FULLY AUTOMATIC OBJECT-ORIENTED HIERARCHICAL CLASSIFICATION FOR DETAILED LAND COVER UPDATING WITH HIGH SPATIAL RESOLUTION REMOTELY SENSED IMAGERIES

摘要

Within large-scale remote sensing applications, the automatic classification of remotely sensed images is still a challenge. For example, the artificial scheme of sample collection cannot meet the needs of rapid information extraction from remotely sensed images. To overcome some difficulties in automatic classification, this paper introduces the technologies of spatial data mining, including change detection and transfer learning, to the process of sample collection, and further proposes a prior knowledge-based automatic hierarchical classification approach for detailed land cover rapidly updating. A series of high-resolution images are then experimented for validating the effectiveness of the proposed approach in mining detailed land cover information. The results show that the proposed algorithm can automatically collect reliable object samples for object-oriented classification. Furthermore, detailed land cover information can be more excellently extracted by complying hierarchical classifier with the ordinary global classifier, which demonstrates the practicability and effectiveness of the proposed method. The approach creates a novel way for employing the methods of knowledge discovery into the field of remote sensing applications.