Invited Talks

Challenges and opportunities for advance remote sensing analysis with Neural Networks

Jiaguo Qi, Michigan State University

Abstract

Advances in space technologies such as satellite remote sensing have made global change studies at such temporal and spatial scales that were not possible in the past. Sensors aboard space satellites are able to acquire images at a range of spatial and temporal resolutions in multispectral channels to capture the electromagnetic properties of the earth system. However, challenges exist in extracting quantitative information from these multi-dimensional images, although some methods have been developed and applied to shown successful. These include efficient computation algorithms to deal with large volume data, better classification methods to achieve the required accuracy, and ways to deal with uncertainties from both imagery data as well as processing methods. Neural networks have also been used to extract information from satellite images but the accuracy and efficiency decrease as uncertainties associated with data pre-processing and data volume increase. This presentation raises the issue of how to take the advantage of new theory development in the neural science as well as computer science to convert massive volume of satellite imagery for global environmental studies.