GEOM30009: Imaging the Environment
1. Image interpretation and electromagnetic radiation:
What are the spectral reflectance curves (4 points)? and how are they useful for identifying different features in multispectral images? Explain your answer by an example (6 points).
2. Human vision and image formation:
What are the different aspects of image resolution (2 points)? Provide a proper definition of the different aspects of image resolution (8 points).
3. Imaging sensors:
Most very high resolution satellite sensors capture images in one panchromatic band and several spectral bands. Provide a comparison of a panchromatic image and a spectral image in terms of spatial resolution (5 points), and spectral resolution (5 points). Explain your answer.
4. Image processing:
A satellite sensor has captured an image that is oflow contrast. Suggest an image processing method that can be applied to improve the image contrast (4 points). Explain in detail how this method works (6 points). Draw a diagram if it helps your explanation.
5. Information extraction:
A common image segmentation method is Region growing.
a) Describe the region growing algorithm for image segmentation (4 points).
b) Give an example of homogeneity criteria that can be used in region growing (2 points).
c) Explain the over-segmentation and under-segmentation errors that might occur during segmentation (4 points).
6. Image classification:
The table below shows a confusion matrix resulting from a supervised classification. Calculate the following:
a) Omission error for each class. (3 points)
b) Commission error for each class. (3 points)
c) Overall classification accuracy. (4 points)
|
|
Ground truth
|
|
Classification
result
|
|
Water
|
Vegetation
|
Land
|
Total
|
|
Water
|
88
|
5
|
2
|
95
|
|
Vegetation
|
10
|
79
|
6
|
95
|
|
Land
|
2
|
16
|
92
|
110
|
|
Total
|
100
|
100
|
100
|
300
|
7. Image rectification:
A simple method for image rectification is resampling by interpolation.
a) Describe the resampling by interpolation method (4 points).
b) What are the common interpolation methods used in resampling (3 points)?
c) Which interpolation method produces better results and why (3 points)?
8. Orthorectification:
Suppose you are asked to produce an orthophoto from an aerial image taken by a camera with the following specifications:
- Focal length: 150 mm
- Frame. dimension: 200 x 200 mm
The following requirements are given for the orthophoto:
- Orthophoto scale: 1:5000
- Maximum tolerable radial error: 0.5 mm
You must choose between three DTMs with the following specifications:
- An SRTM global elevation model with a height accuracy of 16 m. This DTM is freely available.
- A photogrammetric DTM generated from aerial images with a height accuracy of 2 m. The cost of acquiring this DTM is moderate
- Lidar DTM generated from aerial lidar data with a height accuracy of 0.2 m. The cost of this acquiring this DTM is high.
Which of the above DTMs will you choose to generate an orthophoto that meets the user requirements and why? (10 points)
9. 2D measurement in images:
Suppose you are given the task of generating a 2D map of building boundaries from an aerial image captured over an urban area.
a) What process should be applied to the image before it can be used for mapping (2 points)?
b) What other source of data in addition to the image would you need to apply the process (2 point)?
c) What is the specific problem caused by occlusion in this process? (3 points)
d) How can this problem be overcome? (3 points)
10. 3D measurements:
Lidar is a widely used technology for direct acquisition of 3D elevation data. Describe the components of an airborne lidar system (4 points), and explain the type of measurements each component can record (6 points).