Lab 8: Spectral Signature Analysis

Introduction

In this lab, students will learn how to take, graph, and analyze spectral signatures from satellite imagery. This will build upon students priory knowledge of spectral reflectance of Earth's surface features. Twelve spectral signatures will be collected from a Landsat ETM+ image that covers a portion of western Wisconsin and Eastern Minnesota. Students will need to located each of these surface features: standing water, moving water, vegetation (forest), riparian vegetation, crops, urban grass, dry soil (uncultivated), moist soil (uncultivated), rock, asphalt highway, airport runway, and a concrete surface.

Methods

ERDAS IMAGINE 2013 will be used to capture and analyze the spectral signatures.

Figure 1: With the proper image opened in ERDAS, Lake Wissota was picked as the location for standing water. The Drawing tab was selected, as seen above, and polygon, near the left end of the toolbar, was chosen. A small polygon was drawn and selected, making it the active area of interest.

Figure 2: Next, the Signature Editor was opened by navigating to the Raster tab > Supervised > Signature Editor, as seen above.

Figure 3: With the polygon still selected, the Create New Signature(s) from AOI icon (looks like a bent arrow next to the plus sign) was clicked. This created a new entry into the black editor window. The name was changed to Standing Water. The same process continued for each feature until all twelve had been taken.

Figure 4: The signature mean plot for standing water. These graphs are generated by clicking the Display Mean Plot Window icon (looks like a zig-zag line) in the Signature Editor window. By holding shift and clicking to select multiple signatures and clicking the Switch Between Single and Multiple Signature Mode Icon (looks like 3 zig-zag lines) in the Signature Mean Plot window, any number of signatures can be plotted on the same graph.


Results

Figure 5: All twelve signatures plotted together to help visualize trends. It became apparent that three trends seemed to dominate the graph. These trends were broadly categorized as water, vegetation, and land.

Figure 6: Water features. Standing water has a higher reflectance across all spectral channels. This is most likely due to higher sediment or algae content in the standing water when compared to moving water.

Figure 7: Vegetation features. Crops and urban grass have higher reflectance in the visible red band and mid-IR channel when compared to forest vegetation and riparian vegetation. This suggests that crops and urban grasses are under more stress or are more unhealthy than the other types of vegetation.

Figure 8: Land features. Rock has the highest reflectance across all spectral channels. Dry soil has substantially higher reflectance in the mid-IR channel than moist soil. As seen in Figure 6, water has low reflectance in the mid-IR channel because it absorbs most of this radiation. As such, the greater the water content in the soil, the lower its reflectance will be, especially in the mid-IR channel.




Data Sources

UWEC Department of Geography and Anthropology

 

Lab 7: Photogrammetry

Introduction

In this lab, students will learn and experiment with multiple photogrammetric operations. Students will calculate photographic scale and relief displacement, use Erdas Imagine to measure area and perimeter, experiment with stereoscopy by crating an anaglyph, and perform orthorectification on satellite imagery.

Methods

Part 1: Scales, measurements and relief displacement

Photographic scale
Photographic scale, the relationship between the distance observed on a vertical aerial image and its corresponding distance in the real world, can be determined two different ways.

One way is to compare the size of objects measured in the real world with the size of the same objects measured on the image by use of this equation: Scale = pd/gd.  Where, pd is photo distance and gd is ground distance. To convert this fraction into a useful measure of scale, the numerator must be 1. Therefore, the numerator (photo distance) must be multiplied by itself. When working with fractions, what is done to the numerator must also by done to the denominator. After multiplying the numerator and the denominator by the numerator, a representative fraction results.

The other way to find photographic scale is to use the relationship between the focal length of the sensor lens and the flying height of the craft above ground level. This relationship is defined as: Scale = f/(H). Where, f is the focal length and H is the altitude above ground level. If needed, H can be broken down into (H' - h), where H' is the elevation above sea level and h is the elevation of terrain.

Relief displacement
When objects are located at elevations greater than or less than the elevation of the photographs principle point, displacement occurs. At greater elevations features will be displaced away from the principle point and at lower elevations features will be displaced towards the principle point. The extent of the displacement is determined by three factors: 1 the greater the height of the feature, the larger the displacement, 2 the farther the feature is from the principle point, the greater the displacement, and 3 its inverse relationship to the height of the sensor above local datum.

