Development of AI Enabled Solution for Efficient Feature Enrichment from Multiple Data Sources: An Application in Precision Agriculture

The world is surrounded by enormous amount of data almost generated at an unpredictable velocity, in huge volume, variety and veracity. Our traditional systems, however have not yet reached a state where they can efficiently use every bit of this Big data and incorporate it to derive consistent information. Data Fusion is one such technique which can help us achieve this goal. It can be applied to various fields, however for the scope of this paper we have focused on its implementation in Precision Agriculture. While remote sensing has played a major role in precision agriculture by harnessing satellite data as a non-destructive way of information retrieval. The data collected from these satellite varies widely depending on the technique, spatial resolution, temporal resolution, spectral range, viewing geometry of the sensors, thus providing us different amounts of information for different use cases, some in which makes it quiet challenging for the researchers to harness all the information available to attain higher levels of precision, as high as to be able to classify at a sub pixel level while retaining the efficiency and feasibility of the solution. one of the major pain point in agriculture is monitoring large fields and gauging their crop density per square meter. While crop density of a field depends hugely of the soil quality, moisture, fertilizer percentage knowing crop density can have a great impact on yield prediction, sustainable fertilization and overall better through put of a field. Thus, in this paper we have explored the possibility of fusing data from different sensors using CNN based Data Fusion Algorithm to retrieve crop density and segregate patches of field as sparse or dense respectively. The results are quite encouraging.