Spectral Identification of Disease in Weeds Using Multilayer Perceptron with Automatic Relevance Determination

Microbotryum silybum, a smut fungus, is studied as an agent for the biological control of Silybum marianum (milk thistle) weed. Confirmation of the systemic infection is essential in order to assess the effectiveness of the biological control application and assist decision-making. Nonetheless, in s...

Full description

Bibliographic Details
Main Authors: Afroditi Alexandra Tamouridou, Xanthoula Eirini Pantazi, Thomas Alexandridis, Anastasia Lagopodi, Giorgos Kontouris, Dimitrios Moshou
Format: Article
Language:English
Published: MDPI AG 2018-08-01
Series:Sensors
Subjects:
Online Access:http://www.mdpi.com/1424-8220/18/9/2770
Description
Summary:Microbotryum silybum, a smut fungus, is studied as an agent for the biological control of Silybum marianum (milk thistle) weed. Confirmation of the systemic infection is essential in order to assess the effectiveness of the biological control application and assist decision-making. Nonetheless, in situ diagnosis is challenging. The presently demonstrated research illustrates the identification process of systemically infected S. marianum plants by means of field spectroscopy and the multilayer perceptron/automatic relevance determination (MLP-ARD) network. Leaf spectral signatures were obtained from both healthy and infected S. marianum plants using a portable visible and near-infrared spectrometer (310–1100 nm). The MLP-ARD algorithm was applied for the recognition of the infected S. marianum plants. Pre-processed spectral signatures served as input features. The spectra pre-processing consisted of normalization, and second derivative and principal component extraction. MLP-ARD reached a high overall accuracy (90.32%) in the identification process. The research results establish the capacity of MLP-ARD to precisely identify systemically infected S. marianum weeds during their vegetative growth stage.
ISSN:1424-8220