Goals:
The main goals of this lab is to introduce all students to the basic mathematics behind photogrammetric correction. The lab is designed to to train us on photographic scaling, area calculations and calculating relief displacement. From these skills all students will be introduced to stereoscopy and orthorectification both of which are essential skills any remote sensing technician needs to know. We were tasked with three parts to this lab. Pat one introduced us to scales and relief displacement. Part two introduced us to stereoscopy, technique used for enhancing the illusion of depth in an image. Part three of the lab was introducing us to the process of orthorectification, process of removing
Methods:
Part one:Part one of the lab consisted of us students grasping the concepts of image scaling and correction of relief displacement for an image. In order to understand image scaling we were tasked with two objectives calculate the representative fraction of an image using known ground measurements against image measurements. Second was to find the representative fraction using focal camera length and elevation of capture. For the first process one must known the true distance or length of a known object on the ground. Then with in the captured image you will measure the same object or distance. In our case we measured a portion of interstate highway 94. We found that the image distance was 2.7 inches and the ground distance was 105,869 inches this gave us a representative fraction scale of 1:39,000. For the second calculation we knew the height the object was captured and the focal length so we where able to use the calculation S= f/ H-h or scale is equal to focal length divided by Height of camera minus the height of the terrain elevation. This calculation also generated a representative fraction of 1:39,000 supporting the previous claim and double checking our answers to the right scale.
The second section of part one was to introduce us to a tool that is used often within remote sensing applications. The tool we were introduced to was the area/parameter measuring tool. Section two of part one objective was to measure and collect both the area and parameter of dells pond located just south west of the University of Wisconsin Eau Claire. To accomplish this tool we open up a local image of the Eau Calire area and selected the measure perimeters and areas digitizing tool. Then we needed to digitize the area of interest ( in this case dells pond). After we had finished our digitizing a print out of the area and perimeter would show up. One can then change the units to what ever they desire to obtain the output they want. Section three of part one introduced us to relief displacement and how to calculate the total amount of displacement on an object. We accomplished this task by looking at an image taken in 200 of the upper half of UW Eau Calire. With in this image we found a tall industrial smoke stack that was not corrected yet for relief displacement for it was leaning away from the principle point. To calculate how much we needed to move this we had to obtain two measurements. The first measurement was the height of the stack from the base to the very top (.5 inches). Second we needed to find the distance the object was from the principle point in this case it was 8.5 inches away. Last we where given that the actual height of the object which was 3980 ft. Then using the calculation for relief displacement D= h*r/H
h = height of object in image
r = distance from principle point
H = true height of object
We found that the smoke stack was leaning .29 inches away from the principle point. From this information we were then able to correct the image and shift the stack back to its proper placement.
Part two:
 |
| Figure 1: DSM stereoscopic image produced in Erdas of Eau Claire area |
Part two of lab seven introduced us to them few techniques that are commonly used within stereoscopy ( or altering an image to display elevation changes). We would produce two final images from this method one would utilize a digital elevation model which would apply all first return from Lidar dataset and the second image would utilize a digital surface model which was obtained from the last return points of a lidar dataset. The procedure for the model where both the same the only changes where the second image being used either the DSM or the DEM.
To accomplish this we first needed to input the image we wanted to stereoscope and the image used for the model. For the first image it was the ec_city image and the ec_dem image. From here we open the terrain- anaglyph tool and anaglyph generator. Within the anaglyph generator menu we then could select the image we would stereoscope in this case was the eau Claire city photo and the DEM model photo to generate the elevation changes. The output image was then generated and although displayed in figure 1 you can only see the elevation changes with access to 3-d glasses.