Relief displacement (d) is defined as, d = (h*r)/H. Where h is the actual height of the feature, r is the radial distance from the top of the displaced object to the principal point, and H is the height of the sensor above the local datum.

Heads-up digitizing

Figure 1: Erdas Imagine can be used to measure perimeter and area of features on an image. This is done by clicking Measure in the Home tab of the main toolbar (Measure is between Metadata and Paste) to open the Measurement toolbar.

Figure 2: Changing the first button on the left from point to polygon (for area) and polyline (for perimeter), allowed for the proper measurements to be taken by tracing the outside of the desired feature. For this lab, a lagoon was measured.


Part 2: Stereoscopy
Stereoscopy is the science of creating depth in flat images. To analyze an area with a 3D perspective, Erdas Imagine can be used to create an anaglyph image.  An anaglyph is two of the same images superimposed onto each other in different colors, typically red and cyan. When seen through polaroid glasses, the flat image will appear to have depth.

Figure 3: To create an anaglyph in Erdas Imagine, click Terrain and then choose Anaglyph.

Figure 4: The Anaglyph Generation window. Here the input images are added and other parameters can be adjusted. For this lab the Exaggeration was changed from 1 to 2.


Part 3: Orthorectification
Orthorectification is the process of removing positional and elevation (x, y, and z) error from an aerial photograph or satellite image. For this lab, a previously orthorectified image will be used as a reference image to create a planimetrically true orthoimage from an image with substantial error.

Figure 5: To begin, the LPS Project Manager was opened by clicking the Toolbox tab on the main toolbar then choosing the leftmost icon LPS.

Figure 6: A new project must be created by clicking the leftmost icon, Create New Block File (looks like a blank sheet of paper). Here the geometric model options can be modified. For this lab SPOT imagery is being used so the settings were changed accordingly. OK was clicked.

Figure 7: Next, the Block Property Setup window appeared (left). Set... was chosen in the Horizontal section and the Projected chooser window appeared (right). Here the projection was set up accordingly. OK was clicked.

Figure 8: Next, highlight Images in the left column underneath the main toolbar of the LPS Project Manager. Then click the Add frame to the list icon (looks like a piece of paper with an arrow).

Figure 9: Now there is new information in the bottom portion of the window. The red color seen in multiple fields indicates that the sensor settings need to be verified. The Show and Edit Frame Properies icon was clicked, next to the Add frame to the list icon, and Edit... was chosen near the bottom of the window. The Sensor Information window appeared, OK was clicked, and OK was clicked again in the Frame Editor.

Figure 10: The Int field changed to green indicating that the sensor has been verified and the internal orientation information has been supplied.


Next, ground control points (GCPs) can be established. The Start point measurement tool (looks like crosshairs) was clicked and the radio button for Class Point Measurement Tool was filled. Ok was clicked.

Figure 11: The resulting window. This image was taken from the Lab instructions.
A: Main View
B:Tool Palette
C: Detail View
D: Reference Cell Array
E: File Cell Array
F: Reset Horizontal reference source icon
G: Use Viewer As Reference radio button

Before GCPs can be taken, the Reset Horizontal reference source icon (Figure 11 arrow F) must be clicked. For this lab, in the GCP reference source dialog box, the radio button for Image Layer was filled. OK was clicked and the previously orthorectified image was chosen. OK was clicked. The radio button for Use Viewer As Reference (Figure 11 arrow G) was then filled to bring both images onto the screen.

Figure 12: To create a GCP, a suitable area on the reference image was found (left) by using the Select Point icon (looks like an arrow) in the uppermost left corner of the Tool palette (Figure 11 arrow B) to position the inquire boxes around the area of interest. Then using the Create Point icon (looks like crosshairs), next to the Select point icon, the desired position on the reference image was clicked. Then the inquire boxes on the distorted image were moved to the same area as in the reference image and using the Create Point tool a GCP was placed as close as possible to the same spot. Clicking Add, next to the Tool Palette, will create a new blank point.

