Internet of Things‐based root disease classification in alfalfa plants using hybrid optimization‐enabled deep convolutional neural network

Abstract

Summary In this article, water cycle spider monkey optimization‐based deep convolutional neural network (WCSMO‐based deep CNN) is designed to classify root diseases in alfalfa plants. In the proposed method, the alfalfa plant root images are accessed remotely using a camera and transferred to the sink node to classify the disease. The proposed water cycle spider monkey optimization (WCSMO) algorithm performs the routing. Once the root images are received at base station or sink node, the images undergo preprocessing, segmentation, and classification. In the preprocessing module, a median filter is utilized to reduce the image noise. The preprocessed output is subjected to the segmentation phase, which is carried out using enhanced fuzzy C‐means clustering. The segmented results are passed to the classification step, where the deep convolutional neural network (deep CNN) is employed to categorize the root illness after being trained by the proposed WCSMO algorithm, which combines the water cycle algorithm and spider monkey optimization. The developed method achieved effective performance in terms of the metrics like sensitivity, accuracy, and specificity with the value of 0.9169, 0.9198, and 0.92. Moreover, the proposed WCSMO algorithm obtained the average energy and average throughput of 0.26641 J, and 91.483%, respectively.

The State of the Art in Root System Architecture Image Analysis Using Artificial Intelligence: A Review

“…In addition, roots often grow along the path of least resistance, so it has been suggested that roots grow along windows, possibly skewing aberrance and distribution data. Because of the difficulties with nondestructive techniques, a combined approach that involves both the removal of plants from their growing substrate and AI techniques that analyze root system images is growing in use, where images of root systems acquired by shovelomics are analyzed with the aid of computers using machine learning (ML) technologies These techniques have been applied to several major agricultural crops and plant types in studies concerning roots and RSA, such as alfalfa [ 14 , 17 , 33 ], apple ( Malus domestica ) [ 34 ], cotton ( Gossypium herbaceum L.) [ 35 ], maize [ 36 , 37 ], millet ( Setaria italica ) [ 38 ], rice ( Oryza sativa L.) [ 39 , 40 ], and soybean [ 16 ].…”

“…In the original paper, linear discriminant analysis was used to predict status with 76% accuracy. A different research group used the publicly available imagery to classify on the basis of the WCSMO (water cycle spider monkey optimization)-based deep CNN [ 33 ]. They used their model to analyze and classify 264 monochrome camera images of alfalfa and cotton root rot.…”

“…They used their model to analyze and classify 264 monochrome camera images of alfalfa and cotton root rot. The model achieved higher sensitivity, specificity, and accuracy scores (91.69%, 92%, and 91.98% respectively) over the other existing models they compared it to (PA-RPL + DL, EEG routing + DL, Hybrid routing + Canny and Ostu, and PriNergy + ANN) [ 33 ]. This highlights the importance of sharing imaging data and methods for both reproducibility and reuse.…”