Goal and Background:
The goal of this lab exercise is to develop skills in image preprocessing, image enhancement, delineation of a study area from a larger satellite scene, mosaicking multiple image scenes, and building graphical models for remote sensing analytics. Developing skills in these areas is important to building my repertoire of remote sensing skills.
Methods:
Part 1: Image Subsetting (creation of area of interest AOI) of Study Area
Section 1: Subsetting with the use of an Inquire box
A subset image of the Eau Claire-Chippewa area (WI) was created using an Inquire Box in ERDAS IMAGINE 2013. The uploaded TM satellite image of the Eau Claire region was uploaded and an area of interest was created by first using the Inquire Box tool in the Raster tab. Then the Create Subset Image feature was used in the Subset & Chip tool. The coordinates in the subset image were matched to the larger image as well. The results of Part 1, Section 1 can be seen in Figure 1 below.
Section 2: Subsetting with the use of an area of interest shape file
A subset image of the Eau Claire-Chippewa area was created using an area of interest shape file. The selected TM satellite image of the Eau Claire-Chippewa area and provided shapefile of Eau Claire and Chippewa Counties was added to a viewer in ERDAS IMAGINE 2013. The shapefile was selected and saved as an aoi file. Finally, the Subset & Chip tool in the Raster tab was used to create the final subset of the area of interest. The results of Part 1, Section 2 can be seen in Figure 2 below.
Part 2: Image Fusion
A 15 meters panchromatic image of Eau Claire and Chippewa Counties was used to pan-sharpen a 30 meters reflective image of Eau Claire and Chippewa Counties. This was accomplished using the Resolution Merge feature under the Pan Sharpen icon in the Raster tab in ERDAS IMAGINE 2013. In the Resolution Merge window, the Multiplicative algorithm was selected for Method and Nearest Neighbor was chosen as the Resampling Technique. The results of Part 2 can be seen in Figure 3 below.
Part 3: Simple Radiometric Enhancement Techniques
The spectral and radiometric quality of a satellite image of Eau Claire was enhanced through the use of radiometric enhancement techniques. Specifically, haze reduction was performed on the satellite image. This was accomplished by using the Haze Reduction tool under the Radiometric icon in the Raster tab in ERDAS IMAGINE 2013. The results of Part 3 can be seen in Figure 4 below.
Part 4: Linking Image Viewer to Google Earth
A Google Earth image and satellite image of the Eau Claire area were linked to each other using ERDAS IMAGINE 2013. Google Earth obtains images from the GeoEye high resolution satellite system. To link the images, the Connect to Google Earth tool under the Google Earth tab was used. In order to match the image viewer with the Google Earth window, the Match GE to View tool under the Google Earth tab was first used. After that, the Sync GE to View tool was used to complete the match. Linking the image viewer in ERDAS IMAGE 2013 to Google Earth is an example of creating a selective key, which can be used to study areas of interest.
Part 5: Resampling
A satellite image of Eau Claire was resampled to reduce the pixel size from 30x30m to 20x20m. This was accomplished by using the Resample Pixel Size tool under the Raster tab. The image was resampled twice, once using the nearest neighbor method and once using the bilinear interpolation method. Results for the nearest neighbor method and bilinear interpolation method can be seen in Figures 5 and 6 respectively.
Part 6: Image Mosaicking
Two adjacent satellite images were mosaicked to create one continuous image. The mosaicked images cover path 25, row 29, and path 26, row 29. These images were taken in May 1995 by the Landsat TM satellite. To start, the images to be mosaicked were added to the same viewer through the use of Multiple and Raster Options tabs in the Select Layers to Add window. The images were then mosaicked using Mosaic Express and MosaicPro, as discussed in part 6, sections 1 and 2 below, respectively.
Section 1: Image mosaic with the use of Mosaic Express
The two satellite images were first mosaicked using the Mosaic Express tool under the Mosaic icon in the Raster tab. Appropriate values were inputted into the different tabs of the Mosaic Express window to create the mosaicked image. The results of Part 6, Section 1 can be seen in Figure 7.
