Object-based Land Cover Classification and Change Analysis in the Baltimore Metropolitan Area Using Multitemporal...

Area of Science:

• Environmental Science
• Remote Sensing
• Urban Planning

Background:

• Accurate land cover data is vital for urban management, ecosystem monitoring, and planning.
• Urban areas experience dynamic land cover changes requiring effective monitoring techniques.

Purpose of the Study:

• To compare object-based and pixel-based post-classification change detection methods.
• To assess land cover changes in the Gwynns Falls watershed from 1999 to 2004 using high-spatial-resolution aerial imagery.

Main Methods:

• Object-based image classification was applied to aerial imagery from 1999 and 2004.
• Post-classification change detection was performed using both traditional pixel-based and object-based approaches.

Main Results:

• The object-based classification achieved high overall accuracies (92.3% for 1999, 93.7% for 2004).
• The object-based change detection method yielded higher accuracy (90.0%, Kappa 0.854) compared to the pixel-based method (81.3%, Kappa 0.712).

Conclusions:

• Object-based approaches are superior for land cover change detection in urban areas.
• Incorporating spatial information and expert knowledge enhances change detection accuracy.