I-Corps: Computer Vision based Pre-Harvest Yield Mapping

University Of Minnesota-Twin Cities, Minneapolis MN

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is the practical deployment of a computer-vision based yield estimation system in fruit orchards. With this technology, the farmers will be able to obtain a map of the number and size of fruit across their farms from images recorded by a standard camera. This capability will provide farmers with a useful tool in planning their harvest and sales, as well as in managing the long-term health of their plants. It will also potentially reduce the inherent risk in fruit growing, thereby improving the affordability and availability of fresh fruit in the US. Furthermore, by allowing the adoption of precision agriculture techniques to high value crops, this project may help save water and contribute to reduction of runoff pollution from fertilizer and chemicals. This I-Corps project addresses the task of customer discovery for applications of the technology of vision-based yield detection and mapping in fruit orchards. The core of the technology is a set of computer vision algorithms to detect fruit in images under natural farm conditions. This application presents several computer vision task challenges. Fruit may be occluded by leaves and other fruit. Shadows and specularities make color information unreliable. Also, fruit across frames are matched to avoid double counting. Geometric algorithms are used for estimating fruit diameter. This I-Corps project will allow for customer discovery in this and other market spaces, test the product-market fit for the proposed yield estimation technology, and guide future technology development in this area.