lntegrated-sensing

INTEGRATED SENSING


The growing wealth of remote sensing data from hundreds of space based sensors is providing us with enormous new opportunities to better understand the earth at a time when that understanding may be critical. Yet designing, building, and launching these sensors entail enormous projects typically taking many years and costing billions of dollars. Unfortunately these missions are often delayed or canceled. 

Even when the sensors are deployed, they may not provide data for all of the earth, they may not measure the precise spectral bands of interest, or they may not visit sites of interest frequently enough. Even when remote sensing projects are successful, an instrument providing data may fail, or lack funding for continuing operation and be thus be retired. This can result in gaps in critical data records. It is also worth noting that data provided by new sensors is often of greatest use if there was a way to compare current measurements historically, but an accurate reconstruction of historical data is required ssince those measurements do not exist.

Fortunately growing computational power and statistical learning techniques provide a means to leverage the hundreds of satellites already observing the earth, which produce multi-spectral, hyper-spectral or ultra-spectral images. Through data fusion from several sensors, it is possible to produce virtual integrated sensors. These integrated sensors can help address some of this issues that cannot be addressed with hardware sensors alone. 


Projects:


GOES-R_XYZ  

Estimating True Color imagery for GOES-R

Our group has developed a prototype of an algorithm for producing true color bands for the future GOES-R satellite series, despite the fact that it will only have two visible color bands. The algorithm works by integrating data from other bands, and in training, from other satellites, to build a statistical estimator to determine the true color values. This will leverage current data acquisition to create a new virtual data acquisition product. It will also make it easier for users to use GOES-R visible imagery by providing the imagery in color, which is an easier to visualize format.



QIR  

A Multi-band statistical restoration of the Aqua MODIS 1.6 micron band

Another example of this project is currently funded work for an algorithm to reconstruct level-one radiances that may be missing, for example, due to dead detectors on the satellite. Again by integrating data from other bands and possibly other satellites, it is possible to reconstruct the missing data. More broadly, using statistical techniques, we can address a wide range of problems to estimate remotely sensed imagery virtually, by integrating data from statistically related measurements.



Error ProPagation  

Error Mitigation for CCSD Compressed Imager data

To efficiently use the limited bandwidth available on the downlink from satellite to ground station, imager data is usually compressed before transmission. Transmission introduces unavoidable errors, which are only partially removed by forward error correction and packetization. In the case of the commonly used CCSD Rice-based compression, it results in a contiguous sequence of dummy values along scan lines in a band of the imager data.



© 2010, NOAA-CREST CSDIRS, CCNY-Glasslab/Michael Grossberg/Irina Gladkova/Paul Alabi

Powered by LFC — Lightning Fast CMS

All rights reserved