Visual odometry for enhanced spatial phenotyping in agriculture

Overview

This innovative solution employs a UAV (Unmanned Aerial Vehicle) equipped with stereo cameras and GPS to enhance spatial phenotyping in agricultural fields. By creating sparse reconstructions of vegetable layouts, this system improves the accuracy of GPS measurements. Through visual odometry, the motion of the UAV is determined, which reduces GPS errors and allows precise geotagging of phenotypes in open grid plots. This approach combines data from multiple sensors, significantly reducing manual labor and improving geospatial accuracy.

Technical specifications

• Key features:

  • Utilizes stereo camera systems to capture overlapping image pairs
  • Employs computer vision and machine learning for data processing
  • Integrates well-developed algorithms and open-source libraries such as Pytorch, OpenCV, and Detectron
  • Capable of utilizing monocular camera data with additional depth information
  • Designed for use in GPS-compromised environments like orchards

• Technical process:

  • Sequential stereo images determine UAV motion
  • Geospatially accurate phenotyping through geotagging
  • Error reduction by combining multiple GPS measurements

Technology readiness level

This technology is currently at TRL 3, indicating it is in the early stages of development. While it has been conceptualized and some components have been tested in controlled settings, further validation and development are needed to test its full application in field environments.