The Research On Hyperspectral Characteristics Of Mikania Micrantha

The invasion of alien species has become an important issue in the study of global change because of its lack of natural enemies and strong competitiveness,as well as its capture of the living space of native organisms,which seriously threatens the diversity of plants and organisms and the stability of ecological balance.The foundation and premise of scientific management and control of invasive pests lie in good understanding of in-time invasion scope of invasive organisms.Mikania micrantha is a highly dangerous alien invasive species and one of the most harmful forestry plants in China.It is of great significance to improve the monitoring of Mikania micrantha.In recent years,the rapid development of hyperspectral remote sensing technology has provided a new thought for the monitoring of Mikania micrantha.The study area in this paper is Zengcheng Forest Farm in Guangzhou,Guangdong Province.The hyperspectral remote sensing images of the study area are acquired by hyperspectral imager carried by unmanned aerial vehicle(UAV)and the data sources suitable for the study are obtained by preprocessing the images.The hyperspectral characteristics of Mikania micrantha are studied based on the preprocessed hyperspectral remote sensing images.The main contents of this paper are as follows:(1)Primary selection of hyperspectral characteristics of Mikania micrantha:the hyperspectral characteristics of Mikania micrantha in the study area are selected by Optimal index factor(OIF),Adaptive Band Selection(ABS),OBI,Auto-Subspace partition(ASP)+ABS based on the study area,ASP+ABS and SCP based on Mikania micrantha.The combination of six wavebands is therefore obtained.The results show that ASP+ABS based on the study area and ASP+ABS and Spectral characteristic parameters(SCP)based on Mikania micrantha can better reflect the spectral characteristics of Mikania micrantha.(2)Primary selection test of hyperspectral characteristics of Mikania micrantha:Three new images were generated by band combination based on three methods.Support vector machine and spectral angle mapping were used to classify Mikania micrantha in the study area.The response of three band combinations to hyperspectral characteristics of Mikania micrantha was evaluated with the accuracy of classification results.It has been proved by tests that the band combination of 18 bands selected by the ASP+ABS band selection method in the Mikania micrantha area can better reflect the hyperspectral characteristics of Mikania micrantha.The support vector machine is more suitable for Mikania micrantha information extraction based on hyperspectral image than the spectral angle mapping.(3)Study on the hyperspectral texture features of Mikania micrantha:Four texture features of Mikania micrantha were extracted based on the gray level co-occurrence matrix:angular second moment,contrast,correlation,entropy.The corresponding results of the four texture features on Mikania micrantha were evaluated by the classification results of the support vector machine.The hyperspectral features of Mikania micrantha under the characteristic band of angular second moment are more prominent.It is found that the spectral response characteristics of Mikania micrantha are better than the texture features,and the texture features can play an active role in extracting information based on spectral features of Mikania micrantha.(4)Optimization of the hyperspectral characteristics of Mikania micrantha:the 18 bands of the primary selection of Mikania micrantha are sorted by secondary ABS,the first band is sorted for support vector machine classification,and then the prior bands are added one by one in sequence.The accuracy of the results is assessed by the hyperspectral characteristics of Mikania micrantha.The experimental results show that the hyperspectral features of the first 17 bands and angular second moment of Mikania micrantha after secondary ABS sorting are the most prominent,which is better than other combinations.Based on the high spatial resolution of the image of the study area and the physiological characteristics of large-area coverage on the image of Mikania micrantha,this study proposes a new strategy for characteristic band selection,and crops the occurrence area of the target feature on the image.The appropriate band is selected by the information-based band selection to highlight the hyperspectral characteristics of the target feature.In this study,18 bands are selected by the ASP+ABS band selection method in the invasive area of Mikania micrantha,and the hyperspectral characteristics of Mikania micrantha are well represented in both classification methods.