Articles
Fast sorting of defect apple fruit via X-ray imaging and artificial intelligence
Article number
1382_15
Pages
117 – 124
Language
English
Abstract
Apple fruit is stored for several months under a controlled atmosphere to provide consumers with high-quality fruit the whole year.
Unfortunately, internal browning regularly develops in the fruit flesh.
X-ray imaging has recently demonstrated to be an accurate technique to determine internal quality of fruit and vegetables.
This technology has been used to collect data from apples (Braeburn). In this work, a deep learning-based classifier was designed on 2D radiographs collected from healthy and defect fruit stored under hypoxic conditions.
An accuracy of 95±3% was reported for the classifier.
Additionally, the model was evaluated for its robustness on apples from two other orchards and stored under regular air and anoxic conditions, resulting in an accuracy of 92±1%.
Unfortunately, internal browning regularly develops in the fruit flesh.
X-ray imaging has recently demonstrated to be an accurate technique to determine internal quality of fruit and vegetables.
This technology has been used to collect data from apples (Braeburn). In this work, a deep learning-based classifier was designed on 2D radiographs collected from healthy and defect fruit stored under hypoxic conditions.
An accuracy of 95±3% was reported for the classifier.
Additionally, the model was evaluated for its robustness on apples from two other orchards and stored under regular air and anoxic conditions, resulting in an accuracy of 92±1%.
Authors
A. Tempelaere, L. Van Doorselaer, J. He, P. Verboven, B. Nicolaï
Keywords
X-ray CT, deep learning, convolutional neural network, postharvest, internal browning
Online Articles (30)