Ground-Level Classification of a Coral Reef Using a Hyperspectral Camera

Remote Sensing

Tamir Caras²  Arnon Karnieli¹  Eyal Ben-Dor² 

[1] The Remote Sensing Laboratory, Jacob Blaustein Institutes for Desert Research, Ben-Gurion University of the Negev, Sede Boker Campus, Midreshet, Ben-Gurion 84990, Israel; E-Mail

关键词: coral reef;    spectroscopy;    classification;    remote sensing;    hyperspectral;    PLS-DA;   

DOI  :  10.3390/rs70607521

来源: mdpi

【 摘 要 】

Especially in the remote sensing context, thematic classification is a desired product for coral reef surveys. This study presents a novel statistical-based image classification approach, namely Partial Least Square Discriminant Analysis (PLS-DA), capable of doing so. Three classification models were built and implemented for the images while the fourth was a combination of spectra from all three images together. The classification was optimised by using pre-processing transformations (PPTs) and post-classification low-pass filtering. Despite the fact that the images were acquired under different conditions and quality, the best classification model was achieved by combining spectral training samples from three images (accuracy 0.63 for all classes). PPTs improved the classification accuracy by 5%–15% and post-classification treatments further increased the final accuracy by 10%–20%. The fourth classification model was the most accurate one, suggesting that combining spectra from differ conditions improves thematic classification. Despite some limitations, available aerial sensors already provide an opportunity to implement the described classification and mark the next investigation step. Nonetheless, the findings of this study are relevant both to the field of remote sensing in general and to the niche of coral reef spectroscopy.

【 授权许可】

CC BY   
© 2015 by the authors; licensee MDPI, Basel, Switzerland.