Topic

Technologies and technical equipment for agriculture and food industry

Volume

Volume 75 / No. 1 / 2025

Pages : 346-355

Metrics

Volume viewed 0 times

Volume downloaded 0 times

FARMLAND OBSTACLE RECOGNITION BASED ON IMPROVED FASTER R-CNN

基于改进FASTER R-CNN的农田障碍物识别

DOI : https://doi.org/10.35633/inmateh-75-29

Authors

Xiangyu BAI

College of Electrical and Mechanical Engineering, Qingdao Agricultural University;National Key Laboratory of Intelligent Agricultural Power Equipment

Kai ZHANG

National Key Laboratory of Intelligent Agricultural Power Equipment

Ranbing YANG

College of Electrical and Mechanical Engineering, Qingdao Agricultural University;College of Mechanical and Electrical Engineering, Hainan University

Zhiguo PAN

College of Electrical and Mechanical Engineering, Qingdao Agricultural University

Huan ZHANG

College of Electrical and Mechanical Engineering, Qingdao Agricultural University

(*) Jian ZHANG

College of Electrical and Mechanical Engineering, Qingdao Agricultural University;College of Mechanical and Electrical Engineering, Hainan University

Xidong JING

College of Electrical and Mechanical Engineering, Qingdao Agricultural University

Shiteng GUO

College of Electrical and Mechanical Engineering, Qingdao Agricultural University

Sen DUAN

College of Electrical and Mechanical Engineering, Qingdao Agricultural University

(*) Corresponding authors:

zhangjian_qau@163.com |

Jian ZHANG

Abstract

For the accurate detection of obstacles in complex farmland environments, ResNet50 is adopted as the backbone feature extraction network, feature pyramid network (FPN) is utilized to enhance the multi-scale feature fusion capability, and the region of interest alignment (ROI Align) strategy is introduced to improve the candidate box localization precision. The experimental results show that the precision, recall, and mean accuracy (mAP) of the improved model are 91.6%, 89.7%, and 93.8%, respectively, which are improved by 2.7, 2.3, and 3.1 percentage points compared with the original base network, and provide a technical reference for navigation and obstacle avoidance of unmanned agricultural machinery.

Abstract in Chinese

针对复杂农田环境中障碍物的准确检测，采用ResNet50作为骨干特征提取网络，利用特征金字塔网络（FPN）提升多尺度特征融合能力，并引入感兴趣区域对齐（ROI Align）策略提高候选框定位精度。实验结果显示，改进模型的精度、召回率和平均精度（mAP）分别为91.6%、89.7%和93.8%，相比于原基础网络，提升了2.7、2.3和3.1个百分点，为无人农业机械的导航避障提供了技术参考。

Indexed in

Road