6533b7d9fe1ef96bd126c1a1

RESEARCH PRODUCT

Advances in Kernel Machines for Image Classification and Biophysical Parameter Retrieval

subject

description

Remote sensing data analysis is knowing an unprecedented upswing fostered by the activities of the public and private sectors of geospatial and environmental data analysis. Modern imaging sensors offer the necessary spatial and spectral information to tackle a wide range problems through Earth Observation, such as land cover and use updating, urban dynamics, or vegetation and crop monitoring. In the upcoming years even richer information will be available: more sophisticated hyperspectral sensors with high spectral resolution, multispectral sensors with sub-metric spatial detail or drones that can be deployed in very short time lapses. Besides such opportunities, these new and wealthy information sources also come with a price: the analysts are confronted with data showing large and complex feature characteristics. To deal with these new challenges, kernel methods have emerged as a valid, robust and successful framework. The intrinsic regularization implemented in these methods and their low sensitivity to data dimensionality make them natural candidates to solve current remote sensing problems. The flexibility offered by kernel methods allows us to treat heavily nonlinear tasks with elegant methodologies, while still using linear algebra. In the last decade, kernel methods in general, and support vector machines for classification and Gaussian processes for regression in particular, have become standard tools for geospatial data analysis. In this chapter, we first review the main concepts about kernel methods and their use in remote sensing. Then, we review examples of kernel methods for remote sensing image classification and biophysical parameter retrieval.