基于smogw的深度CNN:使用SMoGW-deep CNN分类器进行植物病害检测和分类

IF 0.2 Q4 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
A. Pahurkar, Ravindra M. Deshmukh
{"title":"基于smogw的深度CNN:使用SMoGW-deep CNN分类器进行植物病害检测和分类","authors":"A. Pahurkar, Ravindra M. Deshmukh","doi":"10.3233/web-230015","DOIUrl":null,"url":null,"abstract":"Diagnosing plant disease is a major role to reduce adequate losses in yield production, which further leads to economic losses. The various disease control measures are accessible without a proper diagnosis of the disease which results in a waste of expenses and time. The diagnosis of disease using images leads to unsatisfactory results in the prevalent methods due to the image clarity. It is mainly caused by the worst performance of the existing pre-trained image classifiers. This issue can be controlled by the SMoGW-deep convolutional neural network (deep CNN) classifier for the accurate and precise classification of plant leaf disease. The developed method transforms the poor-quality captured images into high quality by the preprocessing technique. The preprocessed input images contain pixels on their dimension and also the value of the threshold is analyzed by the Otsu method by which the particular disease-affected region is extracted based on the image pixels. The region of interest is separated from the other parts of the input leaf image using the K-means segmentation technique. The stored features in the feature vector are fed forward to the deep CNN classifier for training and are optimized by the SMoGW optimization approach. The experiments are done and achieved an accuracy of 94.5% sensitivity of 94.525%, specificity of 94.6%, precision of 95% with 90% of training data and under K-fold training with 95% of accuracy, 95% of sensitivity, 94.1% of specificity, and 92.1% of precession is achieved for the SMoGW-optimized classifier approach that is superior to the prevalent techniques for disease classification and detection. The potential, as well as the capability of the proposed method, is experimentally demonstrated for plant leaf disease classification and identification.","PeriodicalId":42775,"journal":{"name":"Web Intelligence","volume":null,"pages":null},"PeriodicalIF":0.2000,"publicationDate":"2023-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"SMoGW-based deep CNN: Plant disease detection and classification using SMoGW-deep CNN classifier\",\"authors\":\"A. Pahurkar, Ravindra M. Deshmukh\",\"doi\":\"10.3233/web-230015\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Diagnosing plant disease is a major role to reduce adequate losses in yield production, which further leads to economic losses. The various disease control measures are accessible without a proper diagnosis of the disease which results in a waste of expenses and time. The diagnosis of disease using images leads to unsatisfactory results in the prevalent methods due to the image clarity. It is mainly caused by the worst performance of the existing pre-trained image classifiers. This issue can be controlled by the SMoGW-deep convolutional neural network (deep CNN) classifier for the accurate and precise classification of plant leaf disease. The developed method transforms the poor-quality captured images into high quality by the preprocessing technique. The preprocessed input images contain pixels on their dimension and also the value of the threshold is analyzed by the Otsu method by which the particular disease-affected region is extracted based on the image pixels. The region of interest is separated from the other parts of the input leaf image using the K-means segmentation technique. The stored features in the feature vector are fed forward to the deep CNN classifier for training and are optimized by the SMoGW optimization approach. The experiments are done and achieved an accuracy of 94.5% sensitivity of 94.525%, specificity of 94.6%, precision of 95% with 90% of training data and under K-fold training with 95% of accuracy, 95% of sensitivity, 94.1% of specificity, and 92.1% of precession is achieved for the SMoGW-optimized classifier approach that is superior to the prevalent techniques for disease classification and detection. The potential, as well as the capability of the proposed method, is experimentally demonstrated for plant leaf disease classification and identification.\",\"PeriodicalId\":42775,\"journal\":{\"name\":\"Web Intelligence\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.2000,\"publicationDate\":\"2023-06-28\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Web Intelligence\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.3233/web-230015\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q4\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Web Intelligence","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.3233/web-230015","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

