Lab 8: Spectral Signature Analysis & Resource Monitoring

Goal

The goal of this lab is to become familiar with spectral reflectance signatures of different materials on and near the Earth's surface. Health monitoring of vegetation and soil will also be introduced using NDVI and ferrous mineral indices.

Methods

Part 1: Spectral Signature Analysis

Spectral signatures can be collected either in the field or from software like ERDAS Imagine. In this lab, the spectral signatures from the image below of Eau Claire and Chippewa counties will be taken using ERDAS Imagine.

Figure 1: Image from which spectral signatures are taken in this lab

First the polygon tool was used from the drawing menu to create a small polygon completely within the water (not including any land edges) of Lake Wissota. This created an area of interest (AOI) from which a signature could be taken shown in Figure 2. 

Figure 2: Polygon created as AOI in open water of Lake Wissota

The "Supervised" window was opened and "Signature Editor" was selected. The "Create New Signature from AOI" was selected and the signature from the polygon above was automatically added to the list. The name and color were changed to reflect the source of the signature.

Figure 3: Signature Editor

This process was followed for running water, deciduous forest, riparian forest, coniferous forest, crops, dry soil, moist soil, rock, asphalt highway, airport runway, and parking lot for a total of 12 signatures.

To display a chart, "Display Mean Plot Window" was selected from the Signature Editor. Chart options were changed so that chart background was set to white for better visibility of the signature. 

Part 2: Resource Monitoring

Section 1: Vegetation Health Monitoring

Vegetation health monitoring is often done using the Normalized Difference Vegetation Index (NDVI). The equation for NDVI is:

 

The NDVI for this lab was done on the image of Eau Claire and Chippewa Counties below: 

Figure 3: Image of Eau Claire and Chippewa Counties used for NDVI

To use NDVI in ERDAS Imagine, it is selected from the dropdown menu for "Unsupervised" in the Raster tab. The specifications for this lab are shown in Figure 4.

Figure 4: Specifications for NDVI

A map was then created and symbolized appropriately in ArcMap, which can be seen in the results section.

Section 2: Soil Health Monitoring

Soil health monitoring is often done using the Ferrous Mineral Index (FM). The equation for FM is:

The same image in Figure 3 was also used for the Ferrous Mineral Index. To use FM in , it is selected from the dropdown menu for "Unsupervised" in the Raster tab. The specifications for this lab are shown in Figure 4.

Figure 5: Specifications for FM

A map was then created and symbolized appropriately in ArcMap, which can be seen in the results section.

Results

The spectral signatures for 12 individual features are shown below. Band 1 corresponds with blue, band 2 with green, band 3 with red, band 4 with NIR, and bands 5 and 6 with MIR.

Figure 6: Spectral signature of standing water

Figure 7: Spectral signature of running water

Figure 8: Spectral signature of deciduous forest

Figure 9: Spectral signature of coniferous forest

Figure 10: Spectral signature of riparian forest

Figure 11: Spectral signature of crops

Figure 12: Spectral signature of dry soil

Figure 13: Spectral signature of moist soil

Figure 14: Spectral signature of rock

Figure 15: Spectral signature of asphalt highway

Figure 16: Spectral signature of airport runway

Figure 17: Spectral signature of parking lot

In order to compare the spectral signatures of moist and dry soil, they were put onto the same chart using that option. As expected due to the high absorbance of water, there was higher reflectance in the dry soil.

Figure 18: Comparison of moist and dry soil spectral signatures

An even broader comparison was made of all 12 spectral signatures together. They follow what is to be expected. Standing and moving water vary slightly in the visible spectrum but still share similar signatures that decrease in reflectance as wavelength increases. Rock, airport runway, and dry soil have similar signatures, probably because they are composed of similar materials. Riparian forest and deciduous forest are similar, likely because the riparian forest contains mostly deciduous trees. Forest in general is unique in that its spectral signature is low in the visible and spikes in NIR. All other surfaces tend to dip in the NIR, especially dry, rocky surfaces. Water also has a unique spectral signature that has some reflectance in the visible and dips to almost no reflectance in NIR. Other signatures have more ups and downs, especially in NIR. Athough similar, riparian/deciduous forest and coniferous forest are different due to the surface area of the needles/leaves, the shape of the canopy, and the direction of the branches.

Figure 19: All 12 spectral signatures

The NDVI gives insight into areas of high and low vegetation, as seen by the map in Figure 13. Areas of high vegetation are brighter because compared to other spectral signatures, there is high NIR reflectance and low red reflectance. Areas that are medium gray and black have no or almost no vegetation. They are mostly water, which absorbs most of the NIR band.

