Articles
Machine vision methods in forecast of mango crop
Article number
1395_39
Pages
297 – 304
Language
English
Abstract
A forward estimate of fruit load is required by mango farm managers for planning of packaging material, labour, machinery resourcing, and market sales.
Current practice for fruit load estimation involves manual counting of total fruit tree‑1. The current study was implemented to extend the use of an orchard imaging system in estimation of mango fruit load, working towards a method for implementation by farm management.
A farm vehicle mounted machine vision system was used for estimation of both flowering level and fruit number.
Panicle development was classified to three maturity stages (elongation stage, X; 50% flower opening stage, Y; and fruit set stage, Z), while a correction based on a manual count was applied to fruit count.
Estimates were undertaken on 39 orchard blocks in two growing regions, involving two seasons and two cultivars.
The time of the peak in count of panicles in stage Y was used as the date of flowering event in each block and used in GDD estimates of harvest timing, based on data from imaging of several rows per block through the flowering period.
For estimation of orchard fruit load, imaging on two dates is recommended to capture fruit arising from different flowering events.
Fruit load estimations based on flowering level and fruit count achieved a R2 of 0.93 and 0.98, with slope of 0.92 and 0.62, respectively, relative to packhouse harvest records, across six orchards.
Current practice for fruit load estimation involves manual counting of total fruit tree‑1. The current study was implemented to extend the use of an orchard imaging system in estimation of mango fruit load, working towards a method for implementation by farm management.
A farm vehicle mounted machine vision system was used for estimation of both flowering level and fruit number.
Panicle development was classified to three maturity stages (elongation stage, X; 50% flower opening stage, Y; and fruit set stage, Z), while a correction based on a manual count was applied to fruit count.
Estimates were undertaken on 39 orchard blocks in two growing regions, involving two seasons and two cultivars.
The time of the peak in count of panicles in stage Y was used as the date of flowering event in each block and used in GDD estimates of harvest timing, based on data from imaging of several rows per block through the flowering period.
For estimation of orchard fruit load, imaging on two dates is recommended to capture fruit arising from different flowering events.
Fruit load estimations based on flowering level and fruit count achieved a R2 of 0.93 and 0.98, with slope of 0.92 and 0.62, respectively, relative to packhouse harvest records, across six orchards.
Authors
M. Amaral, A. Koirala, K.B. Walsh
Keywords
count, GDD, heat units, prediction, size, YOLO
Groups involved
- Division Temperate Tree Nuts
- Division Vine and Berry Fruits
- Division Temperate Tree Fruits
- Division Precision Horticulture and Engineering
- Division Plant-Environment Interactions in Field Systems
- Division Tropical and Subtropical Fruit and Nuts
- Working Group Precision Management of Orchards and Vineyards
Online Articles (59)