Lab 8

The purpose of this lab is to gain experience collecting and analyzing spectral signatures. This was done by collecting 12 spectral signatures of different surface features by developing AOIs with the polygon drawing tool in Erdas 2013 and graphing their curves.

Methods: Landsat ETM+ image of Eau Claire, WI in 2000 was used in this project. AOIs were collected from the drawing tool and then included on a graph using a supervised signature editor in Raster options. 12 spectral signature curves were collected and reported on one graph. These included: Standing Water, moving water, vegetation, riparian vegetation, crops, urban grass, dry soil, moist soil, rock, asphalt highway, airport runway, and a concrete parking lot. Each curve was then looked at to determine the most and least amount of spectral reflectance in bands 1-6 and then compared against each other to determine which surfaces were most similar to each other and which ones were least alike.

Results: 

Spectral Reflectance Curve graph created with a
Supervised Spectral Signature Editor.              

Spectral Signatures- Highest/Lowest Reflectance:
Standing water: Highest: band 1 (blue)= ~77um, Lowest: Band 4 (NIR)=~20um
Moving water: High: Band 1 (Blue)=79um Low: Band 6 (MIR2)=26um
Vegetation: High: Band 4 (NIR)= 123umLow: Band 3 (red) and 6 (MIR2) are both ~32um
Riparian vegetation: High: Band 4 (NIR)=109um Low: Band 6 (MIR2)= 36um (close tie with band 3 (red)=37um)
Crops: High: band 4(NIR)= 76um Low: Band 6 (MIR2)= 26um
 Urban Grass: High: Bands 4 (NIR) and 5 (MIR1)  =~116um each. Low: Band 6 (MIR2)= 61um
 Dry soil (uncultivated): High: Band 5 (MIR1)=128um Low: Band 3 (Red)= 81um

Moist soil (uncultivated): High: Band 1 (blue)= 70um Low: band 6 (MIR2)=37um
Rock: High: Band 1(blue)=84um Low: Band 6 (MIR2)= 50um

Asphalt highway: High: band 1 (blue)= 106um Low: Band 4 (NIR)= 63um

Airport runway: High: Band 3(red) and band 5(MIR1)= 153um Low: Band 4 (NIR)= 82um

Concrete surface (parking lot): High: Band 3 (red)= 114um Low: Band 4 (NIR) = 61um

Standing water and the airport runway differ the most in spectral signature curves. Water reflected the most light in band 1 while the airport runway reflected the most light in band 3 and 5. The mean differences between their reflectance values in any given band had a wider range than most other signature curves. This is due to the nature of the absorbency of light in water and concrete surfaces.

Riparian vegetation and vegetation spectral signature curves were the most similar. Both curves reflect the most light in band 4 and reflects the least in bands 6 and 3. Riparian vegetation reflected more light than vegetation in bands 1,2,3, and 6 while regular vegetation reflected more in bands 4 and 5. I expected to see riparian vegetation absorbing more (consequently reflecting less) than regular vegetation across all bands because of the nearby water feeding the plants that would therefore have more water within their systems. However, this was not the case. Riparian vegetation may then not have any more water within their structures than other vegetation and the differences between the two vegetative areas may simply be do to chance and slight differences in health and photosynthetic processes. Both vegetative regions consisted of trees and thus are subject to a fair and similar amount of diffuse reflection and given similar amount of nutrients (not necessarily water as this may not be the limiting nutrient), and if displayed similar species composition, both would exhibit similar reflectance curves. The differences may lie in these stated conditions as well but may only deviate slightly.

Sources:
Wilson, Cyril. (2014). Geog 338: Remote Sensing of the Environment Lab 7 Photogrammetry. Personal collection of  Cyril Wilson, University of Wisconsin-Eau Claire, Eau Claire, Wisconsin.

Lab 7

The goal of this lab was to become familiar with stereoscopy and orthorectification tasks on satellite and aerial photographs. This included creating an anaglyph to observe elevation features in the Eau Claire, WI region and collecting GCPs to create tie points.

Stereoscopy:

Methods: To become more familiar with stereoscopy, I created an anaglyph image of the city of Eau Claire, WI. I did so by running a digital elevation model (DEM) of the region through the Anaglyph Generation tool in Erdas Imagine 2013. I then observed elevation features.

Results:

The anaglyph showed regions of high and low elevations such as valleys near the Chippewa River or Halfmoon Lake and areas of high elevation on the UWEC campus hill. All elevation shifts seemed to be more dramatic than in real life.





Anaglyph of Eau Claire, WI. 


