A Smart Agriculturing IoT System for Banana Plants Disease Detection through Inbuilt Compressed Sensing Devices

Abstract

Abstract The Internet of Things (IoT) solutions for agriculture are rapidly growing and have the potential to transform agriculture in many aspects. In particular, the plant disease detection devices play a vital role in improving the agriculture. The visual monitoring of plants for the onset of diseases is a tedious and time-consuming task for farmers and at the same time it is less accurate. Hence an automated system with environmental data and camera sensors can serve as an alternative and effective solution for manual monitoring of plants. In this paper, a novel and efficient compressed sensing inbuilt plant disease detection device is developed which uses a foreground-based segmentation method and two step feature extraction technique to detect and classify two of the major banana diseases. A database is created for banana bunchy top and sigatokaleaf spot diseases by collecting images in real time from the fields of southern parts of Tamilnadu namely Thadiyankudisai and Thandikudi of Dindigul district, KC Patti, Muthalapuram, Suruli Patti and Kambam of Theni district and ICAR NRCB, Tiruchirapalli. The suggested device's effectiveness has been assessed in terms of the proportion of infected areas, detection accuracy, percentage of feature reduction, and classification accuracy. The prototype of the proposed device is developed and validated using the Raspberry pi board. The findings demonstrate that the suggested device achieves classification accuracy of 97.33% and detection accuracy of 96.75%.

Plant phenotyping relevance

バナナ葉の病徴を画像から検出・分類する装置と画像解析手法を開発し、感染領域・検出精度・分類精度で検証しており、植物病害表現型の取得が中心である。

Code and data availability

The paper describes a self-collected banana disease image database (bunchy top, sigatoka leaf spot, healthy leaves) and an ORB-CS/SVM analysis pipeline, but provides no public deposit or URL for either. The Data Availability statement only offers the datasets from the corresponding author on reasonable request. All URL

No evidence-backed public reproduction asset is currently recorded.