Term Project

Image Processing of Crawford County, including Haze reduction, Laplacian edge detection, and Implementation of the Normalized Difference Vegetation Index for Identifying Susceptible areas along the Kickapoo River.

Introduction

As indicated by the title of this project, and the maps located below, my study area was Crawford County.  Crawford County is about 600 square miles in size with a population of 16,644 people according to the U.S. Census Bureau.  Towns and communities located in Crawford County include Prairie du Chein, Seneca, Soldiers Grove, and Gays Mills.  Crawford County was close to Vernon County where I am originally from.  My family used to go to Crawford County and spend time around the Kickapoo River.  The Kickapoo River is an important feature of Crawford County and it has received the nickname of the crookedest river in Wisconsin.

Maps throughout this post were created using ArcGIS® software by Esri. ArcGIS® and ArcMap™ are the intellectual property of Esri and are used herein under license. Copyright © Esri. All rights reserved. For more information about Esri® software, please visit www.esri.com

In this project, what I wanted to do was take a look at the riparian zones around the Kickapoo River, particularly areas that I think may be at greater risk for pollution from run off due to degrades riparian buffers.  I downloaded an image from the U.S. geological service and did multiple forms of remote sensing analysis.  The processes I performed are listed below

Haze reduction

Image subset

Laplacian edge detection

Normalized Difference vegetation Index

Remote Sensing Analysis

My first step was to download the satellite image that I wanted to use for my project.  I searched on USGS global visualization viewer to find the area that I wanted to study.  I had already signed up for an account and since I had found an image that was listed as downloadable, downloading the satellite image went relatively smoothly.  I then unzipped the downloaded file and saved it to a folder. 

I then opened Erdas Imagine and selected import data under manage data.  I input the TIFF files I had, to get the output img files that I wanted. The next step was to layer stack these image files.  I made a mistake originally by stacking all of the layers 1-7, and the image appeared with strange colorations on it.  I then went back and excluded band 6 the thermal layer and then the image appeared normal.

Stacked Image (USGS, 2014)

 

 

The Image was somewhat hazy, so I performed a haze reduction, clicking under raster and radiometric and then clicking haze reduction and saving my output file in a folder labeled Haze reduction

 

Image with Haze reduction

 

Next I decided to take a subset image of Crawford County.  To do this I changed the projection of the Crawford county shapefile to match the projection of the image.  Then I imported the image, selected it and made it an area of interest and then performed a subset and clip.

Subset Image of Crawford County

 

Next I wanted to more easily distinguish the Kickapoo River, so I applied a Laplacian edge detection to the image.  I went to raster, spatial, convolution and selected Laplacian edge detection and then saved the resulting image in a folder labeled edge effects.  The Kickapoo River can be seen more noticeably in this photograph essentially highlighted in red.

Crawford County with Laplacian edge detection

Next I wanted to look specifically at levels of vegetation, so I went under raster unsupervised and performed a Normalized Difference Vegetation Index saving the resulting image in a folder labeled NDVI.

Crawford County NDVI Image

 

The next step in my project was to sync the viewer to Google earth.  Using the NDVI image and the Laplacian edge detection image I was able to find the Kickapoo River and find areas along it that had little to no vegetation.  I was able to find four areas along the River that have little vegetative barrier and are relatively exposed to pollution and runoff from adjacent farms, mining operations and communities

 

Conclusions

 

 

Sources

 

 

U.S. Geological Survey (2014) Landsat Scene "LT50250302011256PAC01" [Satellite photograph]

            Retrieved from http://glovis.usgs.gov/

ESRI (2013) Topographic Base Map [Raster Layer]

            Retrieved from ArcMap 10.2

ESRI (2013) Imagery Base Map [Raster Layer]

            Retrieved from ArcMap 10.2

ESRI (2013) US Counties [Shapefile]

            Retrieved from ESRI 2013 geodatabase

Google Earth (2014) Crawford County, Wisconsin [Aerial Photography]