Orthorectification:

Methods: I orthorectified SPOT satellite images by collecting GCPs (with the classic point measurement tool) on two reference images using the LPS tool. 9 GCPs were collected using the first horizontal reference image and 2 GCPs were collected from the second horizontal reference image. Vertical reference information was collected by using a digital elevation model image and updating the Z values on the collected GCPs. More GCPs were collected on another reference image according to the GCPs already collected after running the DEM and Z values. A triangulation was then run on the two satellite images being orthorectified.

Results:

 The resulting photograph was an accurate spatial boundary between the two images with very little skewing.







Orthorectified SPOT satellite images

Sources:
Wilson, Cyril. (2014). Geog 338: Remote Sensing of the Environment Lab 7 Photogrammetry. Personal collection of  Cyril Wilson, University of Wisconsin-Eau Claire, Eau Claire, Wisconsin.

Lab 6

This lab focused on two geometric correction methods: Image-to-map rectification using nearest neighbor resampling and image-to-image registration using bilinear interpolation resampling.

Image-To-Map Rectification:

I used a map of Chicago to collect Ground Contrtol Points (GCPs) for a satellite image of Chicago. The rectification was set to a first order polynomial and I collected 4 GCPs and aqcuired a RMS of 0.2 and resampled the image using nearest neighbor.





Image-To-Image Registration:

I used a reference image of Sierra Leon to collect 12 GCPs for a third order polynomial registration and achieved a RMS of 0.07. I then resampled the image using bilinear interpolation. The finished product had a white, hazey appearence, but the features in the images aligned well with the reference image.

Lab 5

The goal of this lab was to become acquainted  with the processes of image mosaicking, band ratioing, and conducting spectral or spatial image enhancement. These processes are important when faced with a situation in which you must collect data from an area that spans across multiple images or you must analyze and interpret the data of an image that may not be clear.

Two image mosaic processes were used and compared against each other to become familiar with the function and to see which process was more useful when stitching images together.

MosaicPro:

I added the images of interest into the MosaicPro tool and selected "Compute Active Area" in the Image Area Options menu. I used "Histogram Matching" to correct the color in order to make a more seamless stitch in the overlapping area. The Overlap Function perameters were set to default in Overlay to set the brightness values to the top image's in the area where they overlap. The end product was more of an unnoticeable seam than the one mosaic made using MosaicExpress. This is because MosaicPro allows you to manipulate more aspects of the mosaicking process.

MosaicExpress:

I experimented with the MosaicExpress tool to create a mosaicked image of the Eau Claire/Chippewa Valley region. This tool successfully created a mosaic without much effort, but because it did not allow you to correct the colors or choose brightness values at the overlay area, the mosaic has a noticeable overlay region and the images do not blend together.

Band Ratioing:

NDVI (Normalized Difference Vegetation Index: NDVI=(NIR-Red)/(NIR+Red)) was performed on an image of Eau Claire in 2011. White portions are a higher value that indicate a higher difference between the NIR and red bands because of differential absorption and reflection. This represents healthy vegetation because photosynthesizers absorb the red light and reflect much of the NIR.

Image Diferencing to Detect Change:

I assessed the pixel differences between 1991 and 2011 in Eau Claire by differencing the images with the Two Input Operators tool in the Two Image Functions option. Only layer 4 was processed. This image's histogram was stretched to be able to incorporate more brightness values in the image and the upper and lower change-no-change threshold values were used to calculate the difference between Eau Claire in 1991 and 2011 in  ModelMaker.

Lab 4

Goals: 

The purpose of this lab exercise was to develop skills in image enhancement and pre-processing techniques as well as the methods for delineation of areas of interest. This was achieved by sub-setting images through use of inquiry boxes and area of interest tools, altering spatial resolution of images for ease of interpretation and analysis, enhancing radiometric quality of images, linking Google Earth and Erdas 2013 viewers in order to understand features with the use of Google Earth as a selective interpretation key, and applying re-sampling methods.

Methods: 

Using inquiry box tools and the delineation of areas of interests in a shapefile, I created smaller subsets from satellite images. To enhance the spatial resolution of satellite images, I pan-sharpened the image using resolution merge. For further understanding of image enhancement for analysis purposes, I resampled up by first using the nearest neighbor method and then using the bilinear interpolation method and increased spatial resolution by reducing pixel size from 30m x 30m to 20m x 20m. I used radiometric enhancement to reduce haze on another image using the "haze reduction" tool on Erdas 2013. 

Results: 

Pansharpened Image with Resolution Merging  Original image is displayed on the left; pansharpened image is displayed on the right

Resampled Images Using Different Methods

Resampled images using nearest neighbor (top right) and bilinear interpolation (bottom right). The original image is displayed on the left. 

Haze Reduction via Radiometric Tools

Original image (left), Reduced haze image (right).