Kathmandu Valley with optical images - General overview - Continued
|  | Merged classification of Kathmandu | |
Rectification
During the previous exercise we produced several classifications, both manual and digital. One of our goals was to overlay a hand-drawn, manual classification on a digitally supervised classification. Both images, the basic raw material, are saved in your Kathmandu folder. Unfortunately the images do not have the same orientation and scale. While scanning the classified image, you changed the orientation and geometry and the scale.
Therefore, a rectification of the scanned image is necessary. A rectification is the equalisation of one image to another in terms of its scale and orientation.
In this case, the hand-drawn map has to be equalised to the digitally classified map 'Kathmandu_2001_maximlike.tif'. This is the time when the "GCP.shp" registration marks come into use.
Choose File>Open. A dialogue box will pop up. Choose the folder Kathmandu and select image Kathmandu_2001_maximlike.tif. Open the HLC_Kathmandu.tif image, too.
Select (activate) the Kathmandu_2001_maximlike.tif image. Choose GIS and Open Theme GCP.shp. A pop-up menu appears. Define the specific projection parameters. All the GEO-TIFFs in this case study have the same projection:
- Transformation Method: UTM
- Datum Name: WGS84
- Ellipsoid Name: WGS84
- UTM Zone: 45
- UTM Hemisphere: North
The registration marks reappear in the same position as drawn in the false-colour image 4,5,3. Now we have the marks on both images and are able to rectify the manual classification.
Choose Tools>Registration>Image to Image and select Kathmandu_2001_maximlike.tif as Master Image and HLC_Kathmandu.tif as Slave Image.
Now we have to define reference points (ground control points / GCPs) in both images. Select (activate) the Kathmandu_2001_maximlike.tif image. Set the first point at the registration mark at the upper left corner, switch to image HLC_Kathmandu.tif, look for the same position and set the first point there. Both coordinates appear in the
GCPs Collection. Click Add Point to fix the positions for the first GCP. Go back to Kathmandu_2001_maximlike.tif and select the second point at the upper middle registration mark. Switch to HLC_Kathmandu.tif and select the same position. Do the same with all the other registration marks. Show List opens a table where all selected GCPs are registered.
Do not try to select the GCPs by choosing the same cursor positions. The positions in both images have to be different, because the scanned image is not in the correct orientation. That is why it has to be rectified.
Choose Options>Warp Image from the GCPs Collection. A new image appears. It is the scanned hand-drawn classification of Kathmandu in the same scale and orientation as all the other GEO-TIFFs provided in this case study.
Save the rectified image as HLC_Kathmandu_REF.tif in the Kathmandu folder.
As it now has the same grid structure and scale, the image HLC_Kathmandu_REF.tif can be used in the same way as
the GEO-TIFF dataset of Kathmandu. That means we can use the image to take measurements, and we can merge it with the satellite images and classifications of Kathmandu.
Choose File>Open. A dialogue box will pop up. Choose the folder Kathmandu and select image Kathmandu_2001_maximlike.tif. Open the image HLCKathmandu_REF.tif, too.
Choose Image>Transparent Overlay. Select Kathmandu_2001_maximlike.tif as BackGround Image and HLC_Kathmandu_REF.tif as ForeGround Image and bring the Opacity slide to 60%.
Observe your manual classification in comparison with the supervised classification. Point out where it corresponds and where it does not correspond. Which seems to be more accurate? Use your city map and additional material to answer this question.
| | | Digitalised river system on PROBA image |
Digitalisation
Did you notice the missing class of 'Rivers' in the supervised classification?
In fact, there is hardly any way of
classifying this feature, due to the resolution of the image. But you were able to draw the rivers in your manual classification. And you rectified the image.
Therefore you can digitalise the river network and merge it with the Landsat data and the digital classifications.
It is a useful and common method to merge pixel and vector data. In this way, particularly tiny and/or additional
information can be shown on the satellite images.
In the following exercises we digitalise the river network in the image HLC_Kathmandu_REF.tif and merge it with
the supervised classification Kathmandu_2001_maximlike.tif, the natural-colour image Kathmandu_Landsat_2001_Band_321.tif and a high resolution Proba image.
Open the image HLC_Kathmandu_REF.tif and choose GIS. A pop-up menu opens. Choose File>New Theme, name the theme 'Rivers' and select Polyline. To start drawing, select Edit>Start Edit.
Digitalise the rivers you classified. When you are done, choose Edit>Stop Edit to stop drawing.
To set up the river style, choose Edit>Properties and set the colour and the thicknes of the vectors. Save Theme 'Rivers' to your Kathmandu folder.
Start another theme and name it 'Airport'. Digitalise the airport the same way you did for the rivers, and save the theme as 'Airport'.
Open the image Kathmandu_proba12321_utm45_20040505.tif. Choose GIS and Open Theme Rivers.shp, specify the WGS84,
UTM Zone 45 projection and set the colour and thickness options.
How accurate are the river vectors? Do they fit with the high resolution image?
Open the image above. There is an error within the river network. Can you find it?
Check your river network very carefully.
If you find mistakes within the system, choose Edit>Start Edit and Edit>Edit Vertex or Move and correct the error(s). Than save the shapefile again.
Open another Proba image and check the rivers again. Correct all errors you can find by using all Proba high resolution images.
The more accurate the rivers are on a high resolution image, the more accurate they will be on a Landsat image.
To check that out, open the natural-colour image of Kathmandu and open the river and the airport shapefiles.
Do the rivers fit onto the image?
Are there still errors? If yes - where and why?
| 
