SBIR Phase I: Using Aerial Imagery Analysis to Manage Stress in Coffee Production

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is significant. Based on the company's experience with the willingness of farmers to pay for such services, the potential commercial market for the envisioned solution for coffee alone would amount to USD 1 billion, based on the estimated 27 million acres currently in production around the world. The commercial potential for additional specialty crops would exceed several times this amount. The company's agronomy team believes that the proposed solution would result in yield improvements of the order of 5% to 10% and a reduction in combined fertilizer and water costs in the range of 3% to 8%. Moreover, the solution would result in higher-quality coffee beans which would improve the commercial viability of small coffee farms in emerging economies. This would have the potential of lifting hundreds of thousands of sub-tropical farmers and their families out of poverty while sustainably protecting soil and water resources. Precision agriculture techniques developed in this project will also enable coffee producers to address the effects of climate change by identifying destructive patterns and prescribing effective ways to maintain or improve yield and quality. This Small Business Innovation Research (SBIR) Phase I project seeks to develop novel deep learning algorithms and computer vision methods for coffee and for specialty crops in general, and thereby identify, through analysis, signatures of critical coffee tree stress conditions. To date, image analysis algorithms developed in academia typically focus on lab environments or controlled research plots. The tasks outlined in this project focus on real world coffee orchards with more demanding conditions. Challenges from the imagery point of view include managing the effects of clouds and the sun angle on the canopy, capturing tree foliage at an angle and identifying individual trees. Challenges from the deep learning area include identifying with confidence anomalous areas of stress or increasing stress for varying coffee strains, soil and topographical conditions within the orchard using multispectral and thermal imagery. From an agronomy standpoint, the project will seek to assist and/or correlate the deep learning findings with soil moisture sensors and other ground truth data. These elements will form the core of a low-cost commercial remote sensing and alerting system that is expected to be available to coffee-growing communities in the near future. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.