Geometric Correction

Introduction:

When remotely sensed images are collected, certain factors can cause the location of the pixels to be different from their actual locations on the earth’s surface. This distortion is fixed through geometric correction. Geometric correction is a process of image transformation using known locations on the earth’s surface called ground control points.

Methods:

The first method I used to correct an image was “image-to-map rectification. I used a  USGS 7.5 minute Digital Raster Graphic topo map to correct a multispectral satellite image. I recorded points at corresponding locations in both images, using easily identifiable landmarks like levees or intersections. After four of thee points were collected polynomial transformation was calculated to correct the multispectral image.

The second method I used to correct an image was “image to image registration”. When correcting the second image, a second image with the same extent was used.  The second image already had ground control points recorded, so I needed to find their locations on the second image to record their locations on the image. After 10 points were recorded, a 3^rd order polynomial transformation was performed.

Figure 1: The results of the second image correction.

Results:

Geometric correction allowed for non-corrected images to be used. It will be an extremely useful skill to know as a remote sensing professional, and I’m sure it will be used often.

LiDAR

Goal:

The goal of this lab was to gain a basic knowledge of how LiDAR data are structured, and how to process them. This was done by creating a number of surface models and an intensity image.

Methods:

In order to view the LiDAR as a point cloud I created a Las Dataset using ArcMap (Figure 1).  Next, I created a Digital Terrain Model (DTM) showing just the LiDAR points classified as "ground". This was done by filtering the Las dataset to show just "ground" points, then using the "LAS Dataset to Raster" tool to output a 2m surface raster (Figure 1). After the DTM was created, I generated a hillshade to emphasize the relief (Figure 1). Next, I created a Digital Surface Model (DSM) by filtering the Las dataset to just show the "First Return" points, and using the "LAS Dataset to Raster" tool to output a 2m raster (Figure 1). I made a hillshade of the DSM, in order to emphasize the surface's texture (Figure 1). The next stage was to create an intensity image, which was done using the "LAS Dataset to Raster", with "Intensity" as the raster value (Figure 1).

Figure 1: The data flow model

Results:

The Las dataset allowed for visual interpretation of numerous features that aren't observable from satellite remote sensing. This is due to the Las dataset's 3D visualization (Figure 2).

Figure 2: The Las dataset allows for observation of the bridge's structure,
something not possible with traditional remote sensing methods.

The DTM and its hillshade allowed for the ground surface to be seen without the distraction of buildings or vegetation. The small pixel size allowed for the interpretation of small fluvial features, such as the lower section of Otter Creek (Figure 3).

Figure 3: Otter Creek

The DSM and its hillshade aided visual interpretation of buildings and vegetation features (Figure 4).

Figure 4: The DSM with Hillshade

The intensity image proved to be a fantastic tool for visual interpretation, as vegetation features and buildings have very high contrast when compared to the ground surface (Figure 5).

Figure 5: The intensity image of UW - Eau Claire's campus

Conclusion:

LiDAR is an extremely useful type of data, as it has an extremely vast range of applications across disciplines. Also, the high density at which it is captured allows for its derivatives to be produced at extremely high spatial resolutions.