Nondestructive Detection Method for the Calcium and Nitrogen Content of Living Plants Based on Convolutional Neural Networks (CNN) Using Multispectral Images

• Published in: Agriculture (Basel) 2022-06, Vol.12 (6), p.747
• Main Authors: Kunstman, Grzegorz, Kunstman, Paweł, Lasyk, Łukasz, Nowak, Jacek Stanisław, Stępowska, Agnieszka, Kowalczyk, Waldemar, Dybaś, Jakub, Szczęsny-Małysiak, Ewa
• Format: Article
• Language: English
• Subjects: Agricultural production; Agriculture; Analytical chemistry; Artificial neural networks; Cell division; Computer vision; Contamination; Cost analysis; Crop diseases; Crops; Data analysis; Fertilizers; Food quality; Image acquisition; Neural networks; Nitrogen; Nondestructive testing; Nutrition; Nutritional status; Physiology; Plant diseases; Plant growth; plant nutrients; Remote sensing; spectral imaging; Spectrum analysis; Wavelengths
• ISSN: 2077-0472; 2077-0472
• DOI: 10.3390/agriculture12060747

Herein, we present the novel method targeted for determination of plant nutritional state with the use of computer vision and Neural Networks. The method is based on multispectral imaging performed by an exclusively designed Agroscanner and a dedicated analytical system for further data analysis with Neural Networks. An Agroscanner is a low-cost mobile construction intended for multispectral measurements at macro-scale, operating at four wavelengths: 470, 550, 640 and 850 nm. Together with developed software and implementation of a Neural Network it was possible to design a unique approach to process acquired plant images and assess information about plant physiological state. The novelty of the developed technology is focused on the multispectral, macro-scale analysis of individual plant leaves, rather than entire fields. Such an approach makes the method highly sensitive and precise. The method presented herein determines the basic physiological deficiencies of crops with around 80% efficiency.