摘要

诊断植物病害是减少产量损失的重要手段,而产量损失又进一步导致经济损失。在没有对疾病进行适当诊断的情况下,可以采取各种疾病控制措施,这导致了费用和时间的浪费。由于图像清晰度的限制,目前常用的疾病诊断方法对疾病的诊断效果并不理想。这主要是由于现有的预训练图像分类器性能最差造成的。这个问题可以通过SMoGW-deep convolutional neural network (deep CNN)分类器来控制,实现对植物叶片病害的精准分类。该方法通过预处理技术将捕获的低质量图像转化为高质量图像。预处理后的输入图像在其维度上包含像素,并且通过Otsu方法分析阈值,通过该方法根据图像像素提取特定的疾病影响区域。使用k均值分割技术将感兴趣的区域与输入叶子图像的其他部分分离开来。将特征向量中存储的特征前馈给深度CNN分类器进行训练,并通过SMoGW优化方法进行优化。实验结果表明,在90%的训练数据下,smogw优化分类器的准确率为94.5%,灵敏度为94.525%,特异性为94.6%,精密度为95%,K-fold训练下,准确率为95%,灵敏度为95%,特异性为94.1%,进动率为92.1%,优于目前流行的疾病分类和检测技术。实验证明了该方法在植物叶片病害分类和鉴定方面的潜力和能力。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
SMoGW-based deep CNN: Plant disease detection and classification using SMoGW-deep CNN classifier
Diagnosing plant disease is a major role to reduce adequate losses in yield production, which further leads to economic losses. The various disease control measures are accessible without a proper diagnosis of the disease which results in a waste of expenses and time. The diagnosis of disease using images leads to unsatisfactory results in the prevalent methods due to the image clarity. It is mainly caused by the worst performance of the existing pre-trained image classifiers. This issue can be controlled by the SMoGW-deep convolutional neural network (deep CNN) classifier for the accurate and precise classification of plant leaf disease. The developed method transforms the poor-quality captured images into high quality by the preprocessing technique. The preprocessed input images contain pixels on their dimension and also the value of the threshold is analyzed by the Otsu method by which the particular disease-affected region is extracted based on the image pixels. The region of interest is separated from the other parts of the input leaf image using the K-means segmentation technique. The stored features in the feature vector are fed forward to the deep CNN classifier for training and are optimized by the SMoGW optimization approach. The experiments are done and achieved an accuracy of 94.5% sensitivity of 94.525%, specificity of 94.6%, precision of 95% with 90% of training data and under K-fold training with 95% of accuracy, 95% of sensitivity, 94.1% of specificity, and 92.1% of precession is achieved for the SMoGW-optimized classifier approach that is superior to the prevalent techniques for disease classification and detection. The potential, as well as the capability of the proposed method, is experimentally demonstrated for plant leaf disease classification and identification.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
Web Intelligence
Web Intelligence COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE-
CiteScore
0.90
自引率
0.00%
发文量
35
期刊介绍: Web Intelligence (WI) is an official journal of the Web Intelligence Consortium (WIC), an international organization dedicated to promoting collaborative scientific research and industrial development in the era of Web intelligence. WI seeks to collaborate with major societies and international conferences in the field. WI is a peer-reviewed journal, which publishes four issues a year, in both online and print form. WI aims to achieve a multi-disciplinary balance between research advances in theories and methods usually associated with Collective Intelligence, Data Science, Human-Centric Computing, Knowledge Management, and Network Science. It is committed to publishing research that both deepen the understanding of computational, logical, cognitive, physical, and social foundations of the future Web, and enable the development and application of technologies based on Web intelligence. The journal features high-quality, original research papers (including state-of-the-art reviews), brief papers, and letters in all theoretical and technology areas that make up the field of WI. The papers should clearly focus on some of the following areas of interest: a. Collective Intelligence[...] b. Data Science[...] c. Human-Centric Computing[...] d. Knowledge Management[...] e. Network Science[...]
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信