Conflation of road network and geo-referenced image using sparse matching

This paper presents an automatic approach to rectify misalignments between a geo-referenced Very High Resolution (VHR) optical image (raster) and a road database (vector). Due to inconsistent representations of road objects in different data sources, the extraction and validation of the homologous road features are complicated. The proposed Sparse Matching (SM) approach is able to smoothly snap the road features from the vector database to their corresponding road features in the VHR image. This novel conflation approach includes three main steps: linear feature preprocessing; sparse matching; feature transformation. Instead of directly extracting the complete road network from the image, which is still a challenging topic for the image processing community, the linear features as road candidates are extracted using Elastic Circular Mask (ECM) and the existing noises are filtered by means of perceptual factors via Genetic Algorithm (GA). With the sparse matching approach, the correspondence between the road candidates from the image and the road features from the vector database can be maximized in terms of geometric and radiometric characteristics. Finally, we compare the transformation results from two different transformational functions i.e. the piecewise Rubber-Sheeting (RUBS) approach and the Thin Plate Splines (TPS) approach for the matched features. The main contributions of this proposed approach include: 1) A novel sparse matching approach especially for conflation framework; 2) Efficient noise filtering in the results from the ECM detector and the GA approach; 3) Numerical comparison of two popular transformational functions. The proposed method has been tested for variant imagery scenario and over 80 percent correct ratio can be achieved from our experiment, at the same time, the average Root Mean Square (RMS) value decreases from 30 meter to less than 10 meter, which makes it possible to use snake-based algorithm for further process.

[1]  Christian Heipke,et al.  Integration of heterogeneous geospatial data in a federated database , 2007 .

[2]  John F. Canny,et al.  A Computational Approach to Edge Detection , 1986, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[3]  Juan B. Mena,et al.  State of the art on automatic road extraction for GIS update: a novel classification , 2003, Pattern Recognit. Lett..

[4]  Ki-Sang Hong,et al.  Road detection in spaceborne SAR images using a genetic algorithm , 2002, IEEE Trans. Geosci. Remote. Sens..

[5]  Gunilla Borgefors,et al.  Hierarchical Chamfer Matching: A Parametric Edge Matching Algorithm , 1988, IEEE Trans. Pattern Anal. Mach. Intell..

[6]  Christian Heipke,et al.  EMPIRICAL EVALUATION OF AUTOMATICALLY EXTRACTED ROAD AXES , 1998 .

[7]  Meng Zhang,et al.  Methods and Implementations of Road-Network Matching , 2009 .

[8]  Emmanuel P. Baltsavias,et al.  Object extraction and revision by image analysis using existing geodata and knowledge: current status and steps towards operational systems☆ , 2004 .

[9]  James M. Keller,et al.  Automated Geospatial Conflation of Vector Road Maps to High Resolution Imagery , 2009, IEEE Transactions on Image Processing.

[10]  Stefan Hinz,et al.  AUTOMATIC ROAD EXTRACTION IN URBAN SCENES — AND BEYOND , 2004 .

[11]  Jan Flusser,et al.  Image registration methods: a survey , 2003, Image Vis. Comput..

[12]  Li Bai,et al.  Local curvature scale: a new concept of shape description , 2008, SPIE Medical Imaging.

[13]  Xiaqing Wu,et al.  Automatic alignment of large-scale aerial rasters to road-maps , 2007, GIS.

[14]  Anand Rangarajan,et al.  A new algorithm for non-rigid point matching , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).

[15]  Yehoshua Sagiv,et al.  Efficient integration of road maps , 2006, GIS '06.

[16]  Hongsheng Li,et al.  Object matching with a locally affine-invariant constraint , 2010, 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[17]  A. Ardeshir Goshtasby,et al.  A comparative study of transformation functions for nonrigid image registration , 2006, IEEE Transactions on Image Processing.

[18]  Yonghuai Liu,et al.  Developing structural constraints for accurate registration of overlapping range images , 2004, Robotics Auton. Syst..

[19]  D. Fritsch,et al.  INTEGRATION OF VECTOR DATA AND SATELLITE IMAGERY FOR GEOCODING , 1998 .

[20]  Jitendra Malik,et al.  Learning to detect natural image boundaries using local brightness, color, and texture cues , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[21]  Jia Cheng-li.  Automatic Road Extraction from SAR Imagery Based on Genetic Algorithm , 2008 .

[22]  R. N. Desmarais,et al.  Interpolation using surface splines. , 1972 .

[23]  Baba C. Vemuri,et al.  A robust algorithm for point set registration using mixture of Gaussians , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[24]  Craig A. Knoblock,et al.  Automatically Conflating Road Vector Data with Orthoimagery , 2006, GeoInformatica.

[25]  Robert C. Bolles,et al.  Parametric Correspondence and Chamfer Matching: Two New Techniques for Image Matching , 1977, IJCAI.

[26]  Demin Xiong,et al.  Semiautomated matching for network database integration , 2004 .

[27]  Nikos Paragios,et al.  Non-rigid registration using distance functions , 2003, Comput. Vis. Image Underst..