期刊名称:Indian Journal of Computer Science and Engineering
印刷版ISSN:2231-3850
电子版ISSN:0976-5166
出版年度:2021
卷号:12
期号:2
页码:428-433
DOI:10.21817/indjcse/2021/v12i2/211202146
出版社:Engg Journals Publications
摘要:Non-imagery field spectroscopic data is very high dimensional and thereby provides appropriate information about the object. Conversely, the high dimension of field spectroscopic data causes a very strong correlation between adjacent wavelengths and heavy redundancy in the information. This experimentation was performed to study the spectral behavior of healthy and unhealthy Sesame leaves on the electromagnetic spectrum (EMS). Three types of sesame leaves viz. healthy(H), damaged level1(LI), and damaged level2(L2) were selected and Non-imagery hyperspectral measurements of healthy and infected leaves were collected in the Visible and Near-infrared region (Vis-NIR ), in the spectral range of 400nm-1050nm. The VNIR region was divided into three sub-regions i) Visible(Vis):400nm-680nm ii) Red Edge(RE):681nm-781nm and iii) Near-Infrared(NIR):782nm-1050nm. ReliefF supervised attribute selection with ranker method was used as filters for dimensionality reduction, further Random Forest and J48(C 4.5) machine learning classifiers were applied on all the wavelengths in the regions as well as the first 20 wavelengths. The maximum classification accuracy was obtained in the visible region and it is revealed that distinct wavelengths affecting the health of the crop were found in the Visible region than RE and NIR regions. The classification accuracy of the Random Forest classifier was found better than the J48 classifier.
关键词:Hyperspectral leaf reflectance; wavelength selection; ReliefF; J48 ; Random Forest.