Tree Based Representations For Fast Information Mining From VHR Images

In this paper, we present tree based representations as enabler for fast information mining from Very High Resolution optical images. Nowadays, VHR images are acquired at resolution where man-made structures are visible. These structures are best associated to the homogeneous image components which are generally captured by a segmentation. The segmentation decreases the number of elements to be managed with respect to the number of pixels. In this paper, the image is mapped onto hierarchical segmentations which have the property of not being constrained by the typical scale of a single segmentation. The image components embedded in the hierarchical segmentations are further characterized by shape and spectral features. The hierarchical nature of the segmentations allows for fast characterization and restitution of the image components. Then, machine learning techniques can be employed for managing the features for performing analysis of the image information content. The paper describes a new machine learning technique which exploits the pre-organization of the features into hierarchical clusterings represented as a tree structure. This machine learning technique is proven to handle millions of elements. Experiments are conducted with satellite images to assess the combination of hierarchical segmentations with the tree based learning method. A first experiment is conducted on an hyper spectral image. Then, two experiments are carried out for the detection of built-up from 2 WorldView-2 multi-spectral images. Finally, a last experiment exploits the learning technique to extract the buildings from several WorldView-2 scenes by exploiting a low resolution build-up mask provided by the consistent analysis of MODIS images.

GUEGUEN Lionel;  OUZOUNIS Georgios;  PESARESI Martino;  SOILLE Pierre; 

2013-03-06

Publications Office of the European Union

JRC71681

978-92-79-26419-1,   

1831-9424,   

http://dx.doi.org/10.2788/49465,    https://publications.jrc.ec.europa.eu/repository/handle/JRC71681,   

10.2788/49465,