Lab 9 Answer Sheet

Classification

Name 1 Name 2

1. What is the minimum number of pixels you should choose to use as training sites for any one feature type (e.g., water)?

2. Based on a visual examination of the true color composite above, which features do you think might be difficult to separate based on their spectral reflectances?

3. Move the cursor to some of the major categories that we are trying to separate (trees, willows, wood, etc.). Which category or categories (if any) is the sage most likely to be confused with from a spectral perspective? Why do you think this is the case?

4. What is the potential disadvantage of using the feature space to define the training sites rather than defining an AOI on the image?

5. Which approach to collecting an AOI did you use to create a spectral signature file for willows? Explain why you chose this approach to “find” willow locations.

6. Is there more or less overlap when you set the standard deviation to a lower value (e.g., 1.0 instead of 2.0). Why is this (i.e., what is the standard deviation representing)?

7. What features overlap (note that the colors of the ellipses correspond to the colors you have assigned each category)? Why do you think these categories overlap?

8. What is the largest standard deviation value you can choose that avoids any overlap.

9. Use the table below to describe the shape and range of the values for your training sets.

Band	Water		Wood
	Uni-, bi- or multi-modal	Range of values	Uni-, bi- or multi-modal	Range of values
1
2
3
4
5
6

7. Are there legitimate variations in wood reflectance that might lead one to accept a training set with bimodal or multi-modal distributions? If so, what features of wood might be driving this multi-modal reflectance?

8. Which features (if any) were most confused? Do the features that were confused fit with your answers to 2, 3 and 7 above?

9. At a first glance, which category is classified the best. In terms of spectral characteristics (reflectance, absorption) why is this the case?

10. Based on your visual analysis, which three categories were more commonly mixed up?

11. Consider the full range of ways in which you can control the classification process, ranging from the imagery you use (think resolution), to ground truthing, to selection of training sites, to image restoration, to image enhancement and classification approaches. What are three actions you might take to improve the classification accuracy for this particular site? Explain why you think those approaches might work for this site.