English

中国农机化学报

中国农机化学报 ›› 2024, Vol. 45 ›› Issue (9): 209-214.DOI: 10.13733/j.jcam.issn.2095-5553.2024.09.032

• 农业信息化工程 • 上一篇    下一篇

基于无人机遥感的大豆倒伏识别研究

吴宇通 1,2,张伟 1,2,3,石文强 1,2,李金阳 1,2,亓立强 1,2   

  1. (1.黑龙江八一农垦大学工程学院,黑龙江大庆,163319; 2.黑龙江省保护性耕作工程技术研究中心,黑龙江大庆,163319; 
    3.农业农村部大豆机械化生产重点实验室,黑龙江大庆,163319)
  • 出版日期:2024-09-15 发布日期:2024-09-04
  • 基金资助:
    财政部和农业农村部:国家现代农业产业技术体系资助(CARS—04—PS30);北方寒地机械化保护性耕作技术创新研究团队(TDJH201808)

Study on soybean lodging identification based on UAV remote sensing 

Wu Yutong1,2,Zhang Wei1,2,3,Shi Wenqiang1,2,Li Jinyang1,2,Qi Liqiang1,2   

  1. (1. College of Engineering,Heilongjiang Bayi Agricultural University,Daqing,163319,China; 2. Heilongjiang Province Conservation Tillage Engineering Technology Research Center,Daqing,163319,China; 3. Key Laboratory of Soybean Mechanization Production,Ministry of Agriculture and Rural Affairs,Daqing,163319,China) 
  • Online:2024-09-15 Published:2024-09-04
  • Supported by:

摘要:

为快速识别大豆倒伏情况,准确提取大豆倒伏面积,提出基于无人机遥感技术的方法对大豆倒伏情况进行判断。采用无人机获取大豆鼓粒期冠层可见光(RGB)图像及数字表面模型(DSM)图像,提取可见光波段信息并构建过绿植被指数(EXG)图像,将 3类特征图像进行图像特征融合,得到 DSM+RGB融合图像,DSM+EXG+RGB融合图像。利用最大似然法对 4种特征融合图像进行监督分类提取大豆倒伏面积,利用混淆矩阵方法验证各图像分类精度。结果表明,RGB图像、DSM图像、DSM+RGB特征融合图像、DSM+EXG+RGB特征融合图像提取倒伏大豆面积的整体精度分别为78. 36%、65. 38%、82. 84%、68. 41%。Kappa系数分别为 0. 75、0. 53、0. 81、0. 58,DSM+RGB特征融合图像提取大豆倒伏面积精度最高。图像特征融合方法可用于评估大豆倒伏情况,为快速提取大豆倒伏面积提供参考。

关键词: 大豆, 倒伏, 无人机, 特征融合, DSM, 最大似然法

Abstract:

To quickly identify the soybean lodging conditions,accurate extraction of the soybean lodging area,a method based on Unmanned Aerial Vehicle(UAV)remote sensing technology is proposed to judge the soybean lodging situation. Red.Green.Blue(RGB) images and digital surface model Digital Surface Model,(DSM) images of soybean drums were obtained by UAV,visible light information was extracted and Excess Green(EXG)images were constructed,by using 3 types of feature images into image features,DSM + RGB fusion images and DSM + EXG + RGB fusion images were obtained. The maximum likelihood method was used to extract the soybean lodging area,and to verify the classification accuracy of each image by using the confusion matrix method. The results showed that the overall accuracy of the lodging soybean area extracted by RGB images,DSM images,DSM + RGB feature fusion images,and DSM + EXG + RGB feature fusion images were,respectively 78. 36%,65. 38%,82. 84%,68. 41%. The Kappa coefficients were 0. 75, 0. 53,0. 81,and 0. 58,respectively,with the highest accuracy in soybean lodging area extracted from DSM + RGB feature fusion images. Image feature fusion method can be used to evaluate soybean lodging and provide a reference for rapid extraction of soybean lodging area.

Key words: soybean, lodging, UAV, characteristic fusion, digital surface model, maximum likelihood method

中图分类号: