As with traditional airborne laser scanning and camera systems, with drone or sUAS based LIDAR and camera systems it is necessary to obtain a good project calibration. The proper calibration of a system allows pieces of collected data, whether that be image frames or flight lines of laser data, collected over the course of the project to be merged together as if they were collected at the same time by one giant sensor. This requires a good relative calibration, where the data from one flight line to another, or one image to another matches well. Terrasolid’s TerraMatch module has evolved over the years to handle various scenarios and assist users in obtaining the desired calibration.
During the calibration workflow we wish to remove the systematic biases. Traditionally, that means removing errors that manifest themselves as rotations in the orientation of the sensor, or vertical offsets. Drone based systems are similar to traditional aerial systems, but have a few additional considerations due to their relatively lower grade position and orientation systems, and relatively small areas of coverage. The main observations are greater temporal variations in position and orientation, and often lack of suitable features for matching.
To identify and correct for these biases one must find features that can be used to observe the relative mismatches of the data set and then through a process of minimizing the mismatches provide the best values to correct the data. TerraMatch accomplishes this through its tie lines process. Observing vertical sections of data distributed in a gridded fashion across the project, then fitting a line of best fit to the individual line and, if desired and applicable, scanner combinations found therein. In Figure 1, the views on the right show both the laser data and the tie line observation colored by flight line.
Figure 1 – Tie Line observations of mismatched flight lines
These tie line observations allow for a quick way to observe the state of the data as we move through the calibration workflow. The tie lines should also be used to determine if a suitable number of features exist in the right combinations and locations to have confidence that TerraMatch can find suitable correction values for the different calibration parameters. In a dataset, like this quarry in Figure 2, we can use the Tie Line Display Mode to see that we have a good distribution of features suitable for observing relative XY positioning not only for the dataset as a whole, but also each flight line, and importantly, over time along each flight line.
Figure 2 – Good XY feature distribution in the quarry project
Whereas, in this second dataset from Japan there is not a good distribution of XY suitable features observable in the laser data. They are either concentrated on one side of the flight line or near the beginning or end of the lines.
Figure 3 – Project without a good distribution of observable XY features in the laser data
Why is this important? We need to keep the features in mind as certain solvable parameters require sloped features in the correct orientations to find a reasonable correction value. For instance, pitch and heading require slopes in the direction of flight. For X and Y, in the scanner reference frame, not the map, one needs slopes in those respective directions.
Figure 4 – TerraMatch Solvable Parameters
The observable features will help one analyze the data and modify the applicable permutations of solvable parameters. For instance, the standard sUAS project calibration workflow is:
When solving for the applicable parameters it is best to understand what errors one is observing and if there are suitable features for correcting those errors. For example, in the dataset above from Japan, one can observe fluctuating heading in the laser data. However, since heading requires a suitable number of XY features throughout the dataset to correct for this then the laser data on its own is not suitable. Fortunately, one can use the imagery flown concurrently with the laser data and the TerraPhoto workflow to determine the heading fluctuation which is applicable to the laser data. The correction determined from the imagery can then be applied to the laser data using TerraMatch’s Apply Correction tool. This is just one example of many scenarios that can be encountered.
Figure 5 – XY Features visible in imagery
In addition, since MicroStation displays features by elevation, the tie lines can provide a suitable visible analysis akin to that observed by the distance coloring by Line Average Z option. See the article on Recognizing Misalignment Patterns for Airborne LIDAR System Calibration.
For instance, in Figure 2, note the characteristics of the tie line display indicate a mismatch with the data as we see sloped surfaces displaying dominance of one flight line on one side of the feature and another on the other side indicating a horizontal mismatch. In Figure 6, the phase one correction has been applied to the tie lines and now we see the tie lines show more of a flight line pattern.
Figure 6 – Tie lines with phase 1 correction applied
Solving the “by flight line corrections” of phase 2 and applying them to the tie lines results in Figure 7, where the tie lines show the by line mismatch has been removed, and we now see more of a fluctuating change in the dataset.
Figure 7 – Tie Lines with the phase 2 corrections applied
Finding a suitable fluctuations correction, as shown in Figure 8, results in a better mix of tie lines from each flight line being visible along the entire length of the flight line and no longer highlighting features.
Figure 8- Tie lines with phase 3 fluctuations correction applied
TerraMatch is a very flexible tool for analyzing and correcting systematic issues. Multiple beam/scanner alignment, flight line to flight line alignment, fluctuating misalignment corrections, and vertical bias removal can all be addressed by the Terrasolid workflow. We regularly conduct onsite training for experienced users who have added drone based LIDAR systems to their portfolio, as well as for new users who are just getting started and need a good piece of software for calibrating and processing their sUAS collected data. Contact us if you’d like more information and training on this workflow.