In this project we used NDWI to detect water bodies from Landsat 8 imagery and then use convectional neural networks to classify the water bodies in to lakes and rivers.
- NDWI with NIR and Green bands is used to detect water bodies.
- NDWI with SWIR1 and Green bands is used to detect water bodies.
- NDWI with SWIR2 and Green bands is used to detect water bodies.
After applying the NDWI threshold, the images are cropped in to sections of 1024x1024 px and segregated in to lake and river segments for training.
After training the model, the model is used to classify a image that has not been previously classified.
and the classification is shown below.
Red color indicates the portion water body is detected as lake and green indicates it is detected as river.