Anais do XIX Simpósio Brasileiro de Sensoriamento Remoto

The remote sensing of night-time lights (NTL) has been growing since the last decade. Recently, the watershed-based partition (WBP) approach was presented as a useful method for identifying patterns of human settlements based on NTL in a novel scale of analysis. In this paper we tested several methodological procedures and then presented the pre-processing techniques needed for implementing the WBP. We submitted the original VIIRS-Day/Night Band data to a series of operations to emulate a topography entry for the segmentation algorithm for detection of watersheds. The processed data were submitted through a gradual exponentiation until the parameters of the segments were stabilized. From the third power on, the changes of the parameters remain above 10%. From the fifth to the sixth power, the changes represent less than 0.6%. The results of the segmentation indicate that the exponentiation can enhance the segmentation process of NTL through the WBP approach.