Goal/background
The goal of this lab was to practice and become familiar with a variety of miscellaneous image functions. This includes (1) creating subsets from a larger image scene, (2) performing a pan sharpen, (3) perform a haze reduction, (4) link a satellite image to Google Earth, which can be used as a selective key, (5) resample images, (6) mosaic pairs of images using both Mosaic Express and MosaicPro, and (7) use graphical modeling to detect binary changes in images. All functions were done using ERDAS IMAGINE software
Methods
Part 1: Image subsetting
Section 1: Subsetting with the use of an Inquire box
1. A large satellite image was opened. An inquire box was first made around the desired area, the city of Eau Claire by right clicking on the image and selecting “Inquire Box.”
2. The “create subset image” menu was opened from the “Subset and Chip” dropdown menu. An output folder was created and the inquire box extents were used for the subset as well.
3. This was run and the a new file of the subset image of the City of Eau Claire was created.
Section 2: Subsetting with the use of an area of interest shape file
1. A large satellite image was opened. A shapefile of the desired subset area, Eau Claire and Chippewa counties, was opened in the same layer.
2. The two objects in the shapefile were activated by holding shift and clicking on them
3. An area of interest was created around the selected objects by clicking “paste from selected object.”
4. The area of interest was saved as a .aoi file. The .aoi file was used in a similar way as the inquire box was used in the previous method. “Create subset image” was selected but this time the extents of the .aoi were used to create the subset.
Part 2: Image fusion (pansharpen)
1. Two maps were opened, each of the same temporal scale. One was multispectral with 30 m spatial resolution, while the other was panchromatic with 15 m spatial resolution.
2. “Resolution merge” was selected from the Pan Sharpen menu. The panchromatic image was chosen as the high resolution image file and the multispectral file was chosen for the multispectral image file.
3. The multiplicative method and nearest neighbor resampling technique were selected and the model was run.
Part 3: Simple radiometric enhancement techniques
1. “Haze reduction” was selected from the radiometric dropdown menu. A 2007 image of Eau Claire was selected for input and an output location and name were designated.
2. All defaults were accepted and the model was run.
2. All defaults were accepted and the model was run.
Part 4: Linking image viewer to Google Earth
1. A 2007 image of Eau Claire was added to the image viewer. The “Connect to Google” option was found via the help search. This opened the Google Earth Pro window.
2. To link views between the Google Earth window and the Erdas Imagine window, “Match GE to View” was selected in the Erdas Imagine window.
3. Views were then synced by clicking “Sync GE to View.
4. At this point, the Google Earth view could be used as a selective interpretation key because of the markers like roads and places as well as very high resolution it provides.
Part 5: Resampling
1. A 2011 image of Eau Claire was added to the image viewer. Metadata showed that the pixel size of this image was 30 m.
2. “Resample pixel size” was opened from the “Spatial” dropdown menu. The 2011 image was used as the input file and an output file location and name were designated.
3. Output cell size was changed from the existing 30 m in both the X Cell and Y Cell to 15 m in both. In addition, the “square cells” option was checked to ensure pixel sizes were square.
4. Nearest neighbor was used for the resample method and the model was run.
5. Steps 2 and 3 were repeated for the same image.
6. Bilinear Interpolation was chosen for resample method instead of Nearest neighbor as in step 4. The model was run and all 3 images were compared.
Part 6: Image Mosaicking
1. Two images captured in May 1995 by the Landsat Tm satellite were used for this task. One covers path 25, row 29 and the other covers path 26, row 29.
2. The “Multiple” tab was opened and “Multiple Images in Virtual Mosaic” was selected.
3. The “Raster Options” tab was opened to ensure that “Background Transparent” and “Fit to Frame” were checked.
4. One of the two image files was selected and brought into the viewer.
5. Steps 2 through 5 were repeated for the other image, which was added to the same layer.
Section 1: Image mosaic with the use of Mosaic Express
1. “Mosaic Express” was selected from the “Mosaic” dropdown menu.
2. Both image files were entered to be used as the data sources for the mosaic. Because order matters, the image on path 25 was added first so that it would be the top image.
3. All subsequent defaults were accepted and an output location and name were designated. The model was run by clicking “finish.”
Section 2: Image mosaic with the use of MosaicPro
1. “MosaicPro” was selected from the “Mosaic” dropdown menu.
2. The “add images” icon was selected. The “Image Area Option” tab was opened and “Compute Active Area” was selected, activating the “Set” button. Because cropping or reducing the spatial extent was not necessary, the automatic “Active Area Options” were used.
3. The image taken of path 25 was selected and added to the MosaicPro window.
4. Steps 2 and 3 were repeated for the image taken of path 26. At this point, both images were together in the MosaicPro window.
