Dvara E-Registry’s Machine Learning Model Allows Observations Through Cloud Cover.
One of the key services offered by Dvara E-Registry is analysing existing agricultural data and turning it into valuable insights in the form of crop advisory for farmers and crop monitoring systems to aid efficient agricultural loans and insurance claim settlements. We do this by using several indices including the Normalized Difference Vegetation Index (NDVI), a widely used index to estimate agricultural activity and monitoring land-use changes. With these indices, land around the world can be studied, making it useful for both targeted field analysis, as well as continental or global-scale vegetation monitoring. Looking at these indices alongside other data streams such as weather data can give further insight into patterns of drought, floods that affect vegetation.
These indices use data from multispectral satellites which offer a wealth of information. However, a key limitation of such satellites is that they do not work when cloud cover is present. Alternate satellites like microwave radar satellites can sense through cloud cover, however the data is not as rich and the data is not easy to interpret.
To mitigate this limitation, Dvara E-Registry has developed a model using synthetic-aperture radar (SAR) bands that are cloud independent. The model is a machine learning based approach that re-constructs the NDVI time-series upon recovery of cloud pixel from SAR data even during cloud cover. This model has been possible through training our existing machine learning model over time series inputs along with other proprietary data.
Through the development of this model, Dvara E-Registry will now be able to provide analytics irrespective of weather conditions. This model will be used for our product KhetScore, a farm score that assesses the performance of land plots from the standpoint of agricultural activity on a near real-time basis. KhetScore is a tech-enabled product that can be used by financial institutions and insurance providers to get real-time data, monitor existing crops and implement innovative methods like loan tranching and underwrite loans to farmers. Products like KhetScore are particularly useful when credit scores are not available, as in the case of new-to-credit farmers.