Review
Deep learning in agriculture: A survey

https://doi.org/10.1016/j.compag.2018.02.016Get rights and content

Highlights

  • •
    Survey on the deep learning technique applied in agriculture.
  • •
    Detailed review of 40 relevant research papers.
  • •
    Examining research area, technical details, data sources and performance achieved.
  • •
    Deep learning offers high precision outperforming other image processing techniques.
  • •
    Discussion on advanced deep learning models used in various agricultural problems.
  • •
    Status, advantages, disadvantages and potential of deep learning in agriculture.
  • •
    Potential future applications in agriculture using deep learning.

Abstract

Deep learning constitutes a recent, modern technique for image processing and data analysis, with promising results and large potential. As deep learning has been successfully applied in various domains, it has recently entered also the domain of agriculture. In this paper, we perform a survey of 40 research efforts that employ deep learning techniques, applied to various agricultural and food production challenges. We examine the particular agricultural problems under study, the specific models and frameworks employed, the sources, nature and pre-processing of data used, and the overall performance achieved according to the metrics used at each work under study. Moreover, we study comparisons of deep learning with other existing popular techniques, in respect to differences in classification or regression performance. Our findings indicate that deep learning provides high accuracy, outperforming existing commonly used image processing techniques.

Introduction

Smart farming (Tyagi, 2016) is important for tackling the challenges of agricultural production in terms of productivity, environmental impact, food security and sustainability (Gebbers and Adamchuk, 2010). As the global population has been continuously increasing (Kitzes et al., 2008), a large increase on food production must be achieved (FAO, 2009), maintaining at the same time availability and high nutritional quality across the globe, protecting the natural ecosystems by using sustainable farming procedures.
To address these challenges, the complex, multivariate and unpredictable agricultural ecosystems need to be better understood by monitoring, measuring and analyzing continuously various physical aspects and phenomena. This implies analysis of big agricultural data (Kamilaris et al., 2017b), and the use of new information and communication technologies (ICT) (Kamilaris et al., 2016), both for short-scale crop/farm management as well as for larger-scale ecosystems’ observation, enhancing the existing tasks of management and decision/policy making by context, situation and location awareness. Larger-scale observation is facilitated by remote sensing (Bastiaanssen et al., 2000), performed by means of satellites, airplanes and unmanned aerial vehicles (UAV) (i.e. drones), providing wide-view snapshots of the agricultural environments. It has several advantages when applied to agriculture, being a well-known, non-destructive method to collect information about earth features while data may be obtained systematically over large geographical areas.
A large subset of the volume of data collected through remote sensing involve images. Images constitute, in many cases, a complete picture of the agricultural environments and could address a variety of challenges (Liaghat and Balasundram, 2010, Ozdogan et al., 2010). Hence, imaging analysis is an important research area in the agricultural domain and intelligent data analysis techniques are being used for image identification/classification, anomaly detection etc., in various agricultural applications (Teke et al., 2013, Saxena and Armstrong, 2014, Singh et al., 2016). The most popular techniques and applications are presented in Appendix A, together with the sensing methods employed to acquire the images. From existing sensing methods, the most common one is satellite-based, using multi-spectral and hyperspectral imaging. Synthetic aperture radar (SAR), thermal and near infrared (NIR) cameras are being used in a lesser but increasing extent (Ishimwe et al., 2014), while optical and X-ray imaging are being applied in fruit and packaged food grading. The most popular techniques used for analyzing images include machine learning (ML) (K-means, support vector machines (SVM), artificial neural networks (ANN) amongst others), linear polarizations, wavelet-based filtering, vegetation indices (NDVI) and regression analysis (Saxena and Armstrong, 2014, Singh et al., 2016).
Besides the aforementioned techniques, a new one which is recently gaining momentum is deep learning (DL) (LeCun et al., 2015, LeCun and Bengio, 1995). DL belongs to the machine learning computational field and is similar to ANN. However, DL is about “deeper” neural networks that provide a hierarchical representation of the data by means of various convolutions. This allows larger learning capabilities and thus higher performance and precision. A brief description of DL is attempted in Section 3.
The motivation for preparing this survey stems from the fact that DL in agriculture is a recent, modern and promising technique with growing popularity, while advancements and applications of DL in other domains indicate its large potential. The fact that today there exists at least 40 research efforts employing DL to address various agricultural problems with very good results, encouraged the authors to prepare this survey. To the authors’ knowledge, this is the first such survey in the agricultural domain, while a small number of more general surveys do exist (Deng and Yu, 2014, Wan et al., 2014, Najafabadi et al., 2015), covering related work in DL in other domains.

