Remote Sensing of the Environment : Rio Grande Delta Marshes
Classification of Stover Point HYMAP data
Stover Point covers an area of estuarine marsh adjacent to the Laguna Madre and is part of the Laguna Atascosa National Wildlife Refuge. This area is a habitat for many migratory birds, and small mammals including bobcats and ocelots. The vegetation here consists of common high estuarine marsh species including Spartina spartinae, Borrichia frutescens, Monanthochloa littoralis, and Batis maritima. Adjacent to the resaca is an almost impregnable layer of dense shrubs and trees.
The classes determined for Stover Point are as follows:
- Water
- Spartina spartinae
- Batis maritima
- Mixed 1 (Borrichia frutescens, Spartina spartinae and Monanthocloa littoralis)
- Sand Flats (Bare Soil)
- Borrichia frutescens
- Trees
- Dense Brush
- Mixed 2 (Borrichia frutescens and Spartina spartinae)
HYMAP data is collected in 126 bands with wavelengths ranging from 0.441µm to 2.478µm. It contains much more infrared information than the CASI data. Figures 4 and 5 show the classified images obtained from the Best Basis and HTC algorithms respectively with the class colors given below. The overall importance of each band is given in Figure 6, Tables 3 and 4 show the classification accuracies with a confusion matrix:
Fig 1. Stover Point - Hierarchical Tree Classifier |
Fig 2. Stover Point - Best Basis Classifier |
Table 1: Confusion matrix(%), Best Basis Classification
Class |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
1 |
100 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
2 |
0 |
82.537 |
0 |
0.9973 |
0.0635 |
0 |
0 |
0 |
23.79 |
3 |
0 |
0 |
99.262 |
0.9973 |
0 |
0 |
0 |
0 |
0 |
4 |
0 |
0.956 |
0.738 |
95.467 |
0.381 |
1.1513 |
0 |
0.1104 |
2.3158 |
5 |
0 |
0.5417 |
0 |
0 |
99.4921 |
0 |
0 |
0 |
0 |
6 |
0 |
0.0637 |
0 |
1.4506 |
0 |
98.1908 |
0 |
0 |
1.8947 |
7 |
0 |
0 |
0 |
0 |
0 |
0 |
99.605 |
1.8764 |
0 |
8 |
0 |
0 |
0 |
0 |
0 |
0 |
0.3953 |
98.0132 |
0 |
9 |
0 |
15.902 |
0 |
1.0879 |
0.0635 |
0.6579 |
0 |
0 |
72.0 |
Overall accuracy: 93.84%
Table 2: Confusion Matrix(%), Hierarchical Classification
Class |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
1 |
100 |
0 |
0 |
0 |
0.0635 |
0 |
0 |
0 |
0 |
2 |
0 |
88.05 |
0 |
0 |
2.6667 |
0.3289 |
0 |
0 |
11.579 |
3 |
0 |
0 |
100 |
0 |
0.127 |
0 |
0 |
0 |
0 |
4 |
0 |
0.5736 |
0 |
98.459 |
0.9524 |
0.9868 |
0 |
0 |
1.6842 |
5 |
0 |
0.478 |
0 |
0.9973 |
95.8095 |
0 |
0 |
0 |
0 |
6 |
0 |
0.0319 |
0 |
0.544 |
0 |
98.0263 |
0 |
0 |
5.0526 |
7 |
0 |
0 |
0 |
0 |
0.0635 |
0 |
100 |
1.3245 |
0 |
8 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
98.6755 |
0 |
9 |
0 |
10.867 |
0 |
0 |
0.3175 |
0.6579 |
0 |
0 |
81.684 |
Overall accuracy: 95.63%
In this case, we obtain a higher overall accuracy from the Hierarchical classification. The important bands for the Best Basis classification mostly turned to be between 40 and 60 (1.02 µm to 1.4 µm) which are infrared bands higher in the spectrum than the CASI data. The classified images look very similar to each other except that we see more tree pixels, a better defined Batis class, and an improved mixed 2 class in the HTC result.
Figure 3. Best basis weighted bands for Stover Point HYMAP data
Last Modified: Wed Apr 14, 1999
CSR/TSGC Team Web
|