| 摘 要: 针对复杂自然环境中茶叶病害的检测问题,提出一种创新的检测模型,旨在为茶叶病害的精确识别提供有力支持。模型设计中,增加了小目标检测层,提升了对微小目标的检测;引入了CA(CoordAttention)注意力机制,增强模型对细节信息的捕捉能力;采用Focal-EIOU损失函数进一步优化算法模型;替换主干网络MobilevitV2,提升了模型的性能。实验结果表明,在同等条件下,与YOLOv8n原模型相比,本研究提出的优化算法实现了平均精度均值3.5百分点的提升,准确率达到91.6%。这些改进措施有效地提高了茶叶病害检测的准确率,而且为茶叶病害检测提供了坚实的理论基础和技术支撑。 |
| 关键词: 目标检测;注意力机制;损失函数;Yolov8n |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 科技部国家外专项目(G2022042005L);甘肃省重点研发计划(23YFWA0013);甘肃省高等学校产业支撑项目(2023CYZC-54);兰州市人才创新创业项目(2021-RC-47);2023年甘肃农业大学美育和劳动教育教学改革项目(2023-09) |
|
| Detection of TeaLeaf Diseases Basedon Coordinate Attention Mechanism and Focal-EIOU |
|
LI Yingtao, WEI Linjing
|
(College of Information Science and Technology, Gansu Agricultural University, Lanzhou 730070, China)
937923581@qq.com; 916277964@qq.com
|
| Abstract: Aiming at the challenges of detecting complex tea leaf diseases in natural environments, this paper proposes an innovative detection model to provide robust support for the accurate identification of tea leaf diseases. The model design incorporates a small object detection layer to enhance the detection of tiny targets. Coordinate Attention (CA) mechanism is introduced to improve the model's capability to capture detailed information. Focal-EIOU loss function is used to further optimize the algorithm model. The backbone network MobilevitV2 has been replaced to enhance model performance. Experimental results demonstrate that under equivalent conditions, the proposed optimized algorithm achieves an average precision improvement of 3.5 percentage points, compared to the original YOLOv8n model, with an accuracy reaching 91.6%. These improvements effectively enhance the accuracy of tea leaf disease detection and provide a solid theoretical foundation and technical support for this task. |
| Keywords: object detection; attention mechanism; loss function; YOLOv8n |
