Lab 4
Goal and Background
The objective of Lab 4 was to introduce students to numerous functions and processes in Erdas Imagine that will build up skills in image processing. The particular processes in Lab 4 include: two methods of delineating a study area from a satellite image, optimizing spatial resolution through pan-sharpening, a radiometric enhancement technique, linking a satellite image to Google Earth, resampling methods, mosaicking images, and binary change detection through simple graphical modeling. Subsequently, students are better prepared for image processing, enhancement, and interpretation with experience with these techniques and methods.
Methods
The first skill introduced was subsetting images in order to focus on an area of interest. One method was using an Inquire Box - a simple, but limiting method because areas of interest are often not simple square or rectangle. Using an image of Eau Claire county taken in 2011, the Inquire Box, located within the Raster tab, is utilized. The Inquire Box is activated by clicking on the image, causing a white box to appear. Positioning the box over the Eau Claire and Chippewa cities, the Inquire Box magnifies the area it encompasses. Click apply in the Inquire Box view. In the Raster tab, from the Subset and Chip drop down window, clicking on Create Subset Image will automatically input the image file on the image viewer. Save the output to a folder and click From Inquire Box, which brings in the coordinates from the Inquire Box. Running the model, Figure 1 is what the output looks like displayed in the viewer.
Using the same Eau Claire 2011 image, display a shapefile of the Eau Claire and Chippewa counties over the image to begin subsetting by the shape of the two counties. Holding down the shift key and clicking on the counties will select them. From the home tab, select Paste from Selected Object. It's then possible to save the parts of the image that was pasted as an AOI file. Once that's completed, return to the Create Subset Image window. This time, bring in the AOI file that was just saved and then run the model. The output is Figure 2.
The next technique to learn is create a higher spatial resolution for an image by pan-sharpening. The two images used are a reflective image and a panchromatic image of the Eau Claire and Chippewa area in 2000, which have 30 meters and 15 meters spatial resolution respectively. Opening the Resolution Merge window from Pan Sharpen Menu in the Raster tab. The panchromatic image is brought into the High Resolution Input file and the reflective image brought into the Multispectral Input file. Save the output to a specific folder. Run the model using a Multiplicative method and a Nearest Neighbor technique.
Haze reduction was the next technique The Haze Reduction tool is located in the Radiometric drop down window in the Raster tab. Using this tool, the haze in an image of the Eau Claire area in 2007 was reduced and enhancing its radiometric resolution.
Bringing the 2011 image of Eau Claire to viewer and fitting it to frame, use the Connect to Google Earth tool in the Google Earth tab. By Matching GE to View, the image is shown in full extant in Google Earth. Feature identification is made easier by syncing the two using the Sync GE to View.
Next, the 2011 Eau Claire image will be resampled up from 30x30 meters to 15x15 meters pixel size. To do this, use the Resample Pixel Size tool from the Spatial in the Raster tab. First, try using the Nearest Neighbor Resampling Method. Not much of a difference is noticeable. Then try Bilinear Interpolation. Zooming in, the pixel sizes are noticeably smaller.
If an area of interest is too large for an image or traverses two adjacent images, mosaicking images together can resolve this. Before adding two 1995 Eau Claire images into the viewer, check Fit to Frame and Background Transparent in Raster Options. In Multiple, activate Multiple Images in Virtual Mosaic. After the images are displayed in viewer, use Mosaic Express in the Mosaic tab. Add the two images in the Express Window and run the model. Figure 3 shows it doesn't create a seamless output. Next try MosaicPro from the Mosaic drop down window. But first, add the same two images to viewer with the same parameters as before. In the MosaicPro window select the Add Images icon to add the images, one after another. For this Mosaic, use Histogram Mapping for the Overlap Areas from the Color Corrections button. This will help create a smoother transformation between the images. Select Process and the Run Mosaic. Figure 4 shows a much smoother output mosaic.
The last technique to learn to complete Lab for is Binary Change Detection to estimate and map brightness values of pixels that have changed, in this case, between 1991 and 2011 in Eau Claire and nearby counties. Access the Two Input Operators interface from Raster-Functions-Two Image Functions window. Insert the 2011 image in Input File #1 and the 1991 image in Input File #2. Change the operator to a - symbol and ALL layers to Layer 4. Run the model. View the output's metadata. Estimate a threshold by adding the mean to the multiplication of the standard deviation and 1.5. Add and subtract this value from the center value of the image's histogram to get the upper and lower limits, where most of the change in values occur (Figure 5). Open Model Maker in the Toolbox tab. Create a model that has two rasters interact with a function that generate a raster output. The two rasters are 2011 and 1991 band 4 image of Eau Claire. The function is subtracting the 1995 image from the 2011 image and adding the constant 127. Send the output to a specific folder. Open the output's metadata. Estimate the threshold by adding the mean to a multiplication of the standard deviation and 3. Open Model Maker again, this time with a raster object connected to a function object that's connected to another raster object. Insert the first model's output into the new input raster. Open the function object and change it to Conditional. Write the function as EITHER 1 IF (file name>202.18) OR 0 OTHERWISE. Send the output raster to the folder with the previous output. Open ArcMap and display the new output over an image of the area in 1991. Figure 6 shows a map of the changed areas generated by Model Maker. EITHER 1
Using the same Eau Claire 2011 image, display a shapefile of the Eau Claire and Chippewa counties over the image to begin subsetting by the shape of the two counties. Holding down the shift key and clicking on the counties will select them. From the home tab, select Paste from Selected Object. It's then possible to save the parts of the image that was pasted as an AOI file. Once that's completed, return to the Create Subset Image window. This time, bring in the AOI file that was just saved and then run the model. The output is Figure 2.
