The article I read (title and link below) discusses the use of a Python script to correct distortion errors in hyperspectral imagery collected by aircraft.
Jensen, R.R., Jackson, M.W. and Lulla, V., 2008: “Single line correction method to remove aircraft roll errors in hyperspectral imagery.” Journal of Applied Remote Sensing 2, pp.1-10.
I found this article to be very interesting and pertinent because it ties together our Remote Sensing course (which I took last fall semester) with the use of Python programming. The problem being addressed here with a Python algorithm is fairly simple. Because planes collecting aerial imagery are subject to pitch, roll and yaw as they follow their paths, any data collected for more than just a single instantaneous frame is prone to distortion. Hyperspectral data is collected in what is referred to as the “push-broom” method: the sensor instantaneously collects a single continuous strip of data at a time, to each side of the plane, perpendicular to the flight path. Any amount of roll by the plane will change the relative view angles along the data strips, and cause distortion in the resultant image.
The usual way to correct this distortion is called the reference line method: the wavy, distorted image of a feature known to be linear on the ground (like a straight road) is digitized. The program then calculates and implements the offsets needed in each row of pixels to make that feature appear straight on the image. This process, though simple in theory, is very meticulous, as it has to look at every single row of pixels in the image. That makes it an excellent candidate for automation by Python scripting.
The correction was accomplished by exporting the coordinates of each pixel in the wavy reference line into an ASCII file. The inputs for the Python program that will correct the distortion are:
• the uncorrected/distorted imagery file, and
• the ASCII file containing the pixel coordinates of the reference line.
The variables used by the program are also very simple:
• the range of column values for the pixels reference line, and
• the average column value for the entire reference line.
The Python script then just offsets each row by the difference between the column value and the average. The effect of this is to straighten out the image of the linear feature, so it appears straight on the corrected image. The positions of the rest of the features in the image are corrected along with it.
Here is a graphic explanation of how pixel rows are offset using a simple Python script to correct the wavy distortion caused by aircraft roll.
 |
| Example of Distortion Correction from Jensen, et al. (2008) |
