Advanced calibration to improve robustness of drone-acquired hyperspectral remote sensing data

Abstract

Airborne hyperspectral remote sensing is becoming increasingly integrated into monitoring and management applications on large spatial scales. The main advantage of hyperspectral remote sensing is the high spectral resolution, which enables classification/separation of objects, which may appear very similar to the human eye. An example of similar objects is growing agricultural crops with/without emerging pest infestations. In agricultural fields, accurate detection of hotspots with emerging pest infestations can be used to optimize precision-agriculture such as spot treatments with pesticides and/or spatio-temporal releases of natural enemies to control pest species. Despite the significant potential associated with the integration of airborne remote sensing into monitoring and management of crops in agriculture, and many other industry sectors, the widespread use of airborne hyperspectral remote sensing data is constrained by a critically important challenge the non-linear dynamics of weather, source-sensor geometry, solar light intensity, and scattering. That is, the sensitivity of hyper-spectral remote sensing data is, on one hand, the clear advantage of this technology, but it is also its limitation unless it is possible to correct the radiometric signals and minimize the effects of varying weather, goniometry, solar illumination, and scattering. Under the assumption of correcting airborne hyperspectral remote sensing data, it is possible to acquire “pure” reflectance profiles from target objects. Such advanced reflectance calibration of airborne remote sensing data to obtain “pure” hyperspectral reflectance will greatly improve both the robustness of input data and the sensitivity of classification algorithms.