Using machine vision in mango orchard management

Z. Wang, A. Koirala, K.B. Walsh
Machine vision has several potential applications in mango orchard management - examples are presented for (i) flowering assessment, towards feeding heat sum maturation models, to inform selective tree spraying and to inform selective harvest of early flowering trees; (ii) to estimate fruit size and (iii) fruit load of orchard trees, to build yield monitoring and mapping systems, and supporting orchard management decisions such as labour and materials acquisition. A utility mounted rig was used, with night imaging employed using LED lighting to reduce light variability, background issues and sunlight interference to a Time of Flight camera. A differential GPS receiver was adopted to trigger camera imaging and correspond acquired data to each tree. Common machine vision techniques such as Otsu's method, colour thresholding, morphological operations, support vector machine, ellipse fitting and a histogram of oriented gradients (HOG) based cascade classifier were used to identify all potential regions of interest (flowers or fruits) in the RGB images. By the utilization of depth information and thin lens formula, the sizes of the well-separated fruits could be calculated and used for representatives. All the information was finally uploaded to an interactive website server for visualisation.
Wang, Z., Koirala, A. and Walsh, K.B. (2019). Using machine vision in mango orchard management. Acta Hortic. 1244, 109-116
DOI: 10.17660/ActaHortic.2019.1244.17
https://doi.org/10.17660/ActaHortic.2019.1244.17
mango orchard, flowering assessment, fruit sizing, yield estimation, heat sum
English
1244_17
109-116

Acta Horticulturae