|
Spectral quantitative fusion is a new direction of hyperspectral remote sensing fusion model and algorithm research in recent years, which is an urgent problem of multi-source hyperspectral remote sensing data processing driven by the application of time series global land cover monitoring. China has two high-resolution hyperspectral satellites in space orbit operating GF5 and ZY1E, and the planed hyperspectral satellites are expected to launch in next few years. Hyperspectral remote sensing data has shown a strong growth trend. High quality quantitative fusion is necessary for hyperspectral remote sensing images of large scale, time series and multiple coverage. Traditional remote sensing fusion models and algorithms mainly solve the problem of panchromatic and multispectral remote sensing image fusion. Hyperspectral fusion is more complex than traditional panchromatic and multispectral fusion. This paper investigates the three key difficulties in hyperspectral reflectance fusion. First, the spectral bands mapping of hyperspectral and multispectral remote sensing is fuzzy, which have great differences in the spectral band range, band center wavelength, spectral resolution, spectral response characteristics between hyperspectral and multispectral; second, the hyperspectral remote sensing image contains multi-dimensional information such as ground spectrum, aerosol, water vapor, etc., which reflect the time-phase characteristics and time-series changes of spatial spectral characteristics in different places. The use of multi-dimensional time-series information in fusion processing is unknown, so the systematic error of reflectivity value cannot be detected and the random error is difficult to control; third, the traditional fusion method generally destroys the physical characteristics of quantitative hyperspectral data. This paper focuses on the key problem mentioned above, and takes the hyperspectral reflectance as the image fusion’s physical quantity, puts forward a quantitative fusion scheme of temporal hyperspectral remote sensing reflectance based on observation confidence, to realize the high spatial resolution’s breakthrough of hyperspectral remote sensing reflectance fusion with the reflectance physical truth value maintained and refined accuracy improvement of temporal hyperspectral remote sensing reflectance. This paper studies a quantitative fusion scheme which could be applied in hyperspectral reflectance refine and improvement for large-scale, discrete-time series hyperspectral remote sensing reflectance data set, and will be helpful for maximizing the quantitative ability of hyperspectral remote sensing data and improve the numerical accuracy of hyperspectral remote sensing reflectivity.
|
