Identification of banana leaf disease based on KVA and GR-ARNet

作     者：Jinsheng Deng Weiqi Huang Guoxiong Zhou Yahui Hu Liujun Li Yanfeng Wang 

作者机构：College of Electronic Information and Physics Central South University of Forestry and Technology Plant Protection Research Institute Hunan Academy of Agricultural Sciences Department of Soil and Water Systems College of Agricultural & Life Sciences University of Idaho College of Systems Engineering National University of Defense Technology 

出 版 物：《Journal of Integrative Agriculture》 (农业科学学报(英文版))

年 卷 期：2024年第10期

页      面：3554-3575页

核心收录：

学科分类：09[农学] 0904[农学-植物保护] 090401[农学-植物病理学] 090402[农学-农业昆虫与害虫防治] 

基　　金：supported by the Changsha Municipal Natural Science Foundation, China (kq2014160) in part by the Key Projects of Department of Education of Hunan Province, China (21A0179) the Hunan Key Laboratory of Intelligent Logistics Technology, China (2019TP1015) the National Natural Science Foundation of China (61902436) 

摘      要：Banana is a significant crop, and three banana leaf diseases, including Sigatoka, Cordana and Pestalotiopsis, have the potential to have a serious impact on banana production. Existing studies are insufficient to provide a reliable method for accurately identifying banana leaf diseases. Therefore, this paper proposes a novel method to identify banana leaf diseases. First, a new algorithm called K-scale VisuShrink algorithm(KVA) is proposed to denoise banana leaf images. The proposed algorithm introduces a new decomposition scale K based on the semi-soft and middle course thresholds, the ideal threshold solution is obtained and substituted with the newly established threshold function to obtain a less noisy banana leaf image. Then, this paper proposes a novel network for image identification called Ghost ResNeSt-Attention RReLU-Swish Net(GR-ARNet) based on Resnet50. In this, the Ghost Module is implemented to improve the network s effectiveness in extracting deep feature information on banana leaf diseases and the identification speed; the ResNeSt Module adjusts the weight of each channel, increasing the ability of banana disease feature extraction and effectively reducing the error rate of similar disease identification;the model s computational speed is increased using the hybrid activation function of RReLU and Swish. Our model achieves an average accuracy of 96.98% and a precision of 89.31% applied to 13,021 images, demonstrating that the proposed method can effectively identify banana leaf diseases.