5. The “Color Correction” icon was selected and “Use Histogram Matching” was checked.
6. In the Histogram Matching dialogue, the Matching Method was switched to “Overlap Areas” so that colors outside the overlapping areas would be preserved.
7. The “Set Output Options” icon was selected. Because map projection, pixel size, or other parameters didn’t have to be changed, all defaults were accepted.
8. The “Set Overlap Function” icon was selected and the default “Overlay” function was accepted.
9. The “Process” menu was opened, an output location and name were designated, and the Mosaic was run.
Part 7: Binary change detection (image differencing)
Section 1: Creating a difference image
1. Images of Eau Claire and four neighboring counties county in 1991 and 2011 were used to estimate and map brightness value changes in pixels.
2. These images were synced.
3. “Two Image Functions” was selected from the “Functions” dropdown menu of the “Raster” toolbar.
4. In the menu, the 2011 image of Eau Claire was added as “Input File #1” and the 1991 image was added as “Input File #2.” An output location and name were designated.
5. The additive operator was changed to the subtractive operator by choosing the (-) from the dropdown menu.
6. Both input files were changed to only use layer 4 in order to simplify the change detection and the model was run.
7. To find the upper and lower limits of change between the two images, information from the general metadata the mean and standard deviation. The mean plus 1.5* standard deviation provides a good estimate of these limits. The median was estimated from the histogram so that lower and upper limits of change could be calculated as shown.
Section 2: Mapping change pixels in difference image using spatial modeler
1. This section involves using the difference image from the previous section to map only the areas that changed between 1990 and 2011. To do this, a model had to be created. A new viewer was opened and “Model Maker” was selected from the “Model Maker” dropdown menu in the “Toolbox” toolbar.
2. The model maker was opened. Three tools were used to outline the model to be used in this section including placing a raster (top), placing function (middle), and connecting those parts (bottom).
3. Double clicking on the raster objects allowed inputs to be selected. NIR bands from 1991 and 2011 were selected for the raster inputs.
4. The function used the equation ΔBVijk = BVijk(1) – BVijk(2) + c where BVijk(1) and BVijk(2) are the two input rasters, c is a constant used so that there are positive brightness values. ΔBVijk is the change in pixel values. ΔBVijk will be the output raster.
5. Again, the change threshold as calculated the same method as above, using the below numbers.
6. Another model was created in which the difference image was the input, the function was of conditional function of either a pixel has a value of one if it is greater than the change threshold or a value of 0 if it is not.
7. To view the resulting image more clearly, it was opened in ArcMap and overlayed on the image from 1991. Symbolization was changed as shown in the results.
Results
Part 1: Image subsetting
Section 1: Subsetting with the use of an Inquire box
Section 2: Subsetting with the use of an area of interest shape file
Figure 1: Image subset from a simple inquire box.
Figure 2: Image subset from an area of interest file.
This allows more flexibility in the shape of the image subset.
Part 2: Image fusion (pansharpen)
Figure 3: Panchromatic image (top) and multispectral image (bottom) used to createPansharpened image (left). Pansharpening allows the contrast and colors from a mulstispectral image to be combined with the better spatial resolution of the panchromatic image.
Part 3: Simple radiometric enhancement techniques
Figure 4: Haze reduction (right) significantly increased the contrast and sharpness of the original image (left). This is a very simple technique and useful when there are things like excess moisture in the atmosphere.
Part 4: Linking image viewer to Google Earth
Figure 5: Linked views of image file and Google Earth. Google Earth provided an extremely detailed selective key that includes road names, places, and very high spatial resolution.
Part 5: Resampling
Figure 6: Images and magnifications (bottom) comparing effects of resampling from the original image (left) by nearest neighbor (center) and bilinear interpolation (right). Bilinear interpolation resulted in smaller pixel size, causing differences in the image that could only be seen at high magnification.
Part 6: Image Mosaicking
Section 1: Using Mosaic Express
Figure 7: Mosaicked image from the automated Mosaic Express program. Because there is almost no user input, the output is not seamless.
Section 2: Using MosaicPro
Figure 8: Mosaicked image from MosaicPro. This uses extensive user input. Color correction was done by histogram matching to make the result almost seamless.
Part 7: Binary change detection (image differencing)
Figure 9: Changes between 1991 and 2011 in Eau Claire and four neighboring counties can be clearly seen using the method outlined.
Sources
- Earth Resources Observation and Science Center, United States Geological Survey (satellite images)
- Price, M. (2014). Mastering ArcGIS 6th edition. McGraw Hill. (shapefile)
