Part 4: Theoretical Core of Image Processing that Interprets Drone Imagery
Absolute Orientation
Just to give you the hint of the series that I’m working on; I’ve covered the core theoretical concepts of image processing. This article is in continuation of the same. Let’s start with defining Absolute Orientation and why it is a crucial part of image processing?
What is Absolute Orientation?
Absolute Orientation introduces the relationship between image coordinate and the corresponding ground coordinates. In this process, the model is scaled and leveled as per the reference datum. This method is used to mark local points in reference to the global points. Each pixel in the imagery is scaled and leveled in such a manner that it prepares a model for the DEM which supports required GSD of the orthomosaic in measuring the various features in the imagery.
Absolute Orientation is understood by the following two processes:
- Scaling the Model
Scaling the model can be understood as scaling the photo/imagery to the original distance on the ground. The model is based on Relative Orientation model. If the X and Y setting of the two projections are equal then the model base is composed only of an X component called bx (image coordinates) and model scale is varied by simply changing the model base by the change in bx.
A minimum of two horizontal control point is required to scale a stereo model. These points are plotted in the adopted model scale on the manuscript map. The manuscript is then positioned under the model and with the floating mark set on one model. It is rotated until it maps the same line that is collinear with the model line on the imagery. The model scale is too small and must be increased by increasing the model base until the line on the model is equal in length to the line on the manuscript map. If the Y and Z setting of the projections are not equal then projectors will have to be moved appropriately.
2. Leveling the Model
This procedure requires a minimum of the 3 vertical control points distributed in the model so that they form a large triangle. A model with vertical control points near each corner that has not been leveled. To stable the model generally four points are preferred, each marked near to the corner in the model, and a fifth point near the center of the model is also desirable. Note that there are two components of tilt in the model. The amount by which the model is out of level, each of these components are determined by reading model elevations of the vertical control point and compose them with their known values.
For correcting X and Y the projector bar is tilted to distribute the relative orientation error.
Thank you, for reading my article. Please, follow the publication for updates. Also, benefit from the shared public projects on Indshine.
