Texture Analysis for Urban Pattern Recognition Using Fine-resolution Panchromatic Satellite Imagery

Author: Pesaresi, Martino

Source: Geographical and Environmental Modelling, Volume 4, Number 1, 1 May 2000 , pp. 43-63(21)

Publisher: Routledge, part of the Taylor & Francis Group

Buy & download fulltext article:

OR

Price: $54.28 plus tax (Refund Policy)

Abstract:

In the framework of the analysis of urban settlement areas using very fine-resolution satellite imagery, a study is presented concerning the discrimination power of textural features with respect to the different terrain patterns present within an urban region. The textural approach chosen in this study is based on the well-known co-occurrence matrix statistical measures derived from the work of Haralick and others in the 1970s. This approach has already been used in the classification of remotely sensed imagery with significant accuracy improvement when compared with the classical radiometric per-pixel approach. However, some problems still exist in the definitions of key parameters for the calculation of the textural measures, as, for example, the pixel displacement vector or the size of the sliding window. In most cases, a specific set of parameters is adopted by referring to a non-explained trial-and-error experience, to subjective thinking, to a generic affirmation led from other experiences in the literature, or not explained at all. A systematic approach is presented for the empirical estimation of the set of texture parameters that more accurately separate the different terrain patterns. Sixteen different urban and non-urban test patterns have been chosen using IRS-1C satellite imagery of the city of Athens in Greece. This approach involves a very wide range of texture parameter combinations (more than 4500 different combinations) and a non-parametric class-separation index calculation. The discrimination of the terrain pattern present in this urban area is subsequently obtained by statistical multidimensional classification of the more discriminating combination of textural features, obtaining accuracy rates greater than 98%.

Document Type: Research Article

DOI: http://dx.doi.org/10.1080/136159300111360

Publication date: May 1, 2000

Related content

Key

Free Content
Free content
New Content
New content
Open Access Content
Open access content
Subscribed Content
Subscribed content
Free Trial Content
Free trial content

Text size:

A | A | A | A
Share this item with others: These icons link to social bookmarking sites where readers can share and discover new web pages. print icon Print this page