Article
Orchard Mapping with Deep Learning Semantic Segmentation
Athanasios Anagnostis
1,2
, Aristotelis C. Tagarakis
1
, Dimitrios Kateris
1,
* , Vasileios Moysiadis
1
,
Claus Grøn Sørensen
3
, Simon Pearson
4
and Dionysis Bochtis
1,5
Citation: Anagnostis, A.;
Tagarakis, A.C.; Kateris, D.;
Moysiadis, V.; Sørensen, C.G.;
Pearson, S.; Bochtis, D. Orchard
Mapping with Deep Learning
Semantic Segmentation. Sensors 2021,
21, 3813. https://doi.org/10.3390/
s21113813
Academic Editor: Asim Biswas
Received: 21 April 2021
Accepted: 27 May 2021
Published: 31 May 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
Institute for Bio-Economy and Agri-Technology (iBO), Centre for Research and Technology–Hellas (CERTH),
GR57001 Thessaloniki, Greece; a.anagnostis@certh.gr (A.A.); a.tagarakis@certh.gr (A.C.T.);
v.moisiadis@certh.gr (V.M.); d.bochtis@certh.gr (D.B.)
2
Department of Computer Science & Telecommunications, University of Thessaly, GR35131 Lamia, Greece
3
Department of Electrical and Computer Engineering, Aarhus University, DK-8000 Aarhus C, Denmark;
claus.soerensen@ece.au.dk
4
Lincoln Institute for Agri-Food Technology (LIAT), University of Lincoln, Lincoln LN6 7TS, UK;
spearson@lincoln.ac.uk
5
farmB Digital Agriculture P.C., Doiranis 17, GR54639 Thessaloniki, Greece
* Correspondence: d.kateris@certh.gr; Tel.: +30-242-109-6740
Abstract:
This study aimed to propose an approach for orchard trees segmentation using aerial
images based on a deep learning convolutional neural network variant, namely the U-net network.
The purpose was the automated detection and localization of the canopy of orchard trees under
various conditions (i.e., different seasons, different tree ages, different levels of weed coverage). The
implemented dataset was composed of images from three different walnut orchards. The achieved
variability of the dataset resulted in obtaining images that fell under seven different use cases.
The best-trained model achieved 91%, 90%, and 87% accuracy for training, validation, and testing,
respectively. The trained model was also tested on never-before-seen orthomosaic images or orchards
based on two methods (oversampling and undersampling) in order to tackle issues with out-of-
the-field boundary transparent pixels from the image. Even though the training dataset did not
contain orthomosaic images, it achieved performance levels that reached up to 99%, demonstrating
the robustness of the proposed approach.
Keywords:
precision agriculture; orchard mapping; deep learning; computer vision; semantic
segmentation; orthomosaic
1. Introduction
The latest advances in sensing technologies dedicated to agricultural systems have led
to the emergence and development of a modern management concept, namely precision
agriculture, which focuses on efficient management of the temporal and spatial variability
of field and crop properties using information and communication technology (ICT) [
1
]. A
plethora of different sensors and technologies are utilized in relation to this concept to form
a detailed view of fields’ properties, capturing the spatial and temporal variability and
searching for the specific factors responsible for their occurrence, which are to be treated
accordingly. Therefore, mapping the field and crop properties is a fundamental aspect in
the application of such management systems.
Remote sensing is defined as the non-contact measurement of crop properties based
on the radiation reflected from the plants, using ground based or aerial platforms, and
it is widely used for mapping tasks in agricultural systems [
2
]. Recent technological
advances have made unmanned aerial systems (UASs), i.e., sensing systems mounted on
unmanned aerial vehicles (UAVs), commercially available. These systems provide high
spatial resolution images and, in combination with their ease of use, quick acquisition
times, and low operational cost, they have become particularly popular for monitoring
agricultural fields [
3
]. Several studies have utilized UASs for crop management purposes,
Sensors 2021, 21, 3813. https://doi.org/10.3390/s21113813 https://www.mdpi.com/journal/sensors