Comparing, validating and improving the performance of reflectance obtention method for UAV-Remote sensing

Abstract

Recently, remote sensing based on the unmanned aerial vehicle (UAV) has become a novel remote sensing technology for researching the characteristics of the earth and the properties of the objects near the earth's surface. Accurate reflectance products are fundamental for optical remote sensing applications. Although the application of remote sensing based on UAV is widely applied, the quality, standards, strategies, and procedures of remote sensing products are faced challenges. To obtain convincing reflectance products, the methods, processes, and strategies obtaining the reflectance products for UAV-remote sensing are validated, compared, and improved. Meanwhile, three methods (ELM, MIR, and ARTM) are exploited to obtain spectral reflectance for hyperspectral images based on UAV in this paper. The comparison with the spectral reflectance of verified targets measured in the field indicates that the spectral reflectance of MIR has higher accuracy than others, with the average spectral absolute error (ASAE) of ±2.5%. By contrast, the ASAE of ELM and ARTM is ±7.0% and ±9.6%, respectively. Besides, based on the principle of radiative transfer, the radiation factors including incident radiation, background radiance, and environmental radiance are analyzed for improving the performance of the three methods. By recalibrating the hyperspectral sensor, the average absolute error of MIR is reduced to ±1.8%. By optimizing the parameters, the performance of ELM and ARTM is improved significantly, and the average absolute error of the two methods is reduced to ±1.7% and ±4.4%, respectively.