Scholarlink

Aims <p>This study presents a novel drone-based method for acquiring navigation coordinates in potato fields to support autonomous agricultural machinery.</p> Methods <p>First, a drone equipped with a high-precision RTK (Real-Time Kinematic) system was used to capture geospatial imagery of potato fields, thereby establishing a potato seedling dataset. A YOLOv8-based approach was proposed to detect the center points of potato seedlings during their early growth stage. This approach replaced the backbone network with MobileNetV3, integrated a Convolutional Block Attention Module (CBAM), and employed SIoU to enhance performance. Subsequently, the center pixel coordinates of each potato plant row were extracted from the detected bounding boxes, fitted using the least squares method, and transformed into spatial coordinates to generate crop row navigation information with actual geographic coordinates.</p> Key results <p>Validation on the potato seedling detection dataset demonstrated that the algorithm achieved an accuracy of 87.7%, a recall of 86.4%, and a mean average precision (mAP) of 90.4%. Compared to SSD, Faster R-CNN, YOLOv5, YOLOv7, YOLOv8 and YOLOv11, the mAP increased by 4.77%, 6.22%, 2.69%, 4.13%, 2.56%, and 5.14%, respectively. The algorithm's compact model size, rapid detection speed, and high precision make it particularly suitable for model migration and deployment.</p> Conclusion <p>Field experiments further confirmed that the generated navigation lines accurately guided tractors along potato ridges. Overall, this study provides a new solution for tractor-assisted navigation coordinate acquisition.</p> Implications and impacts <p>By addressing the limitations of traditional navigation coordinate acquisition methods, this work offers a new approach that significantly reduces labor intensity and crop damage, while promoting the broader adoption of unmanned agricultural machinery in precision agriculture.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Agricultural machinery navigation coordinate extraction driven by YOLOv8-MCS-based potato seedling recognition

  • 期刊论文

  • 发表日期：

    2025年10月14日

查看全文
  • Jian Zhang,
  • Shiting Lyu,
  • Xiangyu Bai,
  • Ranbing Yang,
  • Yang Liu,
  • Huihui Wen

摘要

Aims

This study presents a novel drone-based method for acquiring navigation coordinates in potato fields to support autonomous agricultural machinery.

Methods

First, a drone equipped with a high-precision RTK (Real-Time Kinematic) system was used to capture geospatial imagery of potato fields, thereby establishing a potato seedling dataset. A YOLOv8-based approach was proposed to detect the center points of potato seedlings during their early growth stage. This approach replaced the backbone network with MobileNetV3, integrated a Convolutional Block Attention Module (CBAM), and employed SIoU to enhance performance. Subsequently, the center pixel coordinates of each potato plant row were extracted from the detected bounding boxes, fitted using the least squares method, and transformed into spatial coordinates to generate crop row navigation information with actual geographic coordinates.

Key results

Validation on the potato seedling detection dataset demonstrated that the algorithm achieved an accuracy of 87.7%, a recall of 86.4%, and a mean average precision (mAP) of 90.4%. Compared to SSD, Faster R-CNN, YOLOv5, YOLOv7, YOLOv8 and YOLOv11, the mAP increased by 4.77%, 6.22%, 2.69%, 4.13%, 2.56%, and 5.14%, respectively. The algorithm's compact model size, rapid detection speed, and high precision make it particularly suitable for model migration and deployment.

Conclusion

Field experiments further confirmed that the generated navigation lines accurately guided tractors along potato ridges. Overall, this study provides a new solution for tractor-assisted navigation coordinate acquisition.

Implications and impacts

By addressing the limitations of traditional navigation coordinate acquisition methods, this work offers a new approach that significantly reduces labor intensity and crop damage, while promoting the broader adoption of unmanned agricultural machinery in precision agriculture.