Advanced Progressive Graph Convolutional Networks for Early Detection and Monitoring of Plant Infections and Disease Progression in Smart Agriculture

Abstract

ABSTRACT Plant disease is a physiological or structural problem caused by pathogens such as fungi, bacteria, viruses, or environmental factors, which disrupts plant development, yield, and overall health. Furthermore, the formation of new and more aggressive diseases complicates disease control, making it harder for farmers to preserve their crops while ensuring consistent food production. In this manuscript, to advance Progressive Graph Convolutional Networks enable early detection and continuous monitoring of plant infections in smart agriculture (PGCN‐EDM‐PID) is proposed. Initially, input images of food grains such as rice, wheat, and maize are collected from internet sources. To implement this, the input image is preprocessed using the Adaptive Two‐Stage Unscented Kalman Filter (ATSUKF), which performs resizing, sharpening, cropping, contrast enhancement, brightness adjustment, and Gaussian blurring on the images from the dataset. Then the preprocessed images are augmented based on horizontal flip, width shift, height shift, vertical flip, rotation range, shear, zoom and brightness. Additionally, Make Sense AI is proposed to annotate the images in the dataset under each class. Then the preprocessed and augmented images are fed to Progressive Graph Convolutional Networks (PGCN) to detect and classify the plant diseases. Generally, PGCN does not show adapting optimization approaches to find ideal factors to assure accurate plant disease detection. Therefore, the Augmented Red Panda Optimizer (ARPO) was proposed to optimize the weight parameter of PGCN, which accurately detects the plant disease. Then the proposed PGCN‐EDM‐PID is executed in Python and the performance metrics such as Accuracy, Precision, False Positive Rate (FPR), True Positive Rate (TPR), Specificity, Recall, F1‐score, Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) are analyzed. Performance of the PGCN‐EDM‐PID approach attains high accuracy, high Precision, high Recall when analyzed through existing techniques like Real‐time plant disease dataset improvement and detection of plant disease utilizing DL (PDD‐DPD‐CNN), Detection of plant leaf diseasesusing deep convolutional neural network methods (DPLD‐DCNN), New DL algorithm for cross‐crop detection of plant disease: A generalized model for detecting unhealthy leaves (CPDD‐SVM) methods respectively.

Plant phenotyping relevance

植物画像から病害を検出・分類する画像解析手法の開発と性能評価が研究の中心であり、感染状態という植物表現型を直接推定している。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.