You are here

Classifying agricultural land into crop types using satellite images

Looking at satellite images of an area and being able to tell what crop is being grown, will be a great resource to improve agricultural productivity. Though satellite images show crop growth from planting to harvest and abnormalities, using them to identify crop type accurately is a challenging problem.

The technique of remote sensing, which analyses images taken by satellites provides information about the land in the most efficient manner. Depending on what is on a land cover region, namely, crops, forests, and buildup, depends on light reflected on the Earth surface. By light, we mean the entire spectrum, including what we see with our eyes. Satellite images are composites of the different light spectra reflected by the earth.

The images are accurate enough to characterise agricultural fields. Fields reflect light differently based on the stage of the cropping cycle, and even abnormalities. Comparing images with trips on the ground, called “ground reference information”, can locate problem areas for better management. This information can be used to improve crop production and plan future seasons based on available information.

Crop classification, or distinguishing different types of crops based on satellite images, has been a challenging problem. A huge variety of crops are grown; also, each crop is grown using different methods, even within one area.

This study uses a nature inspired algorithm called “Artificial Immune System”, which is inspired by how our body’s immune system responds to a disease. For instance, the immune system has a memory, which “remembers” previous infections. The cells involved in immune response also improve their response over time to foreign bodies.

Using these principles, the algorithm can cluster satellite images and use them for crop classification. The researchers have tested out the algorithm on satellite image of Mysore district. This proposed method is more accurate than the methods available in the literature.

About the authors:

Lead author J. Senthilnath is a Research Associate at the Department of Aerospace Engineering, IISc. Website: https://sites.google.com/site/jsenthilnath/. Collaborators Nitin Karnwal and Sai Teja are from National Institute of Technology Trichy and National Institute of Technology Surathkal respectively.