            Retrieved from Google Earth Pro

Lab 8

Goals and Background

The point of this lab was mainly to learn skills in measurement and interpretation of spectral signatures.  Spectral signatures are essentially the unique levels of reflectance at various wavelengths.  Understanding spectral signatures can help you with identification of features, as well as with studying the health and quality of certain features from aerial and satellite photographs.  The main goals of this lab was to

1.  Understand how to collect spectral signatures
2.  Understand how to analyze spectral signatures

Methods and procedures

To collect spectral signatures

1.  The first step in this lab was to import the image eau_claire_2000.img into the viewer.
2   Connect the viewer to Google earth in order to more easily identify various features.
3.  Zoom in to an area and draw a polygon, then click on raster, supervised, and spectral editor to bring up the spectral signature for the area inside of the polygon.
4.  Repeat the process, drawing polygons in different features, and adding these spectral signatures to the same signature editor.

To  analyze spectral signatures

1.  Highlight all signatures in the signature editor window.
2.  Compare the spectral signatures in the Signature Mean Plot Window
3.  Switch between single and multiple mode in order to view the spectral signatures individually and within the context of other spectral signatures.

In terms of analyzing, there may be some questions that you will need to look up using outside information.  For instance when trying to understand why vegetation has a certain spectral signature, it may be useful to look up information about cell walls and chlorophyll and understand what wavelengths of electromagnetic radiation they absorb and reflect.

Results and Conclusions

The resulting spectral signatures that I collected in this lab can be seen in the picture below.

Spectral signatures for various features in the Eau Claire area

This lab was beneficial in learning how to collect and analyze spectral signatures.  There is obviously more that can be done with spectral signatures, but I think it was good that we know have a basic understanding of what they are and how to collect them.

Source

Satellite Image:  eau_claire_2000.img, Earth Resources Observation and Science Center.  Retrieved from http://glovis.usgs.gov/

Lab 7

Goals and Background

The purpose of this lab was to learn to perform photogrammetric tasks on aerial photographs including

1.  Calculating scales, measurements, and relief displacements.
2.  Using stereoscopy for analysis
3.  Performing orthorectification

Methods and Procedures

To determine the scale of the Photograph

1.  I measured the size of an object on the computer screen and divided that by its real world height which gave me the scale for the aerial photograph

 

A and B marked on the image

The distance from A to B in the image is 2.75 inches

The distance in the real world is 8822.47 feet

8822.47 multiplied by 12 equals 105869.64 inches

105869.64 divided by 2.75 is approximately 38498

The scale of the aerial photograph is 1 to 38498

2.  For a different aerial photograph I used an equation to determine the scale as shown below

Equation: S= f/(H-h)

S= scale

F= focal lens length

H= altitude above sea level

h= elevation of terrain

F= 152 mm

H= 6096000 mm

h= 242620.8 mm

S= 152 / (6096000-242620.8) 

S= 152 / 5853379.2

 

To measure features on aerial photographs

1.  Use the Measure Polygon digitizing tool in Erdas to draw a polygon around the selected feature and double click when completed

 

To calculate relief displacement

Take the necessary measurements and use the formula below

d= (h*r)/H  

d=relief displacement

h= (.25x3209)  inches

r= (5.5x3209) inches

H= (3980x12) inches

d= (802.25*17649.5)/47760

d= 296.47 inches

For Orthorectification

Orthorectification involved setting many properties that would be lengthy to explain individually.  The main steps that were taken however, is that control and Tie points were selected in order to orthorectify the image.  The purpose of orthorectification is to produce an image that is spatially correct, with relief displacement and other geometric errors removed.

Results and Reflections

The final orthorectified image can be seen in the image below

Orthorectified Image

 

Although the orthorectification process was very time consuming, the process made much more sense after having gone through it.  Overall I think the lab helped improve my skills in photogrammetry. 

 

Sources

(NAIP) images: 
United States Department of Agriculture, 2005.
(DEM) for Eau Claire, WI:

United States Department of

Agriculture Natural Resources Conservation Service, 2010.
Spot satellite images:
Erdas Imagine, 2009.   

