Full
MEN Xiaolong, E-mail: 2879955207@qq.com
Outline
To address the issues of low detection accuracy and poor robustness of apple picking robots caused by complex environmental factors such as branch and leaf occlusion and uneven light distribution in orchards, this study proposes an apple fruit detection algorithm that improves the lightweight model of the YOLOv8n network, aiming to enhance the accuracy and reliability of apple detection in complex scenarios. Firstly, an “MPCA” (multi-scale position-channel attention) is added at the end of the backbone network to precisely capture the position information of apples at different scales, thereby enhancing the model's perception of target position features and effectively reducing the interference of complex orchard environments on detection results. Secondly, a “FocusFeature” feature fusion module is introduced in the neck network, which uses multi-scale depth wise separable convolution to achieve information fusion and feature enhancement, improving the model's detection capability. Finally, the loss function CIoU is replaced with WIoUv3 (WiseIoU), dynamically adjusting the weight of samples in the loss calculation based on the matching quality of predicted and ground truth boxes, effectively enhancing the model's localization accuracy. Experiments show that the improved model achieves an accuracy of 92.2%, a recall rate of 86.4%, and mAP@0.5 and mAP@[0.5: 0.95] of 94.4% and 72.4% respectively in complex orchard environments, representing improvements of 3.4%, 6%, 3.9%, and 2.8% compared to the original algorithm. Ablation experiments reveal that WIoUv3, MPCA, and FocusFeature feature fusion module increase mAP@0.5 by 0.3%, 0.9%, and 1.1% respectively, with performance enhancement achieved through the synergy of multiple modules. The model size is 6.69 MB, only 0.74 MB larger thanthe original model, without significantly increasing the computational load. Compared with mainstream algorithms such as YOLOv5n, YOLOv7-tiny, and YOLOv11n, the improved model increases the accuracy by 3.9%, 7.6%, and 5.7% respectively, and mAP@[0.5: 0.95] by 2.5%, 9.9%, and 13%, demonstrating significant advantages in detection accuracy and generalization ability. It is evident that the improved YOLOv8n algorithm proposed in this study significantly enhances the accuracy and robustness of apple detection in complex orchard environments through the collaborative effect of multiple modules, providing an effective technical solution for the practical application of picking robots and playing a crucial role in promoting the intelligent development of apple harvesting.
30
8
| 科 Family | 属数 Number of genus | 种数 Number of species | 占总种数比例 Percentage of total species (%) | 属 Genus | 种数 Number of species | 占总种数比例 Percentage of total species (%) |
|---|---|---|---|---|---|---|
| 鹅膏菌科Amanitaceae | 2 | 11 | 5.26 | 鹅膏菌属 Amanita | 10 | 4.78 |
| 小菇科 Mycenaceae | 2 | 12 | 5.74 | 丝盖伞属 Inocybe | 5 | 2.39 |
| 多孔菌科 Polyporaceae | 8 | 14 | 6.70 | 蜡蘑属 Laccaria | 5 | 2.39 |
| 红菇科 Russulaceae | 3 | 23 | 11.00 | 小皮伞属 Marasmius | 6 | 2.87 |
| 小菇属 Mycena | 11 | 5.26 | ||||
| 光柄菇属 Pluteus | 5 | 2.39 | ||||
| 红菇属 Russula | 17 | 8.13 | ||||
| 栓菌属 Trametes | 5 | 2.39 |
关闭全屏
