|
|
Unmanned aerial systems equipped with multispectral, thermal, and hyperspectral cameras provide unprecedented potential for the high resolution and precision observation of Earth’s surface. Consequently, these platforms and sensors can provide large amounts of information, which can be used to derive highly detailed classification, mapping, monitoring, and modeling products. Because this technology is relatively new and rapidly evolving, the focus on optical (visible + near-infrared light) has largely been qualitative in nature; however, increased demand for quantitative and temporal monitoring applications of optical data requires the adoption of more rigorous data preprocessing and calibration techniques, similar to those applied in space-borne satellite remote sensing. Due to the low-altitude nature of data acquisition (<120 meters above ground level), atmospheric interference is usually kept to a minimum and requires little correction; however, for any meaningful quantitative study, UAS sensors should at a minimum be radiometrically calibrated to allow for comparisons of these data through time. The calibration process accounts for illumination differences and attempts to normalize the resulting multispectral images, which also minimizes atmospheric impact.
The ideal parameter derived from UAS multispectral sensors is surface reflectance, which is a spectral signature unique to the surface materials imaged. Surface reflectance is the ratio between the incoming solar illumination (irradiance) and the outgoing reflected energy (radiance) from the surface (captured by the sensor), which requires knowledge of the incoming solar irradiation. Other factors unique to the small UAS sensors must also be addressed, such as fisheye effect (lens distortion) and view angles (departure from center of view), which can lead to spurious values. To date, there are no precise or thorough studies that address these challenges, except for coarse-resolution agricultural applications. Therefore, calibration procedures are typically based on instructions provided by the sensor’s manufacturer, resulting in inconsistencies between datasets collected by different sensors. To make effective use of high-precision UAS remote-sensing observations suitable for quantitative analyses and long-term repeated observation (temporal monitoring), three major challenges must be addressed:
(1) Accurate and operational geometric correction
(2) Precise radiometric calibration
(3) Characterization of the dynamic range (range from which useable data can be extracted)
