A novel hybrid fruit fly and simulated annealing optimized faster R-CNN for detection and classification of tomato plant leaf diseases

Abstract

Modern agriculture increasingly relies on technologies that enhance farmers' efficiency and economic growth. One challenge is the accurate identification of disease-affected plants, whose characteristics like structure, size, texture, and color can vary significantly. While there are existing methods to detect and classify these diseases, challenges such as image noise, hyper-parameter selection, and over-fitting can impede prediction accuracy. This paper introduces a hybrid fruit fly and simulated annealing-optimized Faster R-CNN (FS-FRNet) for improved plant leaf disease identification and classification. Our novel FS-FRNet method integrates a Wiener filter for de-noising and a super-resolution method to enhance image quality. By hybridizing the fruit fly optimization algorithm and simulated annealing, the Faster R-CNN's hyper-parameter issues are addressed, and the convergence rate is improved. We applied the FS-FRNet to identify and classify tomato plant diseases like early blight, yellow leaf curl, Septoria leaf, mosaic virus, and late blight. Experimental outcomes on the Plant Village dataset show that our method outperforms existing techniques, achieving 98.3% accuracy, 98.04% precision, and 98.11% recall, thus confirming its efficacy for reliable detection of tomato plant leaf diseases.

Plant phenotyping relevance

トマト葉の病害状態を画像から検出・分類する手法を開発し、データセット上で性能評価しているため、植物表現型（病害状態）の取得・推定が中心である。

Code and data availability

The paper uses the public Plant Village dataset, but this is a generic third-party benchmark, not a paper-specific asset. No author code, trained models, or data availability statements with public URLs appear in the supplied blocks.

No evidence-backed public reproduction asset is currently recorded.