Intelligent image analysis for retrieval of leaf chlorophyll content of rice from digital images of smartphone under natural light

Abstract

The present study describes a new imaging method to acquire rice leaf images under field conditions using a smartphone and modeling approaches to retrieve the leaf chlorophyll (Chl) content from digitized images. Pearson's correlation of image-based color indices of the relative Chl content measured with Soil Plant Analysis Development (SPAD) indicated the suitability of the color models RGB, rgb, and DGCI-rgb. Among the linear regression models, the models based on mean brightness ratio (rgb) alone or in combination with a dark green color index (DGCI-rgb) show a good correlation between the predicted Chl content and relative Chl content. A feed-forward backpropagation-type network was also developed following the optimization of hidden neurons, training, and transfer functions. The predicted Chl contents showed a good correlation with SPAD values. Compared to the linear regression model, the developed artificial neural network model was found to be more efficient in predicting the Chl content, particularly with RGB index.

Smartphone-assisted real-time estimation of chlorophyll and carotenoid contents in spinach following the inversion of red and green color features

“…However, theoretical explanations for all such correlations from the physiological and photochemistry perspectives may not be possible. Investigations employing machine learning for image-based prediction of Chl content (Dutta Gupta et al 2013;Vesali et al 2015Vesali et al , 2017Dutta Gupta and Pattanayak 2017;Odabas et al 2015Odabas et al , 2017Mohan and Dutta Gupta 2019;Hassanijalilian et al 2020) inherently employ black-box algorithms, and hence, their exact method of computation may not be easily defined.…”

“…Furthermore, various methods have been adopted to compensate for the interference by ambient light conditions during leaf image acquisition (Hu et al 2010;Rorie et al 2011;Riccardi et al 2014;Confalonieri et al 2015;Rigon et al 2016;Mohan and Dutta Gupta 2019). However, such methods either .…”

“…Investigations employing machine learning for image-based prediction of Chl content (Dutta Gupta et al 2013; Vesali et al 2015, 2017; Dutta Gupta and Pattanayak 2017; Odabas et al 2015, 2017; Mohan and Dutta Gupta 2019; Hassanijalilian et al 2020) inherently employ black-box algorithms, and hence, their exact method of computation may not be easily defined. Furthermore, various methods have been adopted to compensate for the interference by ambient light conditions during leaf image acquisition (Hu et al 2010; Rorie et al 2011; Riccardi et al 2014; Confalonieri et al 2015; Rigon et al 2016; Mohan and Dutta Gupta 2019). However, such methods either possess limited practical feasibility or make data analysis more complicated, creating the need for a more practicable process.…”

High-Throughput Phenotyping Analysis of Potted Soybean Plants Using Colorized Depth Images Based on A Proximal Platform

“…Strong correlations between the leaf N concentration (LNC) and image color indices are observed [16]. Additionally, the significant relationship between deficiency of nitrogen and color characteristic has been proven in soybean plants in an outdoor environment [17].The color characteristic has been successfully employed to classify healthy plants versus unhealthy plants [18][19][20][21]. Visual inspection and image processing techniques are the major recognition methods for evaluating plant health conditions.…”

“…The color characteristic has been successfully employed to classify healthy plants versus unhealthy plants [18][19][20][21]. Visual inspection and image processing techniques are the major recognition methods for evaluating plant health conditions.…”