2017 Volume 26 Issue 4 Pages 100-114
We describe a method for automatically extracting fruit growth information from multiple apple tree images using a high-resolution image monitoring system. First, Green-Blue Vegetation Index images were created from the green and blue channels in 3,281 high-resolution RGB images taken from April to December 2016. Leaf zones were extracted from the background, and their area and fractal dimension were calculated. Their statistical properties were evaluated, data were sorted, and fruit growth was approximated. Radius (in pixels) was used as an index of fruit growth, and 1,006 measurements were made from 6 fruits in a Web image viewer. When the explanatory variable was the integrated value of the area or the fractal dimension of the leaf area and the objective variable was the daily average value of the fruit radius, the coefficient of determination by logistic curve approximation was ≥0.99. These results suggest that fruit growth information can be extracted automatically from a large number of high-definition images.
