Spectral Reflectance

INTRODUCTION

The following lab demonstrates the ability to measure and interpret spectral reflectance signatures of Earth surface features and materials and to perform basic monitoring of Earth resources using remote sensing band ratio techniques.  

METHODOLOGY

Spectral Signature Analysis

All spectral images were collected from a Landsat ETM+ aerial image of Eau Claire, WI. Using ERDAS Imagine, all spectral images were first digitized, organized by the signature editor, and analyzed using the signature mean plot. The first surface feature digitized was Lake Wissota to give a general overview of the process. The list of additional features collected were:

• Standing water
• Moving water
• Riparian vegetation
• Crops
• Urban grass
• Dry soil
• Moist soil
• Rock
• Ashalt highway
• Airport runway
• Concrete surface

Resource Monitoring

To monitor vegetation health, the normalized difference vegetation index (NDVI) was used which subtracts the red band from NIR divided by the sum of the red band and NIR. The output image was then opened in ArcMap where the gradient was switched to an equal interval 5-class classisification system map. 

The same process was repeated to monitor soil health, except using the ferrous minerals equation (MIR/NIR) instead of the NDVI. 

RESULTS

Spectral Signature Analysis

Figure 1 shows the signature analysis process in ERDAS Imagine. The signature editor on the left displays all areas digitized while the mean plot on the right allows for analysis of the digitized feature.

Figure 1: Lake Wissota digitized with ERDAS signature editor and signature mean plot

Figure 2 shows the signature mean plot for moving water collected by Putnam Rock on the Chippewa River. Water features absorb most energy and have low spectral reflectance as the wavelength increases. The reason for higher reflectance in the blue band is because of the shorter wavelength and the resulting scattering. 

Figure 2: Signature mean plot for moving water in Eau Claire

Figure 3 shows the signature mean plot for a forested area in Eau Claire. Again the blue band has a higher reflectance in the visible light spectrum because of scattering. The reflectance peaks at the NIR band because the vegetation has absorbed all of the energy needed for photosynthesis in the visible spectrum and begins reflecting instead of absorbing.

Figure 3: Signature mean plot for forest area in Eau Claire

Figure 4 shows the signature mean plot for both dry and moist soils in Eau Claire. The largest difference in spectral reflection occur in the visible light spectrum. This is because of the moisture content in the moist soil. At the NIR wavelength, however, the water no longer has the same effect and when analyzing soil health, it's essential to collect data in the first three bands to provide easy identification.

Figure 4: Signature mean plot for dry and moist soils in Eau Claire

Figure 5 shows the signature mean plot for all collected features while Figure 6 shows these features listed in the editor.

Figure 5: Signature mean plot for all features collected in Eau Claire

Figure 6: Signature editor displaying all features collected in Eau Claire

Crops, dry soil, and airport all show similar trends of dipping reflectance on band 4, increasing to band 5 and then decreasing again on band 6.

Forest, urban grass, and moist soil all show similar trends of slowly decreasing from band 1 to band 3, followed by an increase on band 4 and a decreasing reflectance to band 6. All of these have moisture in them that reflects energy similar.

Concrete surface, standing water, and moving water all show a similar trend of slowly decreasing reflectance from band 1 to band 6. This could be because they all have very little reflectance throughout all wavelengths.

Asphalt and rock are very similar in that the reflectance is maintained from band 1 to band 4 and then peaks a little at band 5. This could be because

Concrete has a unique spectral signature. This is because of its unique composition of material.

Resource Monitoring

As shown by Figure 7, the spectral reflectance of the NDVI was re-classified in ArcMap to produce a map showing vegetation health in Eau Claire. This is a cheap and efficient way of vegetation monitoring and can be used for a vast number of purposes. The white in the image indicates an abundance of forest, while the black indicates areas that vegetation doesn't exist. 

Figure 7: Map showing vegetation health in Eau Claire

Figure 8 then shows the same thing but instead for soil monitoring. The white in the image indicates where most ferrous minerals are located and indicate man-made structures that consist of concrete, asphalt, and other materials.

Figure 8: Map showing soil health in Eau Claire

SOURCES

Satellite imagery is from Earth Resources Observation and Science Center, United States Geological Survey. 

Stereoscopy, Orthorectification, & Calculations

INTRODUCTION

The following lab demonstrates knowledge in stereoscopy and performing orthorectification. Key photogrammetric tasks were performed on aerial photographs and satellite imagery to understand calculations of photographic scales, measurements of areas and perimeters, and calculating relief displacement. First, measurements were made from aerial imagery to calculate scales, areas, and relief displacement. Next, anaglyph imagery was created using a digital elevation model (DEM) and a digital surface model (DSM). Finally, orthorectification was performed to produce a planimetrically true orthoimage using ground control points (GCPs).

