•  
  •  
 

Corresponding Author

Simbarashe Jombo

Abstract

Mapping of vegetation at the species level using hyperspectral satellite data can be effective and accurate because of its high spectral and spatial resolutions that can detect detailed information of a target object. Its wide application, however, not only is restricted by its high cost and large data storage requirements, but its processing is also complicated by challenges of what is known as the Hughes effect. The Hughes effect is where classification accuracy decreases once the number of features or wavelengths passes a certain limit. This study aimed to explore the potential of feature selection methods in the classification of urban trees using field hyperspectral data. We identified the best feature selection method of key wavelengths that respond to the target urban tree species for effective and accurate classification. The study compared the effectiveness of Principal Component Analysis Discriminant Analysis (PCA-DA), Partial Least Squares Discriminant Analysis (PLS-DA) and Guided Regularized Random Forest (GRRF) in feature selection of the key wavelengths for classification of urban trees. The classification performance of Random Forest (RF) and Support Vector Machines (SVM) algorithms were also compared to determine the importance of the key wavelengths selected for the detection of the target urban trees. The feature selection methods managed to reduce the high dimensionality of the hyperspectral data. Both the PCA-DA and PLS-DA selected 10 wavelengths and the GRRF algorithm selected 13 wavelengths from the entire dataset (n = 1523). Most of the key wavelengths were from the short-wave infrared region (1300-2500 nm). SVM outperformed RF in classifying the key wavelengths selected by the feature selection methods. The SVM classifier produced overall accuracy values of 95.3%, 93.3% and 86% using the GRRF, PLS-DA and PCA-DA techniques, respectively, whereas those for the RF classifier were 88.7%, 72% and 56.8%, respectively.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.