Date of Degree
Robert M. Haralick
Artificial Intelligence and Robotics | Atmospheric Sciences | Environmental Monitoring | Numerical Analysis and Scientific Computing
MODIS, VIIRS, GOES-R, GOES-16, ACSPO, resampling
Scientists from all over the world make use of remotely sensed data from hundreds of satellites to better understand the Earth. However, physical measurements from an instrument is sometimes missing either because the instrument hasn't been launched yet or the design of the instrument omitted a particular spectral band. Measurements received from the instrument may also be corrupt due to malfunction in the detectors on the instrument. Fortunately, there are machine learning techniques to estimate the missing or corrupt data. Using these techniques we can make use of the available data to its full potential.
We present work on four different problems where the use of machine learning techniques helps to extract more information from available data. We demonstrate how missing or corrupt spectral measurements from a sensor can be accurately interpolated from existing spectral observations. Sometimes this requires data fusion from multiple sensors at different spatial and spectral resolution. The reconstructed measurements can then be used to develop products useful to scientists, such as cloud-top pressure, or produce true color imagery for visualization. Additionally, segmentation and image processing techniques can help solve classification problems important for ocean studies, such as the detection of clear-sky over ocean for a sea surface temperature product. In each case, we provide detailed analysis of the problem and empirical evidence that these problems can be solved effectively using machine learning techniques.
Shahriar, Fazlul, "Machine Learning Approach to Retrieving Physical Variables from Remotely Sensed Data" (2017). CUNY Academic Works.
Artificial Intelligence and Robotics Commons, Atmospheric Sciences Commons, Environmental Monitoring Commons, Numerical Analysis and Scientific Computing Commons