Section snippets

Methodology

The bibliographic analysis in the domain under study involved two steps: (a) collection of related work and (b) detailed review and analysis of this work. In the first step, a keyword-based search for conference papers or journal articles was performed from the scientific databases IEEE Xplore and ScienceDirect, and from the web scientific indexing services Web of Science and Google Scholar. As search keywords, we used the following query:
[“deep learning”] AND [“agriculture” OR ”farming“]
In

Deep learning

DL extends classical ML by adding more “depth” (complexity) into the model as well as transforming the data using various functions that allow data representation in a hierarchical way, through several levels of abstraction (Schmidhuber, 2015, LeCun and Bengio, 1995). A strong advantage of DL is feature learning, i.e. the automatic feature extraction from raw data, with features from higher levels of the hierarchy being formed by the composition of lower level features (LeCun et al., 2015). DL

Deep learning applications in agriculture

In Appendix B, we list the 40 identified relevant works, indicating the agricultural-related research area, the particular problem they address, DL models and architectures implemented, sources of data used, classes and labels of the data, data pre-processing and/or augmentation employed, overall performance achieved according to the metrics adopted, as well as comparisons with other techniques, wherever available.

Discussion

Our analysis has shown that DL offers superior performance in the vast majority of related work. When comparing the performance of DL-based approaches with other techniques at each paper, it is of paramount importance to adhere to the same experimental conditions (i.e. datasets and performance metrics). From the related work under study, 28 out of the 40 papers (70%) performed direct, valid and correct comparisons among the DL-based approach employed and other state-of-art techniques used to

Conclusion

In this paper, we have performed a survey of deep learning-based research efforts applied in the agricultural domain. We have identified 40 relevant papers, examining the particular area and problem they focus on, technical details of the models employed, sources of data used, pre-processing tasks and data augmentation techniques adopted, and overall performance according to the performance metrics employed by each paper. We have then compared deep learning with other existing techniques, in

Acknowledgments

We would like to thank the reviewers, whose valuable feedback, suggestions and comments increased significantly the overall quality of this survey. This research has been supported by the P-SPHERE project, which has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skodowska-Curie grant agreement No 665919.

Research data for this article

Mendeley logo
for download under the CC BY NC 3.0 licence
Data for: Deep Learning in Agriculture: A Survey
Survey on the deep learning technique applied in agriculture.
Detailed review of 40 relevant research papers, examining research area and problem they focus on, technical details on deep learning models, sources of data, pre-processing and data augmentation techniques used, and…

Dataset

relatedwork_analysis_advanced.ods
36KB
relatedwork_analysis_basic.ods
16KB
View dataset on Mendeley Data
Further information on research data

References (72)

Cited by (3268)

  • Using deep transfer learning for image-based plant disease identification

    2020, Computers and Electronics in Agriculture

  • Remote sensing for agricultural applications: A meta-review

    2020, Remote Sensing of Environment

  • A comprehensive review on automation in agriculture using artificial intelligence

    2019, Artificial Intelligence in Agriculture

  • Apple detection during different growth stages in orchards using the improved YOLO-V3 model

    2019, Computers and Electronics in Agriculture

  • Recent advances of hyperspectral imaging technology and applications in agriculture

    2020, Remote Sensing

  • A review on UAV-based applications for precision agriculture

    2019, Information Switzerland
View all citing articles on Scopus
View full text
© 2018 Elsevier B.V. All rights reserved.