Geo-Information Technologies of Object Based Image Analysis (Obia) for Urban Mapping
Questions of Cybersecurity, Modeling and Information Processing in the Modern Socio-Technical Systems. Proceedings of the International Research and Technology Conference. Kursk State University (KSU), Kursk, Russia. Ed. by V. M. Dovgal, Y. N. Bykov, and L. S. Kryzhevich. pp. 109–111.
3 Pages Posted: 10 Jan 2019
Date Written: December 30, 2018
The main purpose of this work is to analyse the urban sprawl and spatial changes in city pattern with a case study of Brussels, Brussels. Current work is aimed at the deriving information from the remote sensed VHR data using a priori knowledge in the Object Based Image Analysis (OBIA) approach. OBIA technology is new and effective tool for urban mapping, as it enables dealing with raster images for detailed and precise cartography. Specific focus of this study is selected urban areas of the city of Brussels, Belgium. The study is performed using panchromatic very high resolution (VHR) image processed in the eCognition software. Application of the a priori knowledge in the OBIA approach towards classification of the satellite imagery for solving problem of the land cover studies is the target aim of this research.
The methodology of current work considered existing works and includes processing and classification of the remote sensing data (satellite very high resolution images) using eCognition software. The operation “Multi-resolution Segmentation” was chosen for image processing, as this is one of the most important image processing tools. During segmentation the image was divided up into large homogeneous regions and isolated shapes into the separate polygons within the study area. This procedure was performed at a different scale factors to adjust local conditions, such as urban structures, contrast factors, topology, etc. The four first layers in the layers legend represent multispectral image, while the fifth layer belongs to the panchromatic image. These two images were processed, in order to benefit from the high spatial and spectral resolution of the images which have different properties. Besides, processing of both of them gave various results: the panchromatic VHR image enabled to achieve very detailed segmentation of the image.
The results are presented by accurate geographic classification of the raster image. The objects are grouped into the separated classes connected with each other according to the hierarchical values of their features. The target objects are detected using existing knowledge which helped to find out what information exists in the objects based on training test areas (TTA).
The classification is done using nearest neighbour principle after the manual defining of the number of sample objects. Alike to the standard pixel-based classification, all spectral bands are also used as input channels in OBIA approach, so that the difference consists not in the data but rather in their methodological processing: while pixel-based classification is based on the classification of each pixel separately, the object-based classification treats together all pixels that belong to one object, which is embedded in eCognition. The method of the multiresolution segmentation procedure using OBIA approach was applied to the image in eCognition software and the image was processed. The classification is based on the segmentation of the whole image into meaningful polygons, according to the fuzzy logic approach and nearest neighbour classifier which is similar to the supervised classification in usual image analysis software. The work proved effectiveness of the object based image analysis approach for satellite data processing in urban mapping.
Keywords: image analysis, city development, urban sprawl, mapping, remote sensing, geospatial planning
JEL Classification: Y92, Q01, Q55, Q56, R11, R52, R33
Suggested Citation: Suggested Citation