The effect of input data transformations on object-based image analysis.

Area of Science:

• Remote Sensing
• Geographic Information Systems (GIS)
• Image Analysis

Background:

• Object-based image analysis (OBIA) is crucial for land cover classification.
• Segmentation quality directly impacts the accuracy of OBIA products.
• The influence of spectral transformations on segmentation needs further investigation.

Purpose of the Study:

• To evaluate the impact of spectral transform images on segmentation quality.
• To assess the effect of these transformations on land cover classification products.
• To explore the relationship between segmentation quality and product accuracy.

Main Methods:

• Compared five spectral image transformations against untransformed spectral bands.
• Utilized land cover classification in Accra, Ghana as a case study.
• Assessed segmentation quality and final product accuracy.

Main Results:

• Spectral transformations can enhance the delineation of landscape objects.
• The impact of transformations is specific to the transformation type and object of interest.
• A notable, though idiosyncratic, effect on segmentation quality was observed.

Conclusions:

• Spectral data transformations offer potential benefits for image segmentation in land cover mapping.
• The choice of transformation should be carefully considered based on the specific application and landscape.
• Further research is needed to optimize transformation selection for improved OBIA accuracy.