Automating mango crop yield estimation

Several imaging technologies were assessed for determination of mango crop load, including hyperspectral and thermal imaging, but with preference given to RGB imaging.
Sets of RGB images of mango fruits on trees were collected under a range of orchard conditions over five seasons.
A number of different image processing algorithms were evaluated, and positive identification rates of up to 80% were achieved.
Approaches included colour and texture filters, contour detection and pixel classification using various approaches.
Issues that affect load estimates, occlusion of fruits by leaves and flower stalks, fruit bunching, shadowing, triggering, and insufficient lighting are discussed.
These image sets are available on request for comparative work using other image processing algorithms.

XXIX International Horticultural Congress on Horticulture: Sustaining Lives, Livelihoods and Landscapes (IHC2014): International Symposia on the Physiology of Perennial Fruit Crops and Production Systems and Mechanisation, Precision Horticulture and Robotics

A.B. Payne, K.B. Walsh, P.P. Subedi

fruit, processing, colour, texture, segmentation, count

https://doi.org/10.17660/ActaHortic.2016.1130.87