METHODOLOGY

Calculations from Aerial Imagery

Scales

To calculate scale from a nearly vertical image, focal length is divided by the flying height of the aircraft. Flying height is calculated from the altitude above sea level (ASL) of the exposure station minus the elevation of terrain.

Areas

To calculate measurement of area on aerial imagery, the area is digitized using a measurement tool which gives you an output of desired units. 

Relief Displacement

To calculate relief displacement, the height of a tall object is measured and converted into a distance using the aerial image's scale. This value is then multiplied by the distance from that object to the principal point of the object. This product is then divided by the height of the camera above the local datum of that image. 

Stereoscopy

To correct for relief displacement and better view height and changes in elevation, anaglyph imagery is created for a 3D effect. Original imagery of Eau Claire with relief displacement was combined with a digital elevation model at a 10-meter spatial resolution. Next, the same function was performed with a LiDar-collected, 2-meter spatial resolution digital surface model. 

Orthorectification

To correct for spatial anomalies, orthorectification was performed on two aerial photos to produce a seamless transition between the two. Both images were uploaded to IMAGINE Photogrammetry Project Manager. The point measurement tool was then activated to collect ground control points between each image and a separate reference image without spatial anomalies. When matching GCPs, the coordinates between images were within 10 meters of each other to maintain accurate corrections. After matching three GCPs, the automatic drive was selected allowing the program to estimate matching GCPs on the reference image.

After 12 GCPs were collected for both images based on the reference image, automatic tie point collection was performed to match GCPs collected from both input images. Triangulation was then performed to match the tie points collected in the overlapped area of the two images. Finally, ortho resampling was performed to correct the original spatial anomalies of both images. 

RESULTS

Calculations from Aerial Imagery

Scales

If the distance between two points were measured by an engineer's chain to be 8,822.47 ft and according to aerial imagery, the distance measured 7 inches, the scale of the image would equal 1:38,498.

If an aircraft took aerial photography at 20,000 ft. ASL of Eau Claire with a 796 ft. ASL elevation, with a focal lenth of 152mm., the scale of the photograph would equal 1:38,509.

Areas

Imagery from ERDAS was carefully digitized and automatically given an outputted area measurement according to the polygon drawn. 

Relief Displacement

The height of the smokestack in the image measured 0.5 in., which coverts to 133.71 ft., using a scale of 1:3,209. The 133-foot smokestack was then multiplied by the distance to the principal point of the image (10.5) and divided by the height of the camera above the local datum of the image (3, 980). As a result, the calculated relief displacement of the image is 0.35 away. 

Stereoscopy

The original Eau Claire aerial image shows clear relief displacement that shows distortion within the image. This can be observed in Figure 1 by the tall buildings of Towers North, a tall campus building located in Eau Claire.

A stereoscopic view of Eau Claire was then produced using a 10-meter spatial resolution digital elevation model which gives an enhanced view of depth and elevation using three dimensions. This is illustrated in Figure 2 and 3. However, polaroid glasses are needed to view the imagery in 3 dimensions.

Figure 2: Input photos of relief displacement and 10-meter DSM

Figure 3: Output stereoscopic image from Figure 2 input images

The same process was repeated, but instead a LiDar-collected DSM at 2-meter spatial resolution was used for the anaglyph image. The results showed a much more in-depth representation of elevation change and height of surface features. 

Figure 4: Enhanced anaglyph photo using LiDar-collected DSM

Orthorectification

While collecting matching GCPs between both input images and the spatially-accurate reference image, all coordinates were within 10 meters to preserve an accurate orthorectification as shown by Figure 5.

Figure 5: Collecting GCPs from input image to a spatially-correct reference image

After collecting all GCPs for both input images, a summary is shown in Figure 6 of the automated tie points generated in the Photogrammetry Project Manager.

Figure 6: Automatic tie point collection between both images and the reference image

The resulting tie points for each input image is shown in totality by Figure 7.

Figure 7: All tie points gathered between the two input images

Figure 8 then shows an overview of both images after performing triangulation calculated based on the tie points generated.

Figure 8: Triangulation from tie point collection

Finally, Figure 9 shows the orthorectified image after being resampled. Both inputted images now have all spatial anomalies orthorectified to produce a seamless transition and accurate spatial data.

Figure 9: Spatially-accurate orthorectified output imagery

SOURCES

National Agriculture Imagery Promgram (NAIP) images are from United States Department of Agriculture, 2005.
Digital Elevation Model (DEM) for Eau Claire, WI is from United States Department of Agriculture Natural Resources Conservation Service, 2010.
Lidar-derived surface model (DSM) for sections of Eau Claire and Cheippewa are from Eau Claire County and Chippewa County governments respectively.
Spot satellite images are from Erdas Imagine, 2009.
Digital elevation model (DEM) for Palm Spring, CA is from Erdas Imagine, 2009.
National Aerial Photography Program (NAPP) 2 meter images are from Erdas Imagine, 2009.