Vision-Based Preharvest Yield Mapping for Apple Orchards

  • 2018-08-13 17:23:16
  • Pravakar Roy, Abhijeet Kislay, Patrick A. Plonski, James Luby, Volkan Isler
  • 1

Abstract

We present an end-to-end computer vision system for mapping yield in an appleorchard using images captured from a single camera. Our proposed system isplatform independent and does not require any specific lighting conditions. Ourmain technical contributions are 1)~a semi-supervised clustering algorithm thatutilizes colors to identify apples and 2)~an unsupervised clustering methodthat utilizes spatial properties to estimate fruit counts from apple clustershaving arbitrarily complex geometry. Additionally, we utilize camera motion tomerge the counts across multiple views. We verified the performance of ouralgorithms by conducting multiple field trials on three tree rows consisting of$252$ trees at the University of Minnesota Horticultural Research Center.Results indicate that the detection method achieves $F_1$-measure $.95 -.97$for multiple color varieties and lighting conditions. The counting methodachieves an accuracy of $89\%-98\%$. Additionally, we report merged fruitcounts from both sides of the tree rows. Our yield estimation method achievesan overall accuracy of $91.98\% - 94.81\%$ across different datasets.

 

Quick Read (beta)

loading the full paper ...