Assessing the Efficacy of Pixel-based and Object-based Classification Techniques and Classifiers for Land Cover Mapping Using Landsat-8 and Sentinel-2 Data in Complex Mountainous Terrain

Mati ur Rehman, Syed Aun Abbas, Abdul Wahab Shah, Raja Tashfeen Muqarrab, Dur E Najaf Raza, Sawaid Abbas
Additional contact information
Mati ur Rehman, Syed Aun Abbas, Abdul Wahab Shah, Raja Tashfeen Muqarrab, Dur E Najaf Raza, Sawaid Abbas: Smart Sensing for Climate and Development, Center for Geographical Information System, University of the Punjab, Lahore, Pakistan. Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University, Hong Kong SAR

International Journal of Innovations in Science & Technology, 2025, vol. 7, issue 9, 42-56

Abstract: Disaster mitigation and climate-resilient planning heavily depend on accurate Land Use and Land Cover (LULC) datasets. Well-classified LULC data optimizes hazard modeling, surface runoff estimation, and sustainable land use planning, enabling informed decision-making and proactive risk reduction. However, supervised LULC classification faces challenges such as selecting optimal Machine Learning (ML) algorithms, differences in spatial and spectral resolution, and seasonal variability. This study adopts a multi-tiered approach to generate effective LULC maps for Gilgit District, Pakistan, by comparing pixel-based classification and object-based image analysis (OBIA) methods. Pixel-based classification was performed on Google Earth Engine (GEE) using Landsat-8 and Sentinel-2 imagery, applying three classifiers: Random Forest (RF), Support Vector Machine (SVM), and k-Nearest Neighbor (k-NN). OBIA involved multi-resolution segmentation, followed by training and classification on image objects using the same algorithms. Validation using independent samples revealed that object-based maps were visually smoother and more realistic. Quantitatively, pixel-based RF yielded the highest accuracy: 82.9% for Landsat-8 and 78.02% for Sentinel-2. In contrast, OBIA k-NN achieved superior accuracy: 81.3% on Landsat-8 and 83.6% on Sentinel-2. Remaining classifiers also provided nearby results in both classification methods. Lower accuracy in Sentinel-2 may be due to within-class spectral variability at 10m spatial resolution, while Landsat-8’s lower resolution (30m) reduced object-based segmentation performance, resulting in object heterogeneity and misclassification. Although pixel-based classification provided promising results, OBIA ultimately demonstrated superior overall accuracy. This study highlights the importance of resolution-context compatibility and algorithm choice in enhancing LULC classification, which is essential for reliable climate-responsive planning, disaster preparedness, and sustainable development

Keywords: LULC; Machine Learning; OBIA vs Pixel-Based; Gilgit; GEE (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journal.50sea.com/index.php/IJIST/article/view/1482/2056 (application/pdf)
https://journal.50sea.com/index.php/IJIST/article/view/1482 (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:abq:ijist1:v:7:y:2025:i:9:p:42-56

Access Statistics for this article

International Journal of Innovations in Science & Technology is currently edited by Prof. Dr. Syed Amer Mahmood

More articles in International Journal of Innovations in Science & Technology from 50sea
Bibliographic data for series maintained by Iqra Nazeer ().