An Advanced Learning Framework for High Dimensional Multi-Sensor Remote Sensing Data

Improve the use of land cover data by developing an advanced framework for robust classification using multi-source datasets: Develop, validate and optimize a generalized multi-kernel, active learning (MKL-AL) pattern recognition framework for multi-source data fusion. Develop both single- and ensemble-classifier versions (MKL-AL and Ensemble-MKL-AL) of the system. Utilize multi-source remotely sensed and in situ data to create land-cover classification and perform accuracy assessment with available labeled data; utilize first results to query new samples that, if inducted into the training of the system, will significantly improve classification performance and accuracy. More »