Radiometric correction of multispectral uas images: Evaluating the accuracy of the parrot sequoia camera and sunshine sensor

Research output: Contribution to journalArticle

Standard

Radiometric correction of multispectral uas images : Evaluating the accuracy of the parrot sequoia camera and sunshine sensor. / Olsson, Per Ola; Vivekar, Ashish; Adler, Karl; Garcia Millan, Virginia E.; Koc, Alexander; Alamrani, Marwan; Eklundh, Lars.

In: Remote Sensing, Vol. 13, No. 4, 577, 02.02.2021.

Research output: Contribution to journalArticle

Harvard

APA

CBE

MLA

Vancouver

Author

RIS

TY - JOUR

T1 - Radiometric correction of multispectral uas images

T2 - Evaluating the accuracy of the parrot sequoia camera and sunshine sensor

AU - Olsson, Per Ola

AU - Vivekar, Ashish

AU - Adler, Karl

AU - Garcia Millan, Virginia E.

AU - Koc, Alexander

AU - Alamrani, Marwan

AU - Eklundh, Lars

PY - 2021/2/2

Y1 - 2021/2/2

N2 - Unmanned aerial systems (UAS) carrying commercially sold multispectral sensors equipped with a sunshine sensor, such as Parrot Sequoia, enable mapping of vegetation at high spatial resolution with a large degree of flexibility in planning data collection. It is, however, a challenge to perform radiometric correction of the images to create reflectance maps (orthomosaics with surface reflectance) and to compute vegetation indices with sufficient accuracy to enable comparisons between data collected at different times and locations. Studies have compared different radiometric correction methods applied to the Sequoia camera, but there is no consensus about a standard method that provides consistent results for all spectral bands and for different flight conditions. In this study, we perform experiments to assess the accuracy of the Parrot Sequoia camera and sunshine sensor to get an indication if the quality of the data collected is sufficient to create accurate reflectance maps. In addition, we study if there is an influence of the atmosphere on the images and suggest a workflow to collect and process images to create a reflectance map. The main findings are that the sensitivity of the camera is influenced by camera temperature and that the atmosphere influences the images. Hence, we suggest letting the camera warm up before image collection and capturing images of reflectance calibration panels at an elevation close to the maximum flying height to compensate for influence from the atmosphere. The results also show that there is a strong influence of the orientation of the sunshine sensor. This introduces noise and limits the use of the raw sunshine sensor data to compensate for differences in light conditions. To handle this noise, we fit smoothing functions to the sunshine sensor data before we perform irradiance normalization of the images. The developed workflow is evaluated against data from a handheld spectroradiometer, giving the highest correlation (R2 = 0.99) for the normalized difference vegetation index (NDVI). For the individual wavelength bands, R2 was 0.80–0.97 for the red-edge, near-infrared, and red bands.

AB - Unmanned aerial systems (UAS) carrying commercially sold multispectral sensors equipped with a sunshine sensor, such as Parrot Sequoia, enable mapping of vegetation at high spatial resolution with a large degree of flexibility in planning data collection. It is, however, a challenge to perform radiometric correction of the images to create reflectance maps (orthomosaics with surface reflectance) and to compute vegetation indices with sufficient accuracy to enable comparisons between data collected at different times and locations. Studies have compared different radiometric correction methods applied to the Sequoia camera, but there is no consensus about a standard method that provides consistent results for all spectral bands and for different flight conditions. In this study, we perform experiments to assess the accuracy of the Parrot Sequoia camera and sunshine sensor to get an indication if the quality of the data collected is sufficient to create accurate reflectance maps. In addition, we study if there is an influence of the atmosphere on the images and suggest a workflow to collect and process images to create a reflectance map. The main findings are that the sensitivity of the camera is influenced by camera temperature and that the atmosphere influences the images. Hence, we suggest letting the camera warm up before image collection and capturing images of reflectance calibration panels at an elevation close to the maximum flying height to compensate for influence from the atmosphere. The results also show that there is a strong influence of the orientation of the sunshine sensor. This introduces noise and limits the use of the raw sunshine sensor data to compensate for differences in light conditions. To handle this noise, we fit smoothing functions to the sunshine sensor data before we perform irradiance normalization of the images. The developed workflow is evaluated against data from a handheld spectroradiometer, giving the highest correlation (R2 = 0.99) for the normalized difference vegetation index (NDVI). For the individual wavelength bands, R2 was 0.80–0.97 for the red-edge, near-infrared, and red bands.

KW - Multispectral camera

KW - Radiometric correction

KW - Unmanned aerial systems

U2 - 10.3390/rs13040577

DO - 10.3390/rs13040577

M3 - Article

AN - SCOPUS:85100745544

VL - 13

JO - Remote Sensing

JF - Remote Sensing

SN - 2072-4292

IS - 4

M1 - 577

ER -