Brazil - Tocantins

Specific Project Objectives & Deliverables

We used an unsupervised object-based classification approach for LULC mapping.

  • Segmentation

« Multiresolution segmentation » (eCognition Developer):
Image Layers = B, V, R, NIR, MIR (Landsat 8 image at original 30m spatial resolution)

Scale parameter = 110
Shape = 0.8
Compactness = 1
Segmentation extract:

  • Classification

Extraction of the median NDVI value per segmented object from the MODIS time series : 

   -> 

The next steps include the classification of the resultant median values. The classification results are expected to improve with respect to last year’s classification results, in particular for the sugarcane fields with a long cultural cycle (18 moths) which should be differentiated from grassland fields with the new longer time series (24-months).

©2015 Joint Experiment for Crop Assessment and Monitoring
© HER MAJESTY THE QUEEN IN RIGHT OF CANADA SA MAJESTE LA REINE DU CHEF DU CANADA (2015)