Elements: A Deep Neural Network-based Drone (UAS) Sensing System for 3D Crop Structure Assessment

Elements：用于 3D 作物结构评估的基于深度神经网络的无人机 (UAS) 传感系统

• 批准号: 2104032
• 负责人: Guoyu Lu
• 金额: $ 58.37万
• 依托单位: Rochester Institute of Tech
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-06-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2104032&HistoricalAwards=false
• 关键词: Elements; Deep; Neural; Network; based

This project develops a 3D reconstruction sensing system that can be installed on unmanned aerial systems (UAS), to be used by agricultural researchers, growers, and service providers to assess crop growth. Applying Artificial Intelligence (AI) technology for large scale agriculture reconstruction applications, the sensing system would be able to estimate crop structure for a large coverage area at a much lower cost than current standards that rely on light detection and ranging (LiDAR). The project would develop and refine a deep neural network-based 3D assessment workflow, based solely on a low cost and lightweight 2D LiDAR and color camera configuration. Researchers, growers, and service providers would be able to extract detailed crop structure and forecast yields, based on a 3D time series of crop growth. The technology would provide a less expensive alternative to the current 3D LiDAR sensor approach, and the sensing system could also be applied to related areas such as high-throughput phenotyping and variation estimation of general terrestrial vegetation. Outreach and extension activities are included, to deliver research outcomes to the stakeholders, including agricultural researchers, growers and service providers. PhD students, undergraduates, and high school students will be trained through this project, including a summer activity training high school students through the Rochester Institute of Technology Imaging Science High School Summer Intern Program.This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Division of Biological Infrastructure within the NSF Biosciences Directorate, and by the Division of Information and Intelligent Systems within the NSF Computer and Information Science and Engineering Directorate.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该项目开发了一种3D重建传感系统，可安装在无人机系统（UAS）上，供农业研究人员、种植者和服务提供商用于评估作物生长。 将人工智能（AI）技术应用于大规模农业重建应用，传感系统将能够以比依赖光探测和测距（LiDAR）的当前标准低得多的成本估计大覆盖区域的作物结构。 该项目将开发和完善基于深度神经网络的3D评估工作流程，仅基于低成本和轻量级的2D LiDAR和彩色相机配置。 研究人员、种植者和服务提供商将能够根据作物生长的3D时间序列提取详细的作物结构并预测产量。 该技术将为当前的3D LiDAR传感器方法提供一种更便宜的替代方案，并且传感系统还可以应用于相关领域，例如高通量表型分析和一般陆地植被的变化估计。 还包括外联和推广活动，向利益攸关方，包括农业研究人员、种植者和服务提供者提供研究成果。 博士生，本科生和高中生将通过这个项目进行培训，包括通过罗切斯特理工学院成像科学高中暑期实习计划培训高中生的暑期活动。高级网络基础设施办公室的这个奖项由NSF生物科学理事会生物基础设施部门共同支持，该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Self-supervised Depth Estimation from Spectral Consistency and Novel View Synthesis

根据光谱一致性和新颖视图合成进行自监督深度估计

• 发表时间: 2022
• 期刊: IEEE International Joint Conference on Neural Network (IJCNN
• 作者: Yawen Lu, Guoyu Lu

An Unsupervised Approach for Simultaneous Visual Odometry and Single Image Depth Estimation

同步视觉里程计和单图像深度估计的无监督方法

• 发表时间: 2022
• 期刊: IEEE International Joint Conference on Neural Network (IJCNN
• 作者: Lu, Yawen;Lu, Guoyu
• 通讯作者: Lu, Guoyu

Self-Supervised Single-Image Depth Estimation From Focus and Defocus Clues

• DOI: 10.1109/lra.2021.3092258
• 发表时间: 2021-10
• 期刊: IEEE Robotics and Automation Letters
• 影响因子: 5.2
• 作者: Yawen Lu;Garrett Milliron;John Slagter;G. Lu

3D SceneFlowNet: Self-Supervised 3D Scene Flow Estimation Based on Graph CNN

• DOI: 10.1109/icip42928.2021.9506286
• 发表时间: 2021-09
• 期刊: 2021 IEEE International Conference on Image Processing (ICIP)
• 作者: Yawen Lu;Yuhao Zhu;G. Lu

AGI for Agriculture

• DOI: 10.48550/arxiv.2304.06136
• 发表时间: 2023-04
• 期刊: ArXiv
• 作者: G. Lu;Sheng Li;Gengchen Mai;Jin Sun;Dajiang Zhu;L. Chai;Haijian Sun;Xianqiao Wang;Haixing Dai