Publication
Land use/cover classification using orbital and ancillary data, neural networks and multiresolution segmentation
| Summary: | In this paper, a land use/cover classification methodology of the rural/ urban fringe is presented, by means of the application of a neuronal network, with resource to the multiresolution image segmentation, construction of complex elements through object oriented analysis and integration of not spectral (ancillary) information (to assist). The study area is the municipality of Almada, located in the south bank of Tagus river and corresponding to one of the core regions of the Lisbon Metropolitan Area (Portugal). The developed procedure is based on 4 phases: (i) image multiresolution segmentation strategy for construction of different scales objects that have good similarity with the shape of the land use/cover final objects (polygons); (ii) objects attributes acquisition, mainly, context, texture, spectral information, shape, among others; (iii) acquisition of statistical auxiliary data proceeding from the Geographic Base of Information Referencing (BGRI in Portuguese); (iv) integration of the data different types in a neuronal network for classification and posterior discriminated analysis of the land use/cover spatial units. Data used in this methodological experimentation was a 2004 HRVIR SPOT image, with fusion between the panchromatic (supermode 2,5 meters) and the multispectral bands (10 meters) through a transformation between RGB-IHS-RGB color spaces, which allowed a final spatial resolution of 2,5 meters for all the bands. This resolution was respected in the images gathered from the alphanumeric database associated to the BGRI. |
|---|---|
| Main Authors: | Rocha, Jorge |
| Other Authors: | Tenedório, José António; Encarnação, Sara; Estanqueiro, Rossana |
| Subject: | Land use/cover Neural networks Object oriented multiresolution segmentation |
| Year: | 2007 |
| Country: | Portugal |
| Document type: | book part |
| Access type: | open access |
| Associated institution: | Universidade de Lisboa |
| Language: | English |
| Origin: | Repositório da Universidade de Lisboa |
| Summary: | In this paper, a land use/cover classification methodology of the rural/ urban fringe is presented, by means of the application of a neuronal network, with resource to the multiresolution image segmentation, construction of complex elements through object oriented analysis and integration of not spectral (ancillary) information (to assist). The study area is the municipality of Almada, located in the south bank of Tagus river and corresponding to one of the core regions of the Lisbon Metropolitan Area (Portugal). The developed procedure is based on 4 phases: (i) image multiresolution segmentation strategy for construction of different scales objects that have good similarity with the shape of the land use/cover final objects (polygons); (ii) objects attributes acquisition, mainly, context, texture, spectral information, shape, among others; (iii) acquisition of statistical auxiliary data proceeding from the Geographic Base of Information Referencing (BGRI in Portuguese); (iv) integration of the data different types in a neuronal network for classification and posterior discriminated analysis of the land use/cover spatial units. Data used in this methodological experimentation was a 2004 HRVIR SPOT image, with fusion between the panchromatic (supermode 2,5 meters) and the multispectral bands (10 meters) through a transformation between RGB-IHS-RGB color spaces, which allowed a final spatial resolution of 2,5 meters for all the bands. This resolution was respected in the images gathered from the alphanumeric database associated to the BGRI. |
|---|
Funded activities
No funded projects found.