Lab 6

Goals and Background

The purpose of this lab was to learn how to preform geometric correction on an image.  There were two forms of rectification in this lab

1.  Image to map rectification
2.  Image to image rectification

Methods

For Image to map rectification

1.  First the images Chicago_2000.img and Chicago_drg.img were brought into the two viewers.  Then I clicked on multispectral and Control Points.

2.  I selected Polynomial for the geometric model, in the next window I accepted the default Image Layer(New Viwer) in the GCP Tool Reference Setup.  I then selected Chicago_drg.img as my reference image.   After this step I accepted all of the default properties. 

3.  I deleted the default GCPs and selected my own.  For the equation being used only 3 pairs of GCPs were needed, but 4 were selected. 

GCP points on Reference image shown on the right and input image shown on the left

4.  The GCP points were moved around until the total RMS error (shown at the bottom of the screen) was less than .5

Total RMS error shown at bottom of the screen

5.  I then clicked on the Display Resample Image Dialog button, I named the output file Chicago_2000gcr.img and saved it to my Lab 6 folder.  I left all of the default parameters and executed the geometric correction.

For Image to Image rectification

The process was similar to the image to map rectification.  For my reference image I used Sierra_Leone_east1991grf.img and for my input image I used Sierra_leone_east1991.img.  The only differences in the process were as followed

1.Under the polynomial model properties I changed the polynomial order to 3

2.  12 GCP pairs were used to account for the higher polynomial order

3.  The resample method was changed to bilinear interpolation

Results

The resulting geometrically corrected images can be seen below

Geometrically corrected image of Chicago

Geometrically corrected image of Chicago

Through this lab I was able to learn the steps that need to be taken in order to properly geometrically correct an image

Sources

Satellite images:  Earth Resources Observation and Science Center, United States Geological Survey.

Digital raster graphic (DRG): Illinois Geospatial Data Clearing House.

Lab 5

Goals and Background

The main purpose of this lab was to introduce important analytical skills in remote sensing.  The specific skills this lab emphasized were:

1.  How to mosaic two adjacent satellite images with Mosaic Express and Mosaic Pro
2.  How to implement the normalized difference vegetation index
3.  How to apply various filters
4.  How to apply various contrast stretches
5.  How to perform a histogram equalization
6.   How to perform binary change detection

Methods

To mosaic two images together using Mosaic Express, the following steps were taken:

1.  I Imported the two images found in the Lab 5 Mosaic subset folder.  When adding these I selected Multiple images in virtual mosaic under the multiple tab, and I selected Background transparent and fit to frame under raster options.  When this was done, the below image could be seen in the viewing window

Two imported images

2.  Images were added into Mosaic Express the image intended to be on top was added first, and the other image was added second.  I clicked next until I got to the output dialog box, where I named the output image and saved it to my lab 5 folder.  The output image can be seen below

Output image from Mosaic Express

To mosaic images together using MosaicPro the following steps were taken:

1.  I Repeated step 1 from the previous section.

2.  I Clicked on Mosaic and then click MosaicPro.  When adding images, I clicked the Image Area Options tab and clicked the compute active area button for both of the images.  I added the image I wanted displayed on top first and the other image second.  When this was done the below image was shown in the MosaicPro window

MosaicPro Window

3.  I clicked on the color corrections icon and checked the use histogram matching option and selected overlap areas and clicked ok.

4.  After selcting the default settings for output image options and set overlap fucntion, I ran the mosaic process and saved the output image to my lab 5 folder.  The resulting mosaic image is shown below.


Mosaic image produced with MosaicPro

To implement the normalized difference vegetation index

1.  I brought the image in, and then went to Raster:Unsupervised:NDVI.  I checked that the sensor said Lansat TM, and then selected NDVI for the function under select function.  I saved the output image in my Lab 5 folder.  The resulting image is shown below


NDVI Image

To apply a low pass convolution image

1.  I brought the high frequency image in and then went to raster: spatial: convolution.  I selected 5X5 Low pass for the Kernel and saved the output image in my Lab 5 folder.  The comparison of the two images can be seen below.



