Radiometric calibration framework for ultra-high-resolution UAV-derived orthomosaics for large-scale mapping of invasive alien plants in semi-arid woodlands: Harrisia pomanensis as a case study

• Published in: International journal of remote sensing 2018-08, Vol.39 (15-16), p.5119-5140
• Main Authors: Mafanya, Madodomzi, Tsele, Philemon, Botai, Joel O., Manyama, Phetole, Chirima, George J., Monate, Thabang
• Format: Article
• Language: English
• Subjects: Accuracy; Agricultural economics; Agricultural land; Agriculture; Aridity; Biodiversity; Calibration; Cameras; Case studies; Confidence intervals; Correlation coefficient; Correlation coefficients; Digital cameras; Empirical analysis; Environmental monitoring; Frameworks; Harrisia pomanensis; Illumination; Image quality; Imagery; Invasive plants; Invasive species; Land cover; Land use; Luminance distribution; Mapping; Mathematical analysis; Precision farming; Procedures; Quantitative analysis; Reflectance; Remote monitoring; Remote sensing; Unmanned aerial vehicles; Woodlands
• ISSN: 0143-1161; 1366-5901
• DOI: 10.1080/01431161.2018.1490503

Orthomosaics derived from consumer grade digital cameras on board unmanned aerial vehicles (UAVs) are increasingly being used for biodiversity monitoring and remote sensing of the environment. To have lasting quantitative value, remotely sensed imagery should be calibrated to physical units of reflectance. Radiometric calibration improves the quality of raw imagery for consistent quantitative analysis and comparison across different calibrated imagery. Moreover, calibrating remotely sensed imagery to units of reflectance improves its usefulness for deriving quantitative biochemical and biophysical metrics. Notwithstanding the existing radiometric calibration procedures for correcting single images, studies on radiometric calibration of UAV-derived orthomosaics remain scarce. In particular, this study presents a cost- and time-efficient radiometric calibration framework for designing calibration targets, checking scene illumination uniformity, converting orthomosaic digital numbers to units of reflectance, and accuracy assessment using in situ mean reflectance measurements (i.e. the average reflectance in a particular waveband). The empirical line method was adopted for the development of radiometric calibration prediction equations using mean reflectance values measured in only one spot within a 97 ha orthomosaic for three wavebands, i.e. red, green and blue of the Sony NEX-7 camera. A scene illumination uniformity check experiment was conducted to establish whether 10 randomly distributed regions within the orthomosaic experienced similar atmospheric and illumination conditions. This methodological framework was tested in a relatively flat terrain semi-arid woodland that is invaded by Harrisia pomanensis (the Midnight Lady). The scene illumination uniformity check results showed that at a 95% confidence interval, the prediction equations developed using mean reflectance values measured from only one spot within the scene can be used to calibrate the entire 97 ha RGB orthomosaic. Furthermore, the radiometric calibration accuracy assessment results showed a correlation coefficient r value of 0.977 (p