IMAGE SET FOR FLOWER DETECTION IN APPLE TREES

Published: 07-06-2020| Version 2 | DOI: 10.17632/bpbsfs26td.2
Contributors:
Dolores del Brio,
Pablo Reeb,
Dario Fernandez

Description

This dataset is composed of 58 apple tree images. They were taken during full bloom in an apple orchard located at National Institute of Agricultural Technology (INTA) Experimental Station, General Roca, Argentina (39° 1’ 34’’ S; 67° 44’ 24.6’’ W).The orchard was established in 2000 with ‘Red Chief’ cultivar apple trees grafted on MM111 rootstock .Trees were planted in a total area of 0.8 ha at a distance of 1.5m between trees by 4m between rows and were trained as espalier. A few days before full bloom, images were obtained from 32 trees, under two conditions: i) natural daylight between 10 am and 13 pm, ii) at night with the artificial flash light of the camera. A black curtain was unfolded behind the trees when images were obtained under daylight conditions in order to avoid interference from neighboring trees. All images were taken with an RGB digital camera (14.1 MP) at approximately 3.0 m from the tree in a 90° angle to the row. An object of known dimensions (a 15x15 cm square) was placed in each tree as a scale reference. Simultaneously, all the flowers on each tree were manually counted. Images taken by using different proximal sensors can be used to estimate the number of flowers or fruits in trees. The accuracy of those methods have been studied and tested by many researchers during the past few years with encouraging results. Since manual counting method in fruit crops is a time-consuming labor and also lead to inaccurate results, image analysis could be used as an alternative procedure. The use of images could provide reliable and consistent data.

Files