Poster Presentation Freshwater Sciences 2023

Data Collection Challenges of a Hyperspectral Camera Mounted on a Large Drone (#667)

Abhishek Verma 1 , Abhimanyu Ghosh 1 , Yuan Xing 1
  1. Engineering and Technology, University of Wisconsin-Stout, Menomonie, Wisconsin, United States

The authors are currently engaged in a project that studies the flow of contaminants from farm fields to open water channels. A Resonon Pika-L hyperspectral camera is used for imaging, mounted on a DJI Matrice 300 drone, to capture images of farm fields and nearby water channels. Since many users, like the authors, are novices in at least one of the two technologies mentioned above, a lot of avoidable errors creep in when integrating the two. Here are some of the early challenges we faced and recommendations to overcome them, in trying to collect hyperspectral images from this setup.

  1. Your drone GPS system may not receive data from enough satellites, likely due to electromagnetic interference (EMI). When installing the GPS/IMU unit on the airborne system, therefore, ensure that the GPS antennae are as far away as possible from other sources of EMI, cable runs are short and avoid coiling of the antennae lines, and the orientation of the GPS/IMU is in its flight position while acquiring satellites.
  2. Calibration of the magnetometer may fail in two possible situations. Firstly, there may be heavy magnetic interference from indoor objects (if calibrating indoors) or even large objects like farm machinery or metal roofs (if calibrating outdoors). Secondly, the gimbal may not be locked tightly in place, leading to unexpected movements while carrying out the recommended motions.
  3. Severe distortions in the images may be caused by an out-of-focus lens or excessive vibrations from the drone. In the first case, refocus the lens by completing radiometric calibration. In the second case, use a different drone mount that adds a stronger damping effect to the gimbal.
  4. Beware of software-related issues. If no data is recorded at the end of the flight, check if the takeoff or landing zone is inside the target area, or if ground elevation from sea level is input incorrectly which can confuse the airborne system to think that it is not in the air. Incorrect configuration of elevation can also result in image distortion when applying georectification to the dataset.