| Title | Classification of a Hyperspectral Agricultural Data Set Using Band Moments for Reduction of the Spectral Dimensionality |
| |
| Author | Staenz, K |
| Source | Canadian Journal of Remote Sensing 22, 3, 1996 p. 248-257, https://doi.org/10.1080/07038992.1996.10855180 |
| Year | 1996 |
| Alt Series | Earth Sciences Sector, Contribution Series 20041502 |
| Publisher | Informa UK Limited |
| Document | serial |
| Lang. | English |
| Media | paper; on-line; digital |
| File format | pdf |
| Released | 2014 07 31 |
| Abstract | The band-moment analysis technique for the reduction of spectral dimensionality of a hyperspectral data set acquired with the Programmable Multispectral Imager (PMI) has been used in combination with
conventional classifiers for labelling agricultural targets. In particular, the impact of this data-reduction procedure on the classification results has been studied with respect to a benchmark data set that includes eight carefully selected bands.
To maximize the classification results, a variety of pre-processing steps were carried out prior to data-reduction/classification addressing problems due to the sensor, calibration, data transcription, and atmosphere. The classification results
indicate that overall accuracies of slightly over 80% could be achieved with the proposed procedure. These accuracy levels were generally achieved with the logistic classifier and partially with the maximum likelihood classifier using six out of
eight features generated with the automated data-reduction procedure. It was also found that these classification results are generally slightly lower than those derived from the benchmark data set. In general, the overall classification performance
is acceptable considering the small size of the fields in combination with the technical constraints of the sensor system. |
| GEOSCAN ID | 218304 |
|
|