In remote sensing applications, the problem of cloud obscuration is widespread. Characterization of these clouds, however, can help to alleviate their effects on scene feature discrimination. Thus, algorithms useful in distinguishing cloudy from clear pixels, and further, in approximating the temperature, particle characteristics, and altitude of the clouds, can improve the quality of many remote sensing missions. To this end, we present here further developments concerning the Cloud Detection and Identification (CloudDI) algorithm, which was first presented in 1998 at the IEEE Aerospace Conference. We concentrate upon using CloudDI to process spectral data in the short and midwave infrared (S/MWIR), and compare our cloud characterization results using multiple bands (e.g., 49 channels) and reduced spectral resolution cases (e.g. 5-8 bands). First, we describe our studies concerning the detection and typing of multiple cloud layers using ARES (Airborne Remote Earth Sensing) flight data. We show that these multiple cloud layers can, in some cases, be detected using suitable CloudD templates. Particular attention is given to modeling the attenuation of the radiance from low-lying clouds by a thin cover of high-altitude cirrus clouds. We also propose a method for directly determining cloud top temperature from midwave infrared (MWIR) bands, using a modified CloudDI routine that removes the reflected solar component from the measured radiance. This temperature information can then be used to infer cloud altitudes. We present our studies concerning the trade-off between the computation time and information value contained in various implementations of CloudDI. In these experiments, we examine the effects of adding and subtracting spectral points, and of replacing the constrained version of the CloudDI algorithm with a (non-iterative) unconstrained implementation. An important part of the CloudDI method is the calculation of the cloud and terrain "templates". In order to optimize this process, we studied three spectral radiance modeling programs - MODTRAN, MOSART and HYPEX - and we delineate some of the differences we have found between them.
Cloud characterization in the S/MWIR using a linear least squares algorithm
2000 IEEE Aerospace Conference. Proceedings (Cat. No.00TH8484) ; 3 ; 351-364 vol.3
01.01.2000
1676830 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Cloud Characterization in the S/MWIR Using a Linear Least Squares Algorithm
British Library Conference Proceedings | 2000
|Least-Squares Frequency-Acquisition Algorithm
NTRS | 1988
|Identification of Linear Time-Varying Systems Using a Modified Least Squares Algorithm
British Library Conference Proceedings | 2000
|NLSQ (non-linear least squares), Programmbeschreibung
TIBKAT | 1977
|Thrust acceleration estimation using an on-line non-linear recursive least squares algorithm
SAGE Publications | 2008
|