| 摘 要: 针对传统的鼠洞、秃斑检测方法速度慢、过程烦琐等问题,提出了一种基于改进 YOLOv7的鼠洞、秃斑检测模型。首先,在 YOLOv7的主干网络中引入BiFormer注意力机制;其次,引入CARAFE算子作为上采样算法;最后,将 MPDIoU 代替原模型的 CIoU 损失函数。实验结果表明,改进模型的准确率、召回率、平均精度均值分别达到了91.6%、86.0%和90.9%,较原 YOLOv7模型分别提升了4.30个百分点、3.40个百分点和5.30个百分点。该模型为高寒草甸高原鼠兔鼠洞、秃斑的精准监测与管理提供了有效的技术支持。 |
| 关键词: 鼠洞 秃斑 注意力机制 CARAFE算子 损失函数 |
|
中图分类号: TP391.4
文献标识码: A
|
|
| Detection of Pika Holes and Bare Patches in Alpine Meadows Based on YOLOv7 |
|
LI Jiazhen1, WANG Lianguo1, HUA Limin2, KONG Yali1, YANG Yang1
|
(1.College of Information Science and Technology, Gansu Agricultural University, Lanzhou 730070, China;
2.College of Grassland Science, Gansu Agricultural University, Key Laboratory of Grassland Ecosystem of the Ministry of Education, Engineering and Technology Research Center for Alpine Rodent Pest Control, National Forestry and Grassland Administration, Lanzhou 730070, China)
1553705843@qq.com; wanglg@gsau.edu.cn; hualm@gsau.edu.cn; 2802448740@qq.com;; yy15928107241@outlook.com
|
| Abstract: In response to the slow and cumbersome process of traditional methods for detecting pika holes and bare patches, this study proposes an improved YOLOv7-based detection model for pika holes and bare patches. First,the BiFormer attention mechanism was introduced into the backbone network of YOLOv7. Second, the CARAFE operator was adopted as the upsampling algorithm. Finally, MPDIoU was used to replace the original CIoU loss function. The results show that the precision, recall, and mean average precision of the improved model reached 91.6% ,
86.0% , and 90.9% , respectively, representing improvements of 4.30 percentage points, 3.40 percentage points, and 5.30 percentage points compared to the original YOLOv7 model. This model provides effective technical support for the precise monitoring and management of pika holes and bare patches in alpine meadows. |
| Keywords: pika holes bare patches attention mechanism CARAFE operator loss function |
