Interactive learning and probabilistic retrieval in remote sensing image archives

We present a concept of interactive learning and probabilistic retrieval of user-specific cover types in a content-based remote sensing image archive. A cover type is incrementally defined via user-provided positive and negat ive examples. From these examples, we infer probabilities of the Bayesian n etwork that link the user interests to a pre-extracted content index. Due t o the stochastic nature of the cover type definitions, the database system not only retrieves images according to the estimated coverage but also acco rding to the accuracy of that estimation given the current state of learnin g. For the latter, we introduce the concept of separability, We expand on t he steps of Bayesian inference to compute the application-free content inde x using a family of data models, and on the description of the stochastic l ink using hyperparameters. In particular, we focus on the interactive natur e of our approach, which provides instantaneous feedback to the user in the form of an immediate update of the posterior map, and a very fast, approxi mate search in the archive, A java-based demonstrator using the presented c oncept of content-based access to a test archive of Landsat TM, X-SAR, and aerial images are available over the Internet [http://www.vision.ee.ethz.ch /similar to rsia/ClickBayes].