Remote Sensing of the Environment : Rio Grande Delta Marshes

Classification of Tiocano CASI data

The Tio Cano Marsh was fortunate to have hyperspectral data collected by both the CASI and HYMAP sensors, which allows for direct comparisons of instruments and results. First we'll consider the CASI data. Figures 7 and 8 show the classified images obtained from the Best Basis and HTC algorithms respectively with the class colors given below.:

Field reconnaissance helped determine the classes for Tio Cano marsh, which are:

1. Trees (Dominated by Salix Nigra)
2. Eleocharis Sp.
3. Distichlis spicata
4. Typha domingensis
5. Mix Vegetation (Typha and Distichlis)
6. Water

Fig 1. Tio Cano (CASI) - Hierarchical Tree Classifier

Fig 2. Tio Cano (CASI) - Best Basis Classifier

Table 1: Confusion matrix (%), Best Basis Classification

Class	1	2	3	4	5	6
1	96.9863	0.885	0	1.8923	0	0
2	0.8219	99.115	0	0	0	0
3	0	0	81.944	0	24.4604	0
4	2.1918	0	0	94.905	3.5971	0
5	0	0	18.056	3.2023	71.9424	0
6	0	0	0	0	0	100

Overall accuracy: 90.82%

Table 2: Confusion Matrix (%), Hierarchical Classification

Class	1	2	3	4	5	6
1	98.6301	2.6549	0	0.8734	0	0
2	0.137	96.46	0	0	0	0
3	0	0	97.222	0	4.3165	0
4	1.2329	0.885	0	94.614	2.1583	0
5	0	0	2.7778	4.5124	93.5252	0
6	0	0	0	0	0	100

Overall accuracy: 96.74%

The Hierarchical classifier again gives a better accuracy, the most important band is obtained as 8 which has a wavelength in the green light's wavelength range. We obtain a better representation of the Eleocharis sp. class from the hierarchical classifier, moreover this technique gives a much better discrimination between the Distichlis and the Mix Vegetation (Typha and Distichlis) classes which also obvious from the confusion matrix. The Best basis classifier seems to over classify the Eleocharis class in this dataset.

Figure 3. Best basis weighted bands for Tio Cano CASI data

Last Modified: Wed Apr 14, 1999
CSR/TSGC Team Web