Land Cover Classification in a Complex Urban-Rural Landscape with Quickbird Imagery

Area of Science:

• Remote Sensing
• Urban Planning
• Geographic Information Systems (GIS)

Background:

• High spatial resolution imagery is crucial for urban land use/cover classification.
• Traditional per-pixel spectral methods struggle with spectral variability, confusion, and shadows.
• Existing methods often yield suboptimal classification performance in complex urban environments.

Purpose of the Study:

• To explore and enhance urban land cover classification using Quickbird imagery.
• To evaluate the effectiveness of incorporating spatial information in classification algorithms.
• To address limitations of traditional spectral-based methods in urban landscapes.

Main Methods:

• Comparison of traditional per-pixel spectral supervised classification.
• Integration of textural and multispectral images for improved classification.
• Application of spectral-spatial and segmentation-based classification techniques.
• Utilized Quickbird imagery in a developing urban area (Lucas do Rio Verde, Brazil).

Main Results:

• Spatial information significantly improves urban land cover classification performance.
• Integrated textural and spectral data enhanced classification outcomes.
• Segmentation-based classification methods demonstrated superior accuracy.
• Spatial approaches effectively mitigated issues like spectral confusion and shadow effects.

Conclusions:

• Incorporating spatial information is vital for accurate urban land cover classification.
• Spectral-spatial and segmentation-based methods offer substantial improvements over traditional techniques.
• These advanced methods provide more reliable data for urban planning and environmental monitoring.