Fine-grain Cluster Estimation of Land Cover Classes using Landsat 8 Multispectral images

What is it about?

Earth observation satellites provide us with ample amount of raw data for land cover analysis. However, annotating these data is a cumbersome process, subjected to human error which compelus to shift from supervised to unsupervised techniques. Although clustering methods are being widely used for the past few years in the field of remote sensing, identifying the number of fine-grain classes present in a region remain a challenging problem. Therefore, we propose a rule-based and neural-network learning technique that can divide the pixels into three standard classes, water bodies, vegetation and vegetation-void. These classes are easier to identify and is not region-specific. Later we apply fine-grain clustering on each of these classes to segregate them into finer groups. Our clustering method identifies the appropriate number of fine-grain classes present in a specific region.

Featured Image

Photo by SpaceX on Unsplash

Photo by SpaceX on Unsplash

Why is it important?

It handles the problem of labeling of data.

Perspectives