an improved approach for dsm generation from high‐resolution satellite imagery

Download An improved approach for DSM generation from high‐resolution satellite imagery

Post on 07-Mar-2017

212 views

Category:

Documents

1 download

Embed Size (px)

TRANSCRIPT

  • This article was downloaded by: [Queensland University of Technology]On: 31 October 2014, At: 09:25Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

    Journal of Spatial SciencePublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/tjss20

    An improved approach for DSMgeneration from highresolution satelliteimageryC. Zhang a & C. S. Fraser ba GIS Center of Excellence , South Dakota State University ,Brookings, SD, 57007, USA E-mail:b Department of Geomatics , University of Melbourne , Melbourne,Vic, 3010, Australia E-mail:Published online: 13 Aug 2010.

    To cite this article: C. Zhang & C. S. Fraser (2009) An improved approach for DSM generationfrom highresolution satellite imagery, Journal of Spatial Science, 54:2, 1-13, DOI:10.1080/14498596.2009.9635175

    To link to this article: http://dx.doi.org/10.1080/14498596.2009.9635175

    PLEASE SCROLL DOWN FOR ARTICLE

    Taylor & Francis makes every effort to ensure the accuracy of all the information (theContent) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

    This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

    http://www.tandfonline.com/loi/tjss20http://www.tandfonline.com/action/showCitFormats?doi=10.1080/14498596.2009.9635175http://dx.doi.org/10.1080/14498596.2009.9635175http://www.tandfonline.com/page/terms-and-conditionshttp://www.tandfonline.com/page/terms-and-conditions

  • SPATIAL SCIENCE 1 Vol. 54, No. 2, December 2009

    An Improved Approach for DSM Generation from High-Resolution Satellite Imagery

    C. Zhang C. S. Fraser

    C. Zhang GIS Center of ExcellenceSouth Dakota State University Brookings, SD 57007, USAchunsun.zhang@sdstate.edu

    C. S. FraserDepartment of GeomaticsUniversity of Melbourne Melbourne, Vic 3010, Australiac.fraser@unimelb.edu.au

    This paper deve lops an improved approach to digital surface model (DSM) generation from high-resolution satellite imagery (HRSI). The approach centres upon an image matching strategy that integrates feature point, grid point and edge matching algorithms within a coarse-to-fine hierarchical process. The starting point is a knowledge of precise sensor orientation, achieved in this case through bias-compensated rational polynomial coefficients (RPCs), and the DSM is sequentially constructed through a combination of the matching results for feature and grid points, and edges at different image pyramid levels. The approach is designed to produce precise, reliable and very dense DSMs which preserve information on surface discontinuit ies. Fol lowing a brief introduction to sensor orientation modelling, the integrated image matching

    algorithms and DSM generation stages are described. The proposed approach is then experimentally tested through the generation of a DSM covering the Hobart area from a stereo pair of IKONOS Geo images. The accuracy of the resulting surface model is assessed using both ground checkpoints and a lidar DSM, with the results indicating that for favourable imagery and land cover, a heighting accuracy of 2 - 4 pixels can be readily achieved. This result validates the feasibility of the developed approach for DSM production from HRSI.

    Keywords: Image matching, IKONOS, high resolution, DSM, accuracy, hierarchical matching strategy

    INTRODUCTIONHigh-resolution satellite imagery (HRSI) is commercially available from a variety of platforms at a range of spatial resolutions, the highest currently being the 41cm ground sample distance (GSD) provided by the recently launched GeoEye-1 satellite. As resolution increases, so does the potential for more accurate and reliable extraction and characterization of finer detail on the earths surface. The ability of HRSI sensors to vary their viewing angle in a single orbit provides the capability of recording stereo or even triple-overlapped images from the

    Dow

    nloa

    ded

    by [

    Que

    ensl

    and

    Uni

    vers

    ity o

    f T

    echn

    olog

    y] a

    t 09:

    25 3

    1 O

    ctob

    er 2

    014

  • Vol. 54, No. 2, December 2009 2 SPATIAL SCIENCE

    same orbital pass. The collected imagery can thus alleviate temporal variability concerns as the minute or so separation between the in-track capture of successive scenes allows consistent imaging conditions, which are then optimal for image matching. The radiometric and geometric characteristics of HRSI make it well suited for digital surface model (DSM) generation (Toutin, 2004; Poon et al., 2005; Krauss et al., 2005; Sohn et al., 2005; Zhang and Gruen, 2006; Poon et al., 2007; Byksalih and Jacobsen, 2008) and feature extraction (Zhang et al., 2005).

