Datenbestand vom 25. Mai 2020
Tel: 089 / 66060798
Mo - Fr, 9 - 12 Uhr
Fax: 089 / 66060799
aktualisiert am 25. Mai 2020
978-3-8439-1068-2, Reihe Bauingenieurwesen
Diego G. Loyola Rodriguez
Methodologies for solving Satellite Remote Sensing Problems using Neuro Computing Techniques
236 Seiten, Dissertation Technische Universität München (2013), Hardcover, B5
The aim of this dissertation is to develop the theoretical ground and tools for retrieving geophysical information from satellite remote sensing data using computational intelligence techniques. The techniques for combining remote sensing and neuro computing, as well as the impressive success of the proposed approaches in demanding and complex Earth observation areas are presented in this work.
While there is an overwhelming number of publications on artificial neural network theory, algorithms, architectures and application; the usage of traceable and repeatable neuro computing practices is almost not found in the current literature. The first new contribution of this work is a novel process model for developing neuro computing systems using either single or combined artificial neural networks. The proposed model is based on state-of-the-art software engineering techniques and it follows an incremental and iterative approach that supports the entire lifecycle of neuro computing systems from requirements gathering, analysis and design, implementation, testing to deployment and maintenance.
The main new contribution of this work is the development of a general methodology for using neuro computing techniques to solve forward and inverse modeling problems in satellite remote sensing. As demonstrated in this dissertation, neuro computing is a powerful tool for extracting information from satellite remote sensing data. The resulting neuro systems provide accurate results, are extremely fast, robust and reliable, and therefore very well suitable for operational environments including (near) real time applications. This methodology is built in the satellite data processor systems for the GOME and GOME-2 sensors that deliver near-real-time and off-line operational products of total ozone concentrations and clouds properties, see http://atmos.eoc.dlr.de/gome.
A new generation of Earth observing satellites will be launched in the near future, carrying instruments that will gather data for studying the Earth with an increase in accuracy, temporal and spatial resolutions. The data volume will raise by two orders of magnitude resulting in a challenge for their timely processing. The neuro computing techniques presented in this work allow not only a very accurate retrieval of geophysical information from current and future satellite sensors but it also provides results in an extremely fast manner. This will help to reach the science and technology goals of improving the prediction of weather, monitoring climate changes and air quality, and other applications with high societal impact.