Figure 13: After two GCPs have been established, the Set automatic (x,y) drive icon can be turned on. This will automatically place a GCP in the distorted image based on the reference. This automatic placement is not perfect and can be off by quite a bit. Therefore, repositioning of the GCP in the distorted image will be necessary but this is helpful in finding the general area of the GCP quickly rather than having to move the inquire boxes around.

As more GCPs are added their locations can be seen in the bottom of the window. For this lab, nine GCPs were created this way and then the horizontal reference source was changed to another image that covered ground not available in the pervious reference image. After 11 GCPs were established the Reset Vertical Reference Source icon was clicked next to the horizontal reference icon. The radio button for DEM was filled and the appropriate DEM of the area was chosen. OK was clicked.

Figure 14: Then, right clicking the first point number in the table at the bottom of the window, choosing select all, and clicking the Update Z values on Selected Points icon (looks like a blue Z), Z values were generated for each GCP.

Then the Type field for each point was set to Full and the Usage field was set to Contorl. This was done by right-clicking the field title and selecting Formula. In the Formula window, the word Full and Usage was typed into the blank section at the bottom to modify the records accordingly. At this point the Point Measurement tool was saved and closed.

Figure 15: What the LPS Project Manager window looks like after the first image has been rectified.

Back at the LPS Project Manager window, another distorted image was added through the same process starting at Figure 8 and finishing before figure 11.

Figure 16: Once the Start Point Measurement tool is selected, GCPs can be established by selecting the Point # field for GCPs that are located on both the first and second image. Points 3,4,7, and 11 were not located on the second image. When a point was highlighted the rectified image on the right automatically shifted to the GCP. Using the inquire boxes, the same area was found manually for the image on the left. Then the Create Point tool was selected and the position on the left image closest to the GCP on the rectified image was clicked. This was repeated for the remainder of the available points.

Next, tie points can be generated. Tie points act like ground control points whose coordinates are uncertain but is visually recognizable on the image. These are generated by the computer and as a result do a much quicker job at finishing the rectifying process compared to manual GCP designation.

Figure 17: The Automatic Tie Point Generation Properties icon (looks like a hand pointing to a blue cross) was clicked to open its associated dialog window. Here tie point generation properties can be modified. For this lab, the settings seen were used along with changing the Intended Number of Points/Image, found in the Distribution tab, was set to 40. Run was clicked.

The tie points could then be checked for accuracy. Save was clicked and then the window was closed. Back at the LPS Project Manager window, triangulation could now be performed. 

Figure 18: The Triangulation dialog window was opened by navigating to Edit > Triangulation Properteis... For this lab, the settings were changed as seen above for the General Tab. In the point tab, the Type was changed to Same weighted values and X, Y, and Z values were changed to 15. Run was clicked.

Figure 19: After the triangulation was complete, a Triangulation Summary box (left) appeared. This report was saved as a text file by clicking the Report option and choosing File > Save As... in the resulting Editor window (right). Accept was clicked in the Triangulation Summary and OK was clicked in the Triangulation window. In the LPS Project Manager window, the Ext field is now green indicating that the external orientation information has now been supplied.

Figure 20: Next, orthorectified images will be created by clicking the Start Ortho Resampling Process icon (Looks like a square divided into four smaller colored squares). In the Ortho Resampling window, the settings were changed to match the figure above.

Figure 21: Then the Advanced tab was clicked and Add.., near the bottom of the window, was selected. The second image was selected in the new Add Single Output window and the Use Current Cell Sizes option was checked. Ok was clicked. Then OK was clicked in the Ortho Resampling window and the process ran and completed.

Figure 22: Here the two ortho images are viewed in the LPS Project Manager. The file name in the left column can be clicked and View selected to see the actual image.




Results

Figure 23: A sample of the anaglyph made in Part 2. Seeing this image through (red/blue) polaroid glasses highlights the differences in elevation around the Chippewa River as it runs through the UW Eau Claire Campus area.

Figure 24: The two orthorectified images overlaying each other. The boundaries of the images and features in the images match up extremely well.


Data Sources

UWEC Department of Geography and Anthropology

Figure 11 taken from lab 7 instructions