Semi-automatic choice of scale-dependent features for satellite SAR image classification

ArticleinPattern Recognition Letters 27(4):244-251 · March 2006with7 Reads
Impact Factor: 1.55 · DOI: 10.1016/j.patrec.2005.08.005 · Source: DBLP
Abstract

In this work we compare two different approaches to the use of multiple scales in the classification process of satellite SAR images. These are (I) the multi-scale co-occurrence texture analysis and (II) the semivariogram approach. Moreover, we propose a scheme for optimizing the co-occurrence window size and the semivariogram lag distances in terms of classification accuracy performance. To improve the results even further, we introduce a methodology to compute the co-occurrence features with a window consistent with the local scale, provided by the semivariogram analysis.Examples of satellite SAR image segmentation for urban area characterization are shown to validate the procedure.

    • "...ness of the GLCM technique for characterizing SAR images has been confirmed. Dell'Acqua et al. [26,27] introduce the GLCM to discriminate the urban areas from the background in SAR images with medium, a..."
      The potential usefulness of the GLCM technique for characterizing SAR images has been confirmed. Dell'Acqua et al. [26,27] introduce the GLCM to discriminate the urban areas from the background in SAR images with medium, as well as high, resolutions. When the GLCM is used, many important parameters need to be considered.
    [Show abstract] [Hide abstract] ABSTRACT: Interests in synthetic aperture radar (SAR) data analysis is driven by the constantly increased spatial resolutions of the acquired images, where the geometries of scene objects can be better defined than in lower resolution data. This paper addresses the problem of the built-up areas extraction in high-resolution (HR) SAR images, which can provide a wealth of information to characterize urban environments. Strong backscattering behavior is one of the distinct characteristics of built-up areas in a SAR image. However, in practical applications, only a small portion of pixels characterizing the built-up areas appears bright. Thus, specific texture measures should be considered for identifying these areas. This paper presents a novel texture measure by combining the proposed labeled co-occurrence matrix technique with the specific spatial variability structure of the considered land-cover type in the fuzzy set theory. The spatial variability is analyzed by means of variogram, which reflects the spatial correlation or non-similarity associated with a particular terrain surface. The derived parameters from the variograms are used to establish fuzzy functions to characterize the built-up class and non built-up class, separately. The proposed technique was tested on TerraSAR-X images acquired of Nanjing (China) and Barcelona (Spain), and on a COSMO-SkyMed image acquired of Hangzhou (China). The obtained classification accuracies point out the effectiveness of the proposed technique in identifying and detecting built-up areas.
    Preview · Article · May 2014 · Remote Sensing
    0Comments 3Citations
    • "...lity of single pixel values because of the influence of the pixel neighborhood on the signal [80], [81], and the missing ability to infer objects of interest corresponding to the visual perception of hu-..."
      This is particularly true for pixel-based land cover classifications in urban areas, which are well-known for their spectral and spatial heterogeneity [33], [72]–[75]. Other drawbacks of traditional approaches are their sensitivity to registration errors between multi-modal inputs [64], [76], [77], the computional effort that comes along with a per-pixel analysis of HSR data [78], [79], the unreliability of single pixel values because of the influence of the pixel neighborhood on the signal [80], [81], and the missing ability to infer objects of interest corresponding to the visual perception of hu- mans [82]. Furthermore, spatial information, like image texture or morphological measures [83], [84], are usually calculated using a moving window, which causes borders between land cover classes to become blurred [85].
    [Show abstract] [Hide abstract] ABSTRACT: This paper focuses on the description and demonstration of a simple, but effective object-based image analysis (OBIA) approach to extract urban land cover information from high spatial resolution (HSR) multi-spectral and light detection and ranging (LiDAR) data. Particular emphasis is put on the evaluation of the proposed method with regard to its generalization capabilities across varying situations. For this purpose, the experimental setup of this work includes three urban study areas featuring different physical structures, four sets of HSR optical and LiDAR input data, as well as statistical measures to enable the assessment of classification accuracies and methodological transferability. The results of this study highlight the great potential of the developed approach for accurate, robust and large-area mapping of urban environments. User's and producer's accuracies observed for all maps are almost consistently above 80%, in many cases even above 90%. Only few larger class-specific errors occur mainly due to the simple assumptions on which the method is based. The presented feature extraction workflow can therefore be used as a template or starting point in the framework of future urban land cover mapping efforts.
    Full-text · Article · Jul 2013 · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
    0Comments 15Citations
    • "... image is an important problem with many applications in computer vision and image processing [21, 10, 27, 30]. Further, a strong link between scale-space/scale-selection theory and biological vision ha..."
      Of particular interest in this work is the detection of tube-like structures through the analysis of the Hessian matrix at multiple scales (section 2.2). Selecting the appropriate scales for the various features in an image is an important problem with many applications in computer vision and image processing [21, 10, 27, 30]. Further, a strong link between scale-space/scale-selection theory and biological vision has been demonstrated in many earlier works [8, 9].