Figure 13: NDVI map of Eau Claire and Chippewa Counties

The FM shows areas of high and low ferrous mineral content, shown by Figure 14. Most of the surface ferrous minerals are west of Lake Wissota in Chippewa Falls and lie mainly in what looks like uncultivated agricultural fields. Urban areas lack ferrous minerals on the surface and water has low minerals.

Figure 14: Ferrous Mineral map of Eau Claire and Chippewa Counties

Sources

Earth Resources Observation and Science Center, United States Geological Survey. (2019). [Satalllite images of Eau Claire and Chippewa counties]. 

Wilson, C. (2019). Spectral signature analysis and resource monitoring. Geog 338: Remote Sensing of the Environment Lab 8.

Lab 7: Photogrammetry

Photogrammetry

Goal

The purpose of this lab was to become acquainted with using tools to turn distorted aerial and satellite images into photogrammetrically correct images. Mathematics including measurements of areas and perimeters, as well as relief displacement will be used. Stereoscopy and orthorectification will also be introduced.

Methods

Part 1: Scales, Measurements, and Relief Displacements

Section 1: Calculating scale of nearly vertical aerial photographs

The image below was used to practice scale calculation. With a known ground distance (8822.47 ft), the photo distance from A to B can be measured with a ruler to obtain the scale of the photo. The equation looks like:

2.625”/8822.47’ = 2.625”/105,869.64” = 1/40,331.29

S = 1 : 40,331”

Figure 1: Image with known ground distance and photo distance measured with ruler.

Scale can also be determined using flying altitude above sea level, elevation of land above sea level, and focal lens length. The scale of the image below was found by using the following equation:

152 mm/ (20,000-796) ft = 152 mm/19,204 ft = 152/5853379.2 = 1/38509

S = 1 : 38,509 mm

Figure 2: Image with known altitude above sea level, elevation of land above sea level, and focal lens length.

Section 2: Measurement of areas on aerial photographs

Sometimes, the areas and perimeters of features must be measured. ERDAS Imagine make this easy by calculating these from polygons drawn on images, with the tools highlighted and measurements recorded as shown in Figure 3. 

Figure 3: Area and perimeter calculation using the measurement tool in ERDAS Imagine

Section 3: Calculating relief displacement from object height

Relief displacement is caused by distance from the principle point of an image. The relief displacement of the image below was calculated using the following equation: 

0.5” of displacement

Real world height = 3209 x 0.5 = 1604.5”

10.4375” away from PP

RD = (1604.5 x 10.4375) / 3980 = 4.21”

Figure 4: Relief displacement of smokestack (labeled "A"). 

Part 2: Stereoscopy

Section 1: Creation of anaglyph image with the use of a digital elevation model (DEM)

Anaglyphs allow viewers of an image to see a 3 dimensional images with the help of polaroid glasses. One way of achieving this is by using a DEM along with a 2 dimensional photograph. The Anaglyph tool in the Terrain window was used. The specifications below were used in the dialogue box before running the model:

Figure 5: Specifications for DEM anaglyph

Section 2: Creation of anaglyph image with the use of a LiDAR derived surface model (DSM)

Another way of achieving this is by using a DEM along with a 2 dimensional photograph. The Anaglyph tool in the Terrain window was used again. The specifications below were used in the dialogue box before running the model:

Figure 6: Specifications for LiDAR anaglyph

Part 3: Orthorectification

Section 1: Create a new project

The images below were taken by the SPOT satellite and require orthorectification.

Figure 7: SPOT satellite images in need of orthorectification

From the help menu, a search was done for "photogrammetry" and "Photogrammetric Project" was selected to begin a new block file project.

"Polynomial-based pushbroom" and "SPOT Pushbroom" were selected from the dropdown menu for Geometric Model Category.

In the Block Property Setup, the Horizontal Reference Coordinate System was set to the following specifications:

Projection Type: UTM.

Spheroid Name: Clarke 1866.

Datum Name: NAD27(CONUS).

UTM Zone field: 11.

North or South Field:  North.

Axis Order: E,N.

Horizontal Units: Meters.

Section 2: Add imagery to the block and define sensor model

The top image in Figure 7 was added to the block file by selecting "Add".

The parameters of the satellite were specified as "SPOT PAN" in "Interior Orientation". Specifications are shown below:

Figure 8: Specifications for Interior Orientation

Section 3: Activate point measurement tool and collect GCP's

"Classic Point Measurement" was selected from the "Point Measurement Tool." With the point measurement window opened, as shown below, the reference layer was set to a previously orthorectified image and the checkbox for "Use Viewer as a Reference" was clicked. 