Original image on left, 5X5 low pass image on right

To apply a high pass convolution filter

1.  I brought in the low frequency and then followed the same steps as the low pass filter, except I selected the 5x5 high pass filter.  The comparison of the two images can be seen below.


Original image on left, 5X5 high pass image on right

To apply a Laplacian convolution filter

1.  I brought in the image and then followed the same steps for the other filters, except selecting 3X3Laplacian filter instead.  The comparison of the two images can be seen below.


Original image on Left, Laplacian edge detection image on right

To apply a minimum maximum contrast stretch

1.  I imported the image, went to general contrast and changed the method to Gausian and applied the change.  The resulting image is shown below.

To apply a piecewise contrast stretch

1.  I imported the image and clicked on General contrast then piecewise contrast.  I set the specific ranges for each mode seen on the image histogram and then clicked apply.  The result is shown below



Piecewise contrast stretch

To perform a histogram equalization
1.  I imported the image and clicked on radiometric and histogram equalization.  I saved the resulting image in my Lab 5 folder. 

To perform Binary change detection

1.  I imported the two images, clicked on raster function then clicked on two image function.  I changed the detection process to layer 4 and then I clicked ok to run image differencing.

2.  I then changed the pixels using model maker, setting the model so that it was for binary change detection. 

3.  I then developed another model maker that would show the changed pixels within the study area.


4.  I mapped the resulting image over an NIR image for the same area, and changed the pixels to a brighter color.  I added a legend and made a few adjusting to make a basic map.

Results

All of the resulting images from this lab I saved to my lab 5 folder.   During this lab, I learned ways that it is possible to take images and adjust them for particular uses or to make them easier to interpret by improving their quality.

Sources:  All image functions were performed in Erdas Imagine and ArcMap.
                All images were made available in our lab folder

Lab 4

Goals and Background

The main purpose of this lab was to familiarize us with the use of several different image reprocessing tools.  The main skills that were emphasized in this lab were

1.  Understand how to take a subset of a larger image
2.  Understand pan-sharpening, and how to increase spatial resolution
3.  Understand radiometric enhancement techniques
4.  Understand how to link an image to Google Earth, and how this can be useful
5.  Understand how to resample satellite images

All of the data used in the lab was made available our Lab 4 folder

Methods


1.  Understand how to take a subset of a larger image:  In order to achieve this, I used the Subset & Clip tool to create subset image.  When I had a specific area that I needed to subset I inserted the shapefile of the desired area, selected the shapefile and clicked on paste from selected object.  Then I was able to make an area of interest which could be designated as the subset image.

2.  Understand pan-sharpening, and how to increase spatial resolution: For the pan-sharpening, I went to Pan Sharpen and clicked on resolution merge.  From there I defined what was the high resolution image, what was the multispectral image, and what I wanted the output file to be. 

3.  Understand radiometric enhancement techniques:  For the image in question there was a bit of haze, so I clicked on radiometric and went to Haze reduction and designated a file for the output image. 

4.  Understand how to link an image to Google Earth, and how this can be useful:  For this section I clicked on Connect to Google Earth and then linked the image that I had in my viewer to Google earth.  Using it, I was able to see what the features were that I was actually looking at, because Google earth had high spectral areal photos where the features were more obvious.

5.  Understand how to resample satellite images:  For this section I clicked on Spatial and then clicked on Resample pixel size.  I tried two different resampling techniques and noticed that nearest neighbor was virtually indistinguishable from the original, while bilinear interpolation produced an image with much smoother lines.

Results

Below are the images that I produced during the Lab


Image Subset

Image subset from Area of Interest

Haze reduction image

Pan-sharpened image



Nearest Neighbor resampled image

Bilinear Interpolation resampled image

 

Sources

All images were provided in our Lab 4 folders
All image editing was done in Erdas Imagine