Filling Voids
OverviewThe object of this operation is to generate elevation data for voids without altering the existing data. If elevation values for voids are estimated from the data at the edges of the voids, the average value generated for a void should be close to the average value had there been no void. (Testing this statement would require another project.)
|
Pre-processingBefore filling voids, an image containing only voids is needed. Using the DEM image for the cone base, RECLASS was used to change all elevation values to one (1). Then GROUP was used to number groups of pixels.After running GROUP, the background was number zero (Group number 0). RECLASS was used to convert the model area (Group number 1) to zero. There were 29 other groups, all voids. These were RECLASSED as one (1). The resulting Boolean image shows the voids as one (1) and the remaining pixels as zero (0). This Boolean image is the voids mask
Method for Filling VoidsSince the voids contain no data, the data must come from the areas peripheral and outside the voids. Using BUFFER, three rows of pixels were added to the groups of pixels in the Boolean image. This is shown in Panel 1, below.RECLASS was used to convert the buffer value to one (1), giving a Boolean image, shown in Panel 2. The resulting Boolean image used as a mask was multiplied by the DEM image for the cone base, giving the result shown in Panel 3. Three rows of elevations have been isolated from the DEM. Some of the voids have already been filled, (not shown). However, the larger voids still have holes. The next step is to grow the voids inward without growing the peripheral areas outward. First, FILTER MAXIMUM 3X3 was used to add rows of pixels to the inside and outside of the voids, as shown in Panel 4. The peripheries of voids were trimmed using the voids mask produced during pre-processing, as shown in Panel 5. (Actually, the mask selected the elevation values, converting the buffer pixels to zero since they are no longer needed.) BUFFER and FILTER MAXIMUM were used three times to fill all the voids.
|
Using the Generated DataThe DEM for the cone base was used as a mask to cover the generated data with OVERLAY. None of the original elevation data was altered. The resulting DEM has no voids.
Deficiencies in the MethodologyThis methodology would not produce good results under certain circumstances, such as voids at the bottom of a cliff at a lake- or sea-shore or for a hillside in RADAR shadow. Also, applying a wide buffer and FILTER MAXIMUM may be a sub-optimum way of using IDRISI. A buffer of one pixel and the MODE or MEDIAN filter might give better results. Each time FILTERing and trimming is applied, a patched DEM should be generated and a new voids mask created to avoid over-smoothing the elevations within the voids.
|
 |
|
|
Return to Project |