Master of Engineering Science
Electrical and Computer Engineering
Planetary Science and Exploration
Cloud coverage has a significant impact on the reflectance maps generated by multispectral cameras mounted on UAV's in small farm settings. The current approach is to calibrate the camera once before every flight and use the radiometric calibration map for the entire flight. In this work we have designed and built a downwelling and upwelling photosensor to work in sync with our custom multispectral camera. A dual reflectance panel implementation for the time dependent radiometric calibration of the multispectral camera is completed. The solar spectral irradiance curve is approximated using the current measurements from the downwelling photosensor and the ground truth curve using a spectrometer. A multilinear regression algorithm was used for this purpose. The solar spectral irradiance curves are used during image acquisition to modify the initial radiometric calibration mapping function. This has potential to improve the usage of the multispectral camera under a wide range of weather conditions.
Summary for Lay Audience
Using multispectral cameras in farms fields is a rapid and accurate method of gathering the data needed to optimize crop growth. Catching diseases spread early can be derived from the data obtained from flying a multispectral camera mounted on an unmanned aerial vehicle over a farm field. The work here aims to provide a solution to one of the current drawbacks with multispectral cameras, the inability to capture accurate image data with change ambient light conditions due to cloud coverage.
This thesis focused on building a dual photosensor instrument which can be matched with a wide range of multispectral cameras in the visual and near infrared spectrum. One of the photosensors called the downwelling photosensor measures the sky with its lens such that it can correct the images from the multispectral camera. The dual photosensor instrument is integrated with a custom multispectral camera so it is built an interfacing microprocessor, which then sends the incoming light irradiance data to the main camera processor.
The multispectral camera is initially relatively radiometrically calibrated before flight using a dual reflectance panel method implemented in this thesis work. This technique allows for increased reflectivity measurement accuracy when imaging with the multispectral camera. The algorithm used to detect both panels used two different object recognition algorithms: the first assumed that the panels lay on a dark surface and the other did not. This work enabled efficient time series measurements of the reflectance panels needed to correct for the cloud coverage effects. Finally, the photosensor data was used to correct for the changing incoming irradiance due to the clouds. The first step was to model the incoming irradiance using the obtained photosensor current data. The solar spectral irradiance was modeled, and the results were tailored to our custom multispectral camera. A radiometric calibration correction model was developed with information from a single channel of the camera, and the by sharing information across all channels to improve the reflectance accuracy of the output image. The correction model showed positive results in which images that were effect by clouds were successfully corrected for by the dual photosensor instrument.
Mitchell, Nicholas S., "Radiometric Correction of MultiSpectral Cameras Using Photosensor Irradiance for Agronomy Applications" (2020). Electronic Thesis and Dissertation Repository. 7539.
Available for download on Friday, December 16, 2022