Section 2: Image mosaic with the use of MosaicPro
The images were then mosaicked using the MosaicPro tool under the Mosaic icon in the Raster tab. Compute active area was selected under Image Area Options in the Add Images window for both images. The radiometric properties at the intersection of the images were synchronized to maintain a smooth color transition at the intersection of the images. This was accomplished by setting Use Histogram Matching and Overlap Areas in the Color Corrections tool of MosaicPro. Additionally, the Set Overlap Function was set to Overlay. The results of Part 6, Section 2 can be seen in Figure 8.
Part 7: Binary Change Detection (Image Differencing)
Brightness values were estimated and mapped for images of Eau Claire and four other neighboring counties, one image from August 1991 and the other from August 2011. See the two sections below for more details.
Section 1: Creating a difference image
Two Landsat TM images of the Eau Claire area were first put into the same viewer. Changes between the images were then analyzed with image differencing. The was done using the Two Image Functions tool under the Functions tab. In the Two Input Operators interface, the 1991 image was subtracted from the 2011 image.
After the image differencing program was run, the image did not show if any areas changed or not. To see the changed areas, a histogram was constructed. Upper and lower thresholds for the data were calculated using Equation 1. The resulting histogram can be seen in Figure 9.
mean + 1.5 standard deviation (1)
Section 2: Mapping change pixels in difference image using spatial modeler
To map changes in the images using spatial modeler, Equation 2 was used to eliminate the negative values in the difference image created in Part 7, Section 1.
ΔBVijk = BVijk(1) – BVijk(2) + c (2)
Where:
ΔBVijk = Change pixel values
BVijk(1) = Brightness values of 2011 image
BVijk(2) = Brightness values of 1991 image
c = constant (127 in this case)
This equation was applied in Spatial Modeler using Equation 3 below. The model was then run to produce an output image.
$n1_ec_envs_2011_b4 - $n2_ec_envs_1991_b4 + 127 (3)
Next, the new threshold of the histogram was calculated using Equation 4. This was done because a constant was used in Equation 3 to eliminate negative brightness values.
mean + (3 x standard deviation of the mean) (4)
To continue, another spatial model was created to show which pixels changed in the images between 1991 and 2011 using the change/no change threshold calculated in Equation 4. The Either If OR operator in Equation 5 was used for the model.
EITHER 1 IF ($n1_ec_91>change/no change threshold) OR 0 OTHERWISE (5)
Finally, the resulting image was mapped in ArcMap 10.2.2 with the overlain image of Eau Claire and four other counties from 1991. The map, displayed in Figure 10, shows the change in pixel values in Eau Claire and four other counties from 1991 to 2011.
Results:
 |
Figure 1: Subset image of Eau Claire-Chippewa Area using an Inquire Box
|
 |
| Figure 2: Subset image of Eau Claire-Chippewa Area using an Area of Interest shapefile |
 |
| Figure 3: Pan-sharpened image of Eau Claire-Chippewa Area Using Image Fusion |
Pan-sharpening in Figure 3 made the image more clear.
 |
| Figure 4: Haze reduction image of Eau Claire using simple radiometric enhancement techniques |
 |
| Figure 5: Resampled image of Eau Claire using Nearest Neighbor method |
 |
| Figure 6: Resampled image of Eau Claire using Bilinear Interpolation method |
Both resampled images in Figures 5 and 6 are very similar to the original image.
 |
| Figure7: Mosaicked Image of Eau Claire using Mosaic Express |
The color transition between the two images in Figure 7 is not very smooth.
 |
| Figure 8: Mosaicked image of Eau Claire using MosaicPro |
 |
| Figure 9: Histogram of differenced image with upper and lower thresholds |
 |
| Figure 10: Map depicting change in brightness values |
Areas that changed over the 20 year period in Figure 10 are close to urban centers, but do not directly reside in urban centers. This could be due to the process of transforming land on the outskirts of cities to agricultural land.
Sources:
Earth Resources Observation and Science Center. (2013). [Satellite images in img. format] United States Geological Survey. Retrieved from http://eros.usgs.gov/.
Price, M. (2014). Mastering ArcGIS 6th Edition. Mastering ArcGIS 6th Edition Dataset [shapefile]. New York: McGraw Hill.
No comments:
Post a Comment