OUR APPROACH
Soon after interacting with some of our clients we have realized that capturing data from field is a costly affair and also challenging. Wherever possible we extract datasets from satellite and improve operational efficiencies of our partners. This helps our partners to re-engineer their exisitng processes and evolve into hybrid models involving better location information at regular intervals.
OUR STACK
Satellite data
We help our partners to harvest spatio-temporal information coming out of rich satellite data resources that are available in public domain. We serve organizations working in the domains of agriculture, water and eco-restoration.
Data Modelling
Multi source data streams are used to created data models. These models are designed to understand correlation between related data sets. These correlation matrices are converted to decision trees to be consumed by our classifiers.
Classifiers
We build custom rule based classifiers which use the correlations generated from the data model. By automation of the classification process of the imagery, in the first iteration we generate binary classification and a set of binaries are used to derive multiple classes.
Feature Extraction
Classified outputs are given to connected component algorithms to generate features. These features fit into any standard compliant feature model of OGC. This process eliminates manual digitization and thus improves efficiency by reducing turn around time.
Decision Trees
Based on application either regression models or decision trees are used to understand changes in the operating environment. This helps in creating information streams depicting the underlying environment through maps. These maps are posted as spatial databases.
Web Publishing
Maps are pulled from spatial databases and are posted as web services which are standard compliant. Our stack is capable of handling both rasters and vectors as xml based or json based web services. This enables us to plugin to any existing information model.