even among different parts of the same tree. Traditional methods, which primarily rely on manual experience for sampling and discrimination, struggle to comprehensively and accurately assess the overall maturity status of the orchard. Furthermore, the differing ripening times of fruits in natural environments, compounded by complex interferences like foliage occlusion and fruit overlap, further increase the difficulty of accurately identifying peach maturity in orchards. According to the maturity standards for harvested peaches specified in the agricultural industry standard of the People's Republic of China, NY/T 586-2002, and the technical code for peach storage, GB/T 26904-2020, this study classifies peach fruit maturity into three categories based on color: 1) Unripe peach (up): the fruit surface is entirely green with no sign of color change; 2) Half-ripe peach (hp): the fruit surface shows a mix of red and green, with the color-changed area being less than 50%; 3) Ripe peach (rp): the color-changed area on the fruit surface exceeds 50%. To address the challenges, this paper proposes an improved YOLOv8n model for identifying peach maturity in orchards. First, the original Conv module was replaced with our proposed PMSEConv (Peach Multi-Scale Efficient Convolution) module to enhance the model's ability to capture multi-scale contextual features and reduce missed and false detections. Second, a novel Peach Occlusion Attention mechanism (POAttention) was introduced in the Neck section to improve the model's detection accuracy for peach maturity in high-density and occluded environments. Finally, the EIoU loss function was adopted to optimize bounding box regression, enhancing the model's adaptability and practicality in complex orchard environments. Ablation experiment results indicate that the improved YOLOv8n model possesses significant comprehensive performance advantages. Compared to the baseline YOLOv8n, its precision and recall rates increased by 5.2 and 2.8 percentage points, respectively, and the mean average precision (mAP@0.5) improved by 5.9 percentage points. Simultaneously, model complexity was effectively controlled, with FLOPs reduced by 0.5 G and parameters by 0.29 M, achieving a dual optimization of detection accuracy and computational efficiency. To further validate its advancement, this model was comprehensively compared with current mainstream object detection models, including RT-DETR, YOLOv3-tiny, YOLOv5, YOLOv6, YOLOv9, YOLOv10, YOLOv11, and YOLOv12. The comparative experimental results show that the improved YOLOv8n model achieves optimal results on key metrics: precision reaches 84.5%, recall reaches 76.0%, and mAP@0.5 reaches 83.8%, fully demonstrating its excellent performance in identifying fruit maturity in complex orchard environments. The lightweight and efficient characteristics of this model can provide strong technical support for peach-picking robots and intelligent orchard management systems, facilitating precise and efficient automated operations.