Title | ||
---|---|---|
Z-Score distance: A spectral matching technique for automatic class labelling in unsupervised classification |
Abstract | ||
---|---|---|
The paper presents a post-classification tool that automatically labels classes in classified imagery by matching their spectral characteristics to reference spectra. Unlike the Spectral Angle Mapper (SAM) and other spectral matching classifiers, it labels clusters of pixels rather than individual pixels. This new method can be used to label or re-label classes generated by any existing classifier, either supervised or unsupervised. In other words, it can be used in conjunction with existing classification approaches or as a part of an ensemble classifier. A Landsat 5 TM image of an agricultural area was used for performance assessment. The spectral signatures (reference spectra) were extracted from a hyperspectral Hyperion data set. The technique produced a map of higher accuracy (51%) in comparison to maps produced by manual class labeling (40% to 45% accuracy, depending on the analyst); it also outperformed the SAM classifier (39%), but underperformed in comparison to the Maximum Likelihood classification (53% to 63% depending on the analyst). |
Year | DOI | Venue |
---|---|---|
2014 | 10.1109/IGARSS.2014.6946801 | Geoscience and Remote Sensing Symposium |
Keywords | Field | DocType |
geophysical image processing,hyperspectral imaging,image classification,maximum likelihood estimation,remote sensing,Landsat 5 TM image,Spectral Angle Mapper classifier,Z-Score distance,agricultural area,automatic class labelling,classification approaches,classified imagery,ensemble classifier,hyperspectral Hyperion data set,manual class labeling accuracy,maximum likelihood classification,performance assessment,pixel clusters,post-classification tool,reference spectra,spectral characteristics,spectral matching classifiers,spectral matching technique,spectral signatures,unsupervised classification,automation,class labelling,spectral library,unsupervised classification | Computer vision,Pattern recognition,Computer science,Remote sensing,Hyperspectral imaging,Standard score,Pixel,Artificial intelligence,Classifier (linguistics),Spectral matching,Spectral signature,Maximum likelihood classification | Conference |
ISSN | Citations | PageRank |
2153-6996 | 0 | 0.34 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Ilia Parshakov | 1 | 0 | 0.34 |
Craig A. Coburn | 2 | 1 | 2.17 |
Karl Staenz | 3 | 115 | 23.05 |