The second image was generated using the same steps however, instead of using the DEM for the model image we used the DSM. The output is shown in figure 2 but as stated before without 3-d glasses you will not be able to see the elevation changes.
 |
| Figure 2: DEM stereoscopic image produced in ErDas of Eau Claire Area |
Part three:
Orthorectification is a skill set that every remote sensor should be able to accomplish. It is the process of geometrically correcting an image for all three coordinates (x,y,z). The process although challenging is very helpful for it allows any user to be able to correct their image for analysis in any dimension. For this portion of the lab we had multiple steps to accomplish our goal of a rectified image output. First we needed to start a new project within the program. Then by selecting and referencing a horizontal image with GCP's we would be able to correct a second image and use tie points to bring them together. Once the tie points where placed we then could triangulate the images and orthorectify them. This would produce a final image that was corrected in all three dimensional coordinates.
The process of this was as follows. First import a image to be used for horizontal reference. In this case we used the spot_pan image and a spot_panb image both of which where pancromanic band images that had very high spatial resolution making the GCPs easy to collect. After setting the model category to polynomial based push broom we then needed to chose a projection of the images this case was Nad27(CONUS) with a UTM zone of 11. Now we would be able to collect GPC of the images and reference them to the correct geospatial location. On the left of the collection screen we had our reference image which was the xs_ortho and the right was the image we where trying to alter this case was the spot_pan. GCP collection window can be seen in figure 3. The collection of all points where exactly the same other than after the first three points the system will automatically place the reference point for you but you can alter the location to lower your overall root mean square to aim for the ideal of being below .5. The collection would start by adding a new point and then hitting the collection button (referenced as a crosshairs in the upper right corner).
 |
| Figure 3: collection window for GCP and orthorectification |
Once you collected the twelve GCP the horizontal reference is now complete and you will move on to set the vertical reference. To set the vertical reference you need to select the reset vertical reference located in the tools part in the upper right portion of the point measurement window. From there we used the DEM option and selected the pal_springs_dem model for our reference. Once selected the z field was now filled in for all of our GCP points we previously collected. After changing the type and usage to full and control the points were now corrected.
The next step was to tie the spot pan image to spot pan b image. To do this we will follow similar steps to the GCP process in the previous steps but will be looking for all locations that overlap in the two images. Once we have collected the tie points we then are able to run the orthorectification of the two images. The output will generate a tied image that is corrected for all three dimensional coordinates. Figure 4
 |
| Figure 4: orthorectified output image |
Results:
The outputs from this lab produced many useful techniques and skills. From this lab we where able to better understand scaling of an image using both ground and image comparison and image referenced scaling. The second portion of the lab introduced us to the application of area and parameter analysis. This tool set is very useful within remote sensing application for it allows users to be able to measure any area of interest or study are they are trying to analyze.
Last we produced both stereoscopic images which can help user to see elevation changes in the areas surface and learned the process of orthorectification which allows users to be able to geometrically correct any image in all three dimensions.
The outputs of the stereoscopy were very interesting for the DSM and DEM model had very different outcomes. The DEM model produced a very aggressive output of the landscape for it used the first returns of the lidar dataset making the elevation easier to see but not a true representation of the actual landscape. Whereas the DSM, was more subtle in its elevation changes making it harder to see some of the areas but a more accurate representation of the natural landscape.
The output of the orthorectification was a little disappointing for even after all of the work and getting my RMS under .5 the images did not match up very well. They where mostly spatially corrected but around critical areas like rivers and roads the images did not fully line up correctly. Although this time I was not fully successful in the rectification I know that the next time I will need to run a model like this I will be more confident in my ability to produce a useful output image.
sources:
National Agriculture Imagery Program (NAIP) images are from United States Department of
Agriculture, 2005.
Digital Elevation Model (DEM) for Eau Claire, WI is from United States Department of
Agriculture Natural Resources Conservation Service, 2010.
Lidar-derived surface model (DSM) for sections of Eau Claire and Chippewa are from Eau
Claire County and Chippewa County governments respectively.
Spot satellite images are from Erdas Imagine, 2009.
Digital elevation model (DEM) for Palm Spring, CA is from Erdas Imagine, 2009.
No comments:
Post a Comment