Figure 9: Point measurement window with orthorectified image set as reference layer (left)

At this point GCP's could begin to be collected. The Select Point tool was used to move the inquire box as appropriate. The "Add" button was selected for point 1 on the reference image and subsequently for point 1 on the distorted image. X and Y reference coordinates as well as X and Y file coordinates were provided by the lab to ensure accuracy. This process was completed for 9 points.

Another horizontal reference source was brought in for the last 2 points by selecting "Reset Horizontal Reference Source." The same GCP selection process was used for this source for points 11 and 12.

The Point Measurement was saved.

The Vertical Reference Source dialogue was opened and set to a DEM of Palm Springs. All the previously gathered points were highlighted and their Z values were updated to the reference DEM, as shown below.

Figure 10: Z values updated from DEM of Palm Springs

Section 4: Set Type and Usage, add a 2nd image to the block & collect its GCP's

The Type and Usage columns (shown also in Figure 10) was right clicked for each point and updated to "Full" and "Control", respectively, shown below. The Point Measurements were then saved and closed.

Figure 11: Type and Usage columns updated to "Full" and "Control"

The same GCP collection process was done for the second image to be orthorectified shown in Figure 7 by clicking "Add Images" in the contents window. Interior Orientation was set the same as with the previous image. 

Figure 12: Both SPOT images in Point Measurement viewer

A total of 11 points were collected.

Section 5: Automatic tie point collection, triangulation and ortho resample

The Automatic Tie Point Generation Properties icon was selected from the Point Measurement tool. The specifications were set to the following:

General Tab:

Image used: All Available

Initial Type: Exterior/Header/GCP

Image Layer Used: 1

Distribution Tab:

Intended Number of Points/Image: 40

Keep All Points: unchecked

The model was run and tie point accuracy was checked. The Point Measurements were then saved and closed.

The Block Triangulation Properties window was opened. The specifications were set to the following:

General Tab:

Iterations With Relaxation: 3

Image Coordinate Units for Report: Pixels

Point Tab:

Ground Point Type and Standard Deviations: Same Weighted Values

X, Y, and Z: 15

Advanced Options Tab:

Simple Gross Error Check Using: checked

Times of Unit Weight: 3

The triangulation was run and the Triangulation Summary Dialogue was saved.

Start Ortho Resampling Process was clicked to begin orthorectification. The specifications were set to the following:

General Tab:

DTM source: DEM

Output Cell Sizes: 10

Advanced Tab:

Resampling Method: Bilinear Interpolation

Add Single Output was selected and the second SPOT image was set as the input image. Current cell sizes were used. The ortho resampling process was run and completed.

Results

Part 1: Scales, Measurements, and Relief Displacements

Section 1: Calculating scale of nearly vertical aerial photographs

It is relatively rare to have a ground distance measurement to determine scale with. But when available, the photo distance is divided by the ground distance to obtain the scale. The scale of the image for this section was S = 1 : 40,331”.

If a ground distance is not known, scale can also be calculated from flying altitude above sea level, elevation of land above sea level, and focal lens length. The scale from the image was S = 1 : 38,509 mm.

Section 2: Measurement of areas on aerial photographs

The polygon in figure 3 was drawn and the perimeter and area of it were automatically computed. They were (in different units, with could be chosen from a dropdown list):

Area: 38.10 hectares, 94.14 acres

Perimeter: 4116.38 meters, 2.56 miles

Section 3: Calculating relief displacement from object height

The calculation of relief displacement is used to correct those distortions. Using the final RD of 4.21”, it can be determined that it will need to be a negative correction to account for the positive RD.

Part 2: Stereoscopy

Section 1: Creation of anaglyph image with the use of a digital elevation model (DEM)

This type of anaglyph image requires aerial imagery and a DEM. The resulting anaglyph below can be viewed using polaroid glasses. 

Figure 13: Anaglyph using DEM

Section 2: Creation of anaglyph image with the use of a LiDAR derived surface model (DSM)

This type of anaglyph image requires aerial imagery and a LiDAR surface model. The resulting anaglyph below can be viewed using polaroid glasses.

Figure 14: Anaglyph using LiDAR surface model

Part 3: Orthorectification

The orthorectification process for two images taken by a spot satellite was very successful. Below is an image of both images together. Features are aligned well, which can be seen especially well using the swipe tool.

Figure 15: Result of orthorectification of two previously distorted images

Sources

Wilson, C. (2019). Photogrammetry. Geog 338: Remote Sensing of the Environment Lab 7.