Multispectral Sensor Calibration and Characterization for sUAS Remote Sensing.

Area of Science:

• Remote Sensing
• Optical Engineering
• Photogrammetry

Background:

• Multispectral sensors used in remote sensing rely on accurate radiometric calibration and vignette correction parameters.
• Factory-provided calibration parameters can degrade over time due to sensor usage in real-world conditions.
• Regular recalibration is essential to maintain sensor accuracy for reliable data acquisition.

Purpose of the Study:

• To develop and present an end-user laboratory method for computing vignette correction and radiometric calibration functions.
• To compare the proposed laboratory method against the MicaSense factory-provided method for radiance computation.
• To quantify errors in radiance, reflectance, and vegetation indices (NDVI, NDRE) using both calibration methods.

Main Methods:

• An end-user laboratory procedure was established for radiometric calibration and vignette correction.
• The proposed method and MicaSense's method were applied to laboratory images captured using a traceable light source.
• Error propagation analysis was performed to assess accuracy in digital counts to radiance conversion, reflectance, NDVI, and NDRE.

Main Results:

• The proposed laboratory method demonstrated significantly lower average percent errors in radiance across multiple spectral bands compared to MicaSense's factory method.
• Errors in radiance for the proposed method ranged from 0.72% to 3.70%, while MicaSense's method showed errors from -17.08% to 32.81%.
• The proposed method produced more accurate NDVI (0.897 ± 0.007 vs. 0.876 ± 0.005) and NDRE (0.435 ± 0.038 vs. 0.239 ± 0.026) values compared to the ground reference.

Conclusions:

• The developed end-user laboratory method provides a more accurate approach to radiometric calibration and vignette correction for multispectral sensors.
• This improved calibration methodology leads to reduced errors in radiance and reflectance data, enhancing the reliability of remote sensing applications.
• The proposed method significantly improves the accuracy of vegetation health indices (NDVI, NDRE), crucial for ecological and agricultural monitoring.