The next technique to learn is create a higher spatial resolution for an image by pan-sharpening. The two images used are a reflective image and a panchromatic image of the Eau Claire and Chippewa area in 2000, which have 30 meters and 15 meters spatial resolution respectively. Opening the Resolution Merge window from Pan Sharpen Menu in the Raster tab. The panchromatic image is brought into the High Resolution Input file and the reflective image brought into the Multispectral Input file. Save the output to a specific folder. Run the model using a Multiplicative method and a Nearest Neighbor technique.
Haze reduction was the next technique The Haze Reduction tool is located in the Radiometric drop down window in the Raster tab. Using this tool, the haze in an image of the Eau Claire area in 2007 was reduced and enhancing its radiometric resolution.
Bringing the 2011 image of Eau Claire to viewer and fitting it to frame, use the Connect to Google Earth tool in the Google Earth tab. By Matching GE to View, the image is shown in full extant in Google Earth. Feature identification is made easier by syncing the two using the Sync GE to View.
Next, the 2011 Eau Claire image will be resampled up from 30x30 meters to 15x15 meters pixel size. To do this, use the Resample Pixel Size tool from the Spatial in the Raster tab. First, try using the Nearest Neighbor Resampling Method. Not much of a difference is noticeable. Then try Bilinear Interpolation. Zooming in, the pixel sizes are noticeably smaller.
If an area of interest is too large for an image or traverses two adjacent images, mosaicking images together can resolve this. Before adding two 1995 Eau Claire images into the viewer, check Fit to Frame and Background Transparent in Raster Options. In Multiple, activate Multiple Images in Virtual Mosaic. After the images are displayed in viewer, use Mosaic Express in the Mosaic tab. Add the two images in the Express Window and run the model. Figure 3 shows it doesn't create a seamless output. Next try MosaicPro from the Mosaic drop down window. But first, add the same two images to viewer with the same parameters as before. In the MosaicPro window select the Add Images icon to add the images, one after another. For this Mosaic, use Histogram Mapping for the Overlap Areas from the Color Corrections button. This will help create a smoother transformation between the images. Select Process and the Run Mosaic. Figure 4 shows a much smoother output mosaic.
The last technique to learn to complete Lab for is Binary Change Detection to estimate and map brightness values of pixels that have changed, in this case, between 1991 and 2011 in Eau Claire and nearby counties. Access the Two Input Operators interface from Raster-Functions-Two Image Functions window. Insert the 2011 image in Input File #1 and the 1991 image in Input File #2. Change the operator to a - symbol and ALL layers to Layer 4. Run the model. View the output's metadata. Estimate a threshold by adding the mean to the multiplication of the standard deviation and 1.5. Add and subtract this value from the center value of the image's histogram to get the upper and lower limits, where most of the change in values occur (Figure 5). Open Model Maker in the Toolbox tab. Create a model that has two rasters interact with a function that generate a raster output. The two rasters are 2011 and 1991 band 4 image of Eau Claire. The function is subtracting the 1995 image from the 2011 image and adding the constant 127. Send the output to a specific folder. Open the output's metadata. Estimate the threshold by adding the mean to a multiplication of the standard deviation and 3. Open Model Maker again, this time with a raster object connected to a function object that's connected to another raster object. Insert the first model's output into the new input raster. Open the function object and change it to Conditional. Write the function as EITHER 1 IF (file name>202.18) OR 0 OTHERWISE. Send the output raster to the folder with the previous output. Open ArcMap and display the new output over an image of the area in 1991. Figure 6 shows a map of the changed areas generated by Model Maker. EITHER 1
IF ( $n1_ec_91>
change/no change threshold value
) OR 0 OTHERWISE
Results
Figure 1: Area if interest delineated using an Inquire Box
Figure 2: Area of interest delineated using a shapefile.
Figure 3: A mosaic image generated by Mosaic Express.
Figure 4: A mosaic image generated by MosaicPro.
Figure 5: Histogram showing the upper and lower limits of the changed/no change areas between the 1991 and 2011 images.
Figure 6: Map showing changed areas in Eau Claire and nearby counties.
Sources
Earth Resources Observation and Science Center, United States Geological Survey (n.d.) Satellite Images. Reston, VA.
Price, Maribeth (2014). Mastering ArcGIS 6th Edition Dataset. Redlands, CA: McGraw Hill.
Comments
Post a Comment