Multipoint Geometric Correction
Goals and Learning Objectives
This particular
lab was designed to introduce students to an important aspect of preprocessing remotely
sensed images. The technique of geometric correction is commonly used before
any information is extracted from the image to ensure the quality and accuracy
of the image obtained.
Methods
The preprocessing technique is
multipoint geometric correction is used to resample an image which contains geometric
errors. To correct for these errors, a reference image is with a known scale,
coordinate system, and units. In ERDAS, the Geometric Correction tool is used,
and for the first portion of the lab an image-to-map rectification process is
conducted. A USGS DRG (digital raster graphic) is used to Collect GCP (ground
control points) to rectify the TM image. Ground control points link one spatial
location with the same area on the image that needs correcting. Once the
reference image and the distorted images are imported into ERDAS Geometric
correcting window, a polynomial function is used to correct the image. For the
image of Chicago, a first order polynomial is used. This means we must only
collect 3 pairs of GCP for an accurate rectification, a fourth point is added
to complete the formula and the display resample image tool used to perform the
task. Before the image can be fully corrected, a quality check of the RMSE is
performed to ensure the correction is accurate. For each application there is
an acceptable value for rmse, in this case the RMSE is less then 2 (total).
Image to Image Rectification
This second
portion of the lab uses a very similar process to correct the distorted image.
Excapt this time a 3rd order polynomial function is used, which
means that (9-12) pairs of points much be collected. Once the collection of the
GCP’s is completed, and the RMSE is <1 total, then the Display Resample
Image tool is used. This tool uses bilinear interpolation to execute the
geometric output. Bilinear interpolation is a more computationally expensive
process, but leads to a much more cohesive, smooth output.
Figure 2: Displaying parts of Eastern Sierra Leone, left
image contains a variety of geometric distortions. The right image is the
reference image used.
Conclusions
This lab
introduced students to two very important preprocessing techniques. While collecting
GCP can be very time consuming and tedious, the results speak for
themselves. A quality, geometrically proportional
image is produced which can be used for a multitude of projects. Often times,
images have some level of distortion, and it is the mappers job to correct for
these errors so the output map is of the highest production value possible.