    This paper outlines a recent development in the generation of DSMs from HRSI, namely an approach incorporating an improved image matching phase. The method uses a coarse-to-fine hierarchical solution with a combination of several image matching algorithms, thus providing dense, precise and reliable results. Following a brief account of sensor modeling and orientation aspects related to precise image georeferencing, the proposed image matching approach for automatic DSM generation is described. The hierarchical approach employs a processing chain that produces an initial DSM through feature point matching, which is then subsequently refined via grid point and edge matching.

    The improved DSM generation approach has been experimentally assessed using a number of HRSI stereo scenes and in this paper a comprehensive evaluation of the approach applied to IKONOS imagery over the Hobart HRSI test field (Fraser and Hanley, 2003; 2005) is reported. This experimental application over varying terrain with a large height range and comprising city, urban and forest land cover, utilized a lidar derived DSM as a reference surface for accuracy evaluation.

    SENSOR MODELING AND IMAGE ORIENTATION

    High-resolution imaging satellites operate with a linear array scanner where images are obtained with a pushbroom sensor. Thus, the imagery is composed of consecutive scan lines where each line is independently acquired with perspective projection in

    accordance with the collinearity equation incorporating time dependent attitude angles and perspective centre position. To mathematically describe the object-to-image space transformation, the rational polynomial coefficient (RPC) model is now widely accepted and extensively used (Fraser and Hanley, 2003; Grodecki and Dial, 2003; Jacobsen, 2003; Baltsavias et al., 2005). RPCs are derived from the satellites orbital position and orientation data via a rigorous sensor model (Grodecki and Dial, 2003). The general form of the RPC model is given as:

    where

    and ln and sn are the normalized line and sample coordinates, LS and L0 the line scale and line offset, SS and S0 the sample scale and sample offset, and U, V and W the normalized latitude, longitude and ellipsoidal height.

    B e cau se R PCs are d e r ive d f ro m orientation data originating from the satellite ephemeris and star tracker observations, without reference to ground control points (GCPs), geopositioning biases are inherently present. These biases can be accounted for by introducing additional parameters into the original RPC model (Fraser and Hanley, 2003; Fraser et al., 2006). The general bias compensated model is then given as:

    Here, F (l,s) and G (l,s) are functions to model bias. Theoretically, any order of function can be used, however while higher order functions require more GCPs, they do not necessarily increase geopositioning

    00 ,n S n Sl l L L s s S S (1)

    3

    0

    3

    0

    3

    0i j k

    kjiijk WVUaNum ,

    3

    0

    3

    0

    3

    0i j k

    kjiijk WVUbDen

    (3)

    0

    0

    ),,(),,(

    ),(

    ),,(),,(),(

    SSWVUDenWVUNum

    slGs

    LLWVUDenWVUNumslFl

    SS

    S

    SL

    L

    (4)

    ( , , )( , , ) ,( , , ) ( , , )

    SLn n

    L S

    Num U V WNum U V Wl sDen U V W Den U V W

    (2)

    Dow

    nloa

    ded

    by [

    Que

    ensl

    and

    Uni

    vers

    ity o

    f T

    echn

    olog

    y] a

    t 09:

    25 3

    1 O

    ctob

    er 2

    014

  • SPATIAL SCIENCE 3 Vol. 54, No. 2, December 2009

    accuracy (Fraser et al ., 2006). In the implementation adopted for the present study, an affine distortion model introduced in Fraser and Hanley (2003) has been adopted.

    After the bias compensation process, bias-corrected RPCs can be generated and these allow bias-free application of RPC positioning without the need to ref