Multispectral image classification from axiomatic locally finite spaces-based segmentation

  • Jose Valero
  • Ivan Lizarazo

Resumen (es_ES)

El análisis de imagenes basado en objetos geográficos (GEOBIA por su sigla en inglés) comienza generalmente definiendo elementos más gruesos del espacio geométrico u objetos de imagen, agrupando píxeles cercanos con base en grafos (a, b)-conectados como definiciones de vecindario. En este enfoque, sin embargo, pueden no cumplirse algunos axiomas topológicos requeridos para garantizar una correcta representación de las relaciones de conexión. Por lo tanto, la definición convencional de límites de objetos de imagen, presenta ambigüedades debido a que los contornos unidimensionales están representados por píxeles bidimensionales. En este trabajo, la segmentación se lleva a cabo mediante un nuevo enfoque basado en espacios axiomáticos localmente finitos (proporcionados por complejos cartesianos) y sus matroides orientados asociados. Para probar el enfoque propuesto, los segmentos de la imagen basada en ALFS fueron clasificados usando el algoritmo de máquina de soporte vectorial (SVM por su sigla en inglés) usando la respuesta a filtros direccionales como un canal adicional. El enfoque propuesto utiliza un enfoque multiescala para la segmentación, que incluye análisis de textura y de afinidad espectral en la definición de límite. La propuesta se evaluó comparativamente con la representación de píxeles convencionales en un pequeño subconjunto del conjunto de datos de referencia GEOBIA2016. Los resultados muestran que la exactitud de la clasificación se incrementa en comparación con la segmentación convencional de pixeles.

Resumen (en_US)

Geographical object-based image analysis (GEOBIA) usually starts defining coarse geometric space elements, i.e. image-objects, by grouping near pixels based on (a, b)-connected graphs as neighbourhood definitions. In such an approach, however, topological axioms needed to ensure a correct representation of connectedness relationships can not be satisfied. Thus, conventional image-object boundaries definition presents ambiguities because one-dimensional contours are represented by two-dimensional pixels. In this paper, segmentation is conducted using a novel approach based on axiomatic locally finite spaces (provided by Cartesian complexes) and their linked oriented matroids. For the test, the ALFS-based image segments were classified using the support vector machine (SVM) algorithm using directional filter response as an additional channel. The proposed approach uses a multi-scale approach for the segmentation, which includes multi-scale texture and spectral affinity analysis in boundary definition. The proposed approach was evaluated comparatively with conventional pixel representation on a small subset of GEOBIA2016 benchmark dataset. Results show that classification accuracy is increased in comparison to a conventional pixel segmentation.

Descargas

La descarga de datos todavía no está disponible.

Referencias

Arbeláez P., Maire M., Fowlkes C., Malik J.( 2011). Contour Detection and Hierarchical Image Segmentation. IEEE Trans. Pattern Anal. Mach. Intell, vol. 33, no 5, p. 898–916.
Brun L., Domenger J-P, Mokhtari M. (2003). Incremental modifications of segmented image defined by discrete maps. Journal of Visual Communication and Image Representation, vol. 14, no 3, p. 251 – 290, http://www.sciencedirect.com/science/article/pii/ S1047320303000233.
Cramer M. (2010). The DGPF-Test on Digital Airborne Camera Evaluation Overview and Test Design, http://www.ifp.uni-stuttgart.de/dgpf/DKEP-Allg.html.
De Loera J. A., Rambau J., Santos F. (2010). Triangulations: Structures for Algorithms and Applications. First ed. Springer Publishing Company, Incorporated.
Fukuda K. (2004). Lecture notes on oriented matroids and geometric computation. Technical Report. RO-2004.0621, course of Doctoral school in Discrete System Optimization, EPFL 2004.
Grady L. (2012). Targeted image segmentation using graph methods. Department of Image Analytics and Informatics Siemens Corporate Research Princeton, NJ. http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.303.7046&rep=rep1&type=pdf
Kovalevsky V.A. (2005). Algorithms in Digital Geometry Based on Cellular Topology. Proceedings. Springer Berlin Heidelberg, p. 366–393.
Kovalevsky, V.A. (1989). Finite Topology as Applied to Image Analysis. In Computer Vision, Graphics, and Image Processing, vol. 42 no 2, p. 141–161.
Kovalevsky V. A. (2001). Algorithms and Data Structures for Computer Topology, In: Bertrand G., Imiya A., Klette R. (eds) Digital and Image Geometry. Lecture Notes in Computer Science, vol. 2243. Springer, Berlin, Heidelberg, p. 38-58.
Kovalevsky V. A. (2006). Axiomatic Digital Topology. Journal of Mathematical Imaging and Vision, vol. 26, no 1-2, p. 41–58.
Kovalevsky V. A. (1984). Discrete Topology and Contour Definition. Pattern Recognition Letters, vol. 2, no 5, p. 281-288.
Kovalevsky V. A. (2008). Geometry of Locally Finite Spaces. International Journal of Shape Modelling, vol. 14 no 02, p. 231–232.
Leung T. K., Jitendra M. (1996). Detecting, Localizing and Grouping Repeated Scene Elements from an Image. Proceedings of the 4th European Conference on Computer Vision. ECCV ’96, London, UK, vol 1, p. 546–555. Springer-Verlag.
Malik J., Belongie S., Leung T., Shi J. (2001). Contour and Texture Analysis for Image Segmentation. International Journal of Computer Vision, vol. 43, no, p. 7–27.
Oxley J. G. (2006). Matroid Theory (Oxford Graduate Texts in Mathematics). New York, NY, USA. Oxford University Press, Inc
Pont-Tuset J., Arbeláez P., Barron J. T., Marqués F., Malik J. (2015). Multiscale Combinatorial Grouping for Image Segmentation and Object Proposal Generation. CoRR abs/1503.00848. http://arxiv.org/abs/1503.00848.
Tang Pang-Ning, Steinbach M., Kumar V. (2004). Introduction to Data Mining, (First Edition). Boston, MA, USA: Addison-Wesley Longman Publishing Co., Inc.
Tso B., Mather P. (2009). Classification Methods for Remotely Sensed Data. CRC Press, Taylor & Francis Group.
Valero J., Arbeláez P., Lizarazo I. (2017). A Combinatorial Approach for Hyperspectral Image Segmentation. Cham: Springer International Publishing.AG 2017, p. 334–348
Whitney H. (1935). On the abstract properties of linear dependence American Journal of Mathematics, vol. 57, p. 509–533.
Worboys M., Duckham M. (2004). GIS: A Computing Perspective, 2nd Edition. Boca Raton, FL, USA: CRC Press, Inc.
Cómo citar
Valero, J., & Lizarazo, I. (2019). Multispectral image classification from axiomatic locally finite spaces-based segmentation. UD Y La geomática, (13). Recuperado a partir de https://revistas.udistrital.edu.co/index.php/UDGeo/article/view/15230
Sección
Artículo de investigación científica y tecnológica