Main content area

Computerized Seed and Range Selection Method for Flood Extent Extraction in SAR Image Using Iterative Region Growing

Chakraborty, Arunangshu, Chakraborty, Debasish
Journal of the Indian Society of Remote Sensing 2019 v.47 no.4 pp. 563-571
algorithms, reflectance, remote sensing
This study presents a novel method to capture the flood-affected area in SAR image. It initially locates a pixel in HH-polarized SAR image whose intensity value is equal or close to minimum intensity value in that image. This is adapted since SAR reflectance values of flooded area are less than the other regions due to the water surface smoothness that makes the flood surface a specular reflector with nearly no return to the sensor. Thereafter, the identified seed point is confirmed locally based on two parameters corresponding to intensities and percentage of occurrence of intensities around the seed. Densely populated range around the seed point is computed in the second step. Subsequently in the third step, from the seed point, regions are grown till the intensity value of that point is within the range. These three steps are continued till all flooded regions are captured in SAR image. The algorithm works with minimum human interaction. This method is validated by applying on RADARSAT-2 data and is found that the classification accuracy is 95%, in comparison with “mean shift” and “LPQ”.