Object Based Land Cover Extraction Using Open Source Software


  • Abhasha Joshi Instructor, Land Management Training Center, Dhulikhel Kavre
  • Janak Raj Joshi Director, Land Management Training Center, Dhulikhel Kavre
  • Nawaraj Shrestha Lecturer, Kathmandu University
  • Saroj Shrestha MTECH Student, University of Technology, Sydney
  • Sudarshan Gautam Training Chief, Nepalgunj Technical School, Nepalgunj




Land Cover, OBIA' Segmentation, Spring


Land cover is observed bio-physical cover of the earth’s surface and is an important resource for global monitoring studies, resource management, and planning activities. Traditionally these land resources were obtained from imagery using pixel based image analysis. But with the advent of High resolution satellite imagery and computation techniques these data are now widely being prepared using Object based Image Analysis (OBIA) techniques. But mostly only algorithm provided in commercial software and Ecognition in particular is being used to study OBIA.

This paper aims to assess the application of an open source software Spring for OBIA. In this Study 0.5 meter pan sharpened Geo-Eye image was classified using spring software. The image was first segmented using region growing algorithm with similarity and area parameter. Using hit and trail method best parameter for segmentation for the study area was found. These objects were subsequently classified using Bhattacharya Distance. In this classification method spectral derivatives of the segment such as mean, median, standard deviation etc. were used which make this method useful. However the shape, size and context of the segment can’t be accounted during classification. i.e. rule based classification is not possible in spring.

This classification method provides satisfactory overall accuracy of 78.46% with kappa coefficient 0.74. This classification method gave smooth land cover classes without salt and pepper effect and good appearance of land cover classes. However image segmentation and classification based on additional parameters such as shape and size of the segment, contextual information, pixel topology etc may give better classification result.

Nepalese Journal on Geoinformatics -12, 2070 (2013AD): 26-30


Download data is not yet available.




How to Cite

Joshi, A., Joshi, J. R., Shrestha, N., Shrestha, S., & Gautam, S. (2013). Object Based Land Cover Extraction Using Open Source Software. Journal on Geoinformatics, Nepal, 12, 26–30. https://doi.org/10.3126/njg.v12i0.9070