Making the Images
|
|
|
||
|
All of this was straighforward, except for the decision to select the histogram equalization option. The region depicted is relatively dry. Average annual rainfall is 30 cm (12 inches) to 50 cm (20 inches), with virually no rainfall in July when this image was acquired. (Close inspection of the scene reveals circular patterns that indicate rotary irrigation systems, for example 25 km north of Mt Shasta's peak at the eleven o'clock position.) Most of the scene is either entirely lacking in vegetation or supports vegetation that is not growing vigorously. Thus a conventional natural color composite would appear bluish, while a composite with histogram equalization appears more natural. Preparing the image with an outline of the Pleistocene debris flow was a little more complicated, mainly because the image available was not georeferenced and thus could not be co-registered with the satellite image. Co-registering these two images means to place one image on top the other so that the ground locations represented in the first image correspond to same the ground locations represented in the second image. To do this we need to use a common frame of reference, which might be one of the images, say the first. We would then line up all the locations of the second image with the corresponding locations in the first image. This is commonly called rubber-sheeting because one image is streched to fit the other. To do this operation it is not necessary for either of the images to be geo-referenced (coded with latitude and longitude or other geographical coordinates). Fortunately, the Landsat TM image was already georeferenced to the UTM-10N (Universal Transverse Mercator 10 North). How to Read UTM. At first, the difference in resolution of the two images gave cause to doubt that georeferencing might be successful. The image outlining the debris flow (the outline image) has only one row of pixels (picture elements) for every two rows in the Landsat image. While few features are clearly defined, about ten features appeared potentially useable: places where roads cross rivers, the peaks of Mt Shasta and Mt Shastina, and the dam at Lake Shastina. To produce a georeferenced DEBRIS.RST file and its accompanying DEBRIS.RDC file the coordinates of four features in both images would be sufficient to run IDRISI's RESAMPLE routine. In fact, four of the ten ground control points (GCPs) selected had to be discarded because they did not fit the geometry computed for rubber-sheeting. With six GCPs the average error reported by IDRISI was 60 meters, perhaps not good enough for serious research, but acceptable for illustration purposes. Locating suitable gound control points and recording their locations was the difficult part of the task.
Next, the outline image had to be converted into a vector image, an image which IDRISI recognizes as forming lines that can be superimposed on the Landsat image (which is in raster format). There may be a simpler way to do this, but the following method works: By way of reassurance, once an image has been georeferenced and/or co-registered with another image, sending it out of IDRISI for cosmetic surgery will not undo all the work of resampling (rubber-sheeting) so long as one rule is followed:
Getting the image back into IDRISI is simple. Import the image as DEBRIS_BOOL.RST using the BMPIDRIS option. The reference system will be plane, as recorded in the DEBRIS_BOOL.RDC file. The information needed for georeferencing is in the DEBRIS.RDC file, accompanying the DEBRIS.RST file. Close the DEBRIS_BOOL file. Copy and rename DEBRIS.RDC to DEBRIS_BOOL.RDC. When IDRISI next displays the DEBRIS_BOOL image Composer>Layer Properties will indicate UTM-10N as the reference system.
|
Cleaning up the ImageBetween exporting the image from IDRISI and reimporting it, the image needs to be cleaned up. As shown in the images to the right, the detail must be removed from the original image to produce a black and while image, called Boolean because the image has only two values. There are many ways to obtain the Boolean image. Most graphics programs have tools such as magic wand and paint bucket, a paint brush for the small details, adjustment to maximum contrast, and a speckle–removing filter. The choice of white as foreground and black as background is immaterial, since this can be changed in IDRISI. |
 |
|
Making a Vector Image of the AvalancheIDRISI has the facility for adding a vector layer on top of a raster layer. A line or polygon vector image of the avalanche outline can be overlain on the Landsat raster image as in the image at the top of this page. A polygon-vector image can be produced directly using IDRIS's POLYVEC routine. To produce a line-vector image an indirect method is needed. Use BUFFER to produce a one-pixel wide line outside the outline. Use RECLASS to make a negative image of the outline image, then use BUFFER to produce a one-pixel wide line inside the outline. Add the two images. This gives a two-pixel wide line straddling the outline of the image. Now IDRISI can run a vector line down the center of this line using LINEVEC. Symbol Workshop can be used to make a symbol file that will display red lines in the image, otherwise the lines will be black. Note that this outline has islands, so there are several lines to color.
Displaying the Raster and Vector Images TogetherAfter displaying the natural-color composite image the line-vector image is added using COMPOSER and the appropriate symbol file is chosen to obtain red instead of black lines.
|
Return to LocationGo to Cascade Range menu |