https://scholars.lib.ntu.edu.tw/handle/123456789/639771
標題: | Application of Hyperspectral Imaging for Identification of Types and Levels of Pest Damage on Tea Leaves | 作者: | Lin, Yu Hung Lin, Xiu Rui SHIH-FANG CHEN |
關鍵字: | hyperspectral imaging | machine learning | plant disease | tea foliar lesion | 公開日期: | 1-一月-2023 | 來源出版物: | 2023 ASABE Annual International Meeting | 摘要: | Tea is one of the most popular drinks worldwide, and the quality of tea shoots determines the quality. Many factors, such as leaf tenderness and damage level by pests and diseases, influence the quality of tea shoots. Visible images can be applied to judge the leaf tenderness by colors and inspect the pest damage by observable symptoms. However, it is not easy to determine the level of pest damage when the symptoms are indistinct. Hyperspectral imaging (HSI) is a nondestructive technique and has the advantage for detecting the unapparent symptoms of tea shoots by spectral information. Hence, the study aims to identify the levels of pest damage on tea shoots using HSI. Tea mosquito (Helepeltis fasciaticollis Poppius) and yellow thrips (Scirtothrips dorsalis Hood) are two of the most common pests occurring in tea plantations, and they were selected as the target pests in this study. According to the severity of pest damage, the levels of the foliar lesion were categorized into three classes for each pest, with a total of seven categories, including healthy. The classification models were developed and compared by three machine learning (ML) and one deep learning method: random forest (RF), k-nearest neighbors algorithm, extreme learning machine, and one-dimensional convolutional neural network (1D-CNN). The RF model reached an accuracy of 71%; the 1D-CNN model achieved the best performance with 79% accuracy. The results showed that applying the 1D-CNN model on HSI could identify different levels of pest damage on fresh tea leaves. The next phase in this study will focus on performing the feature wavelength selection corresponding to different damage levels and types of pests. |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85183588874&doi=10.13031%2faim.202300398&partnerID=40&md5=6432335f018f230951360e6f2f222282 https://scholars.lib.ntu.edu.tw/handle/123456789/639771 |
ISBN: | 9781713885887 | DOI: | 10.13031/aim.202300398 |
顯示於: | 生物機電工程學系 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。