Novel Vegetation Indices to Identify Broccoli Plants Infected With Xanthomonas campestris pv. campestris
Metadatos
Mostrar el registro completo del ítemEditorial
Frontiers
Materia
Biotic stress Climate change Hyperspectral reflectance imaging Machine learning Thermography
Fecha
2022-06-23Referencia bibliográfica
Pineda M, Pérez-Bueno ML and Barón M (2022) Novel Vegetation Indices to Identify Broccoli Plants Infected With Xanthomonas campestris pv. campestris. Front. Plant Sci. 13:790268. doi: [10.3389/fpls.2022.790268]
Patrocinador
MCIN/AEI RTI2018-094652-B-I00; ERDF: A way of making Europe; Consejo Superior de Investigaciones Cientificas (CSIC) through the Unidad de Recursos de Informacion Cientifica para la Investigacion (URICI)Resumen
A rapid diagnosis of black rot in brassicas, a devastating disease caused by
Xanthomonas campestris pv. campestris (Xcc), would be desirable to avoid significant
crop yield losses. The main aim of this work was to develop a method of detection of
Xcc infection on broccoli leaves. Such method is based on the use of imaging sensors
that capture information about the optical properties of leaves and provide data that can
be implemented on machine learning algorithms capable of learning patterns. Based on
this knowledge, the algorithms are able to classify plants into categories (healthy and
infected). To ensure the robustness of the detection method upon future alterations in
climate conditions, the response of broccoli plants to Xcc infection was analyzed under
a range of growing environments, taking current climate conditions as reference. Two
projections for years 2081–2100 were selected, according to the Assessment Report of
Intergovernmental Panel on Climate Change. Thus, the response of broccoli plants to
Xcc infection and climate conditions has been monitored using leaf temperature and five
conventional vegetation indices (VIs) derived from hyperspectral reflectance. In addition,
three novel VIs, named diseased broccoli indices (DBI1-DBI3), were defined based on
the spectral reflectance signature of broccoli leaves upon Xcc infection. Finally, the nine
parameters were implemented on several classifying algorithms. The detection method
offering the best performance of classification was a multilayer perceptron-based artificial
neural network. This model identified infected plants with accuracies of 88.1, 76.9, and
83.3%, depending on the growing conditions. In this model, the three Vis described in
this work proved to be very informative parameters for the disease detection. To our
best knowledge, this is the first time that future climate conditions have been taken into
account to develop a robust detection model using classifying algorithms.