    [Show abstract] [Hide abstract] ABSTRACT: Many feature detection algorithms rely on the choice of scale. In this paper, we complement standard scale-selection algorithms with spatial regularization. To this end, we formulate scale-selection as a graph labeling problem and employ Markov random field multi-label optimization. We focus on detecting the scales of vascular structures in medical images. We compare the detected vessel scales using our method to those obtained using the selection approach of the well-known vesselness filter (Frangi et al 1998). We propose and discuss two different approaches for evaluating the goodness of scale-selection. Our results on 40 images from the Digital Retinal Images for Vessel Extraction (DRIVE) database show an average reduction in these error measurements by more than 15%.
    Full-text · Conference Paper · Jul 2010
    0Comments 8Citations
    • "...every day more the leaning is to process images with a multi-scale or hierarchical technique [17], [11], that offer as result a segmentation of the image for each scale or level, respectively. Multi-scal..."
      As an example, given an aerial image, we could distinguish the sea, the land and the forest areas (in a low detailed analysis), but also (in a high detailed study) we could find distinct deep areas in the sea or different types of forest, or even (in a higher detail level) look for marks in the sea with concrete characteristics that could represent oil dumping. This is why every day more the leaning is to process images with a multi-scale or hierarchical technique [17], [11], that offer as result a segmentation of the image for each scale or level, respectively. Multi-scale techniques are mainly based on study the evolution of the contours through different scales, by blurring the image more in each scale and then performing a new segmentation.
    [Show abstract] [Hide abstract] ABSTRACT: In image segmentation it is well known that a given image can be analyzed with different detail levels, this is why some hierarchical approaches have been proposed to give a different segmentation for each detail level. Most of these proposals are specially designed for precise and well defined regions. However regions usually have blurred contours, soft color shades, and brightness that give rise to the problem of the imprecision in the regions. In this paper we face both problems considering the imprecision of the regions at the definition of the criteria to obtain a hierarchy detail levels. Concretely, we propose to calculate a similarity relation between fuzzy regions, based on two measures that take into account the imprecision in the transition between the regions, as well as the likeness of their characteristics. Then we use this fuzzy similarity relation to obtain a nested hierarchy of fuzzy segmentations by means of its alpha-cuts. In this way we obtain a tool to easily change the detail level and obtain a new fuzzy segmentation of the image, just changing the value of alpha.
    Full-text · Conference Paper · Jul 2008
    0Comments 1Citation
    • "...of classification accuracy performance, as effective as a long exhaustive search of the best scale [22]. The textural bands of Heterogeneity using different window size (3*3, 7*7, 11*11) are shown in As ..."
      Dell' Acqua, F. et al. extracted texture feature by using co-occurrence matrix and semi-variogram analysis for mapping urban density classes in satellite SAR data. The results show that the joint use of co-occurrence texture features and semi-variogram analysis to optimize co-occurrence window size can be, in terms of classification accuracy performance, as effective as a long exhaustive search of the best scale [22]. The textural bands of Heterogeneity using different window size (3*3, 7*7, 11*11) are shown in As shown in theFig.
    [Show abstract] [Hide abstract] ABSTRACT: The areas of the land consolidation projects are generally small, so the remote sensing images used in land-cover classification for the land consolidation are generally high spatial resolution images. The spectral complexity of land consolidation objects results in specific limitation using pixel-based analysis for land cover classification such as farmland, woodland, and water. Considering this problem, two approaches are compared in this study. One is the fixed window size co-occurrence texture extraction, and another is the changeable window size according to the result of semi-variogram analysis. Moreover, the methodology for optimizing the co-occurrence window size in terms of classification accuracy performance is introduced in this study. Zhaoquanying land consolidation project is selected as an example, which located in Shunyi District, Beijing, China; texture feature is extracted from SPOT5 remote sensing data in the TitanImage development environment and involved in classification. Accuracy assessment result shows that the classification accuracy has been improved effectively using the method introduced in this paper.
    Full-text · Article · Jul 2008 · WSEAS Transactions on Computers
    Yan Huang Yan Huang Anzhi Yue Anzhi Yue Su Wei Su Wei +3 more authors... Daoliang Li Daoliang Li
    0Comments 4Citations
    • "...ay be obtained by the same SAR image, using for instance the textural feature analysis proposed in [21]. A more efficient way would be however to use ancillary data. ..."
      Damage maps will be computed for the whole data set, but they will be significant in urban areas only. This requires a urban/surrounding discrimination by means of a preliminary step, which may be obtained by the same SAR image, using for instance the textural feature analysis proposed in [21]. A more efficient way would be however to use ancillary data.
    [Show abstract] [Hide abstract] ABSTRACT: In this paper, the problem of rapid earthquake damage detection in urban areas using multitemporal synthetic aperture radar data is addressed. It is shown that the combination of intensity and phase features enhances the damage pattern extracted from the data temporal stack using a spatially aware classifier. Moreover, the use of ancillary data, easily available for urban areas, further improves the accuracy by discarding uninteresting parts of the scene and forcing homogeneous classification within city blocks to avoid "class-blurring" effects consequential to the window-based computation of relevant measures. The procedure is validated based on results for the town of Bam, Iran, and compared with ground-based survey maps
    Full-text · Article · Jul 2007 · IEEE Transactions on Geoscience and Remote Sensing
    0Comments 79Citations
Show more