Acquisition

Prior to going out in the field and flying, we needed to diligently plan acquisition heights and flight speeds to ensure that proper density and accuracy requirements were met throughout the area of interest.

Density is naturally a function of speed and scan rate. The faster the flying speed, the less dense the data will be. Flying 13 knots (15mph) and at 50 meters should yield roughly 175-200 points per meter squared. On this particular project we were able to achieve 175 points/meter^2 which is much higher than any LiDAR standard. This higher density also helps with accuracy. The more points you have, the less interpolation is required in post processing.

Before flying, our base station needed to be on a landmark (known survey point) or in a flat area in where we could create our own static point. Being in a flat area helps with Z (height) correction. The base station is used in initial post processing for trajectory calculation. If the base station is also in the data, it can double as a ground control point. If our goal is absolute (global) accuracy, the base station runs for more than 6 hrs for an OPUS CORS solution. This gives us a ground point more accurate than our raw LiDAR data. This is generally in the range of 2-3mm, while our LiDAR scanner has a relative accuracy of 2-3 cm.

In order to create accurate and dependable data, it’s important to have a QA/QC plan in place. Part of our QA (Quality Assurance) plan was to check GNSS and ephemeris information days before our flights began. We wanted to make sure our PDOP (Position Dilution of Percision) values are well below 2. The Trimble GNSS Planning Tool is a great online resource for planning out satellite constellation and GNSS coverage.

GNSS Planning Iono Map

GNSS Planning DOP Plot

Initial Processing

After data acquisition, the flight and INS data are compiled to create a trajectory of the flights. This information is run through Inertial Explorer, a software program that calculates the position, velocity, and attitude of the aircraft at all times. Since UAVs are airborne, GNSS coverage is almost always excellent. But in the case of GNSS outage or delay, Inertial Explorer will use the onboard IMU to fill in the gaps of the trajectory. This is why having a very capable IMU system on the UAV is important. Fiber Optic IMU’s are preferred, while the MEMs (mechanical) versions are acceptable. The STIM300 is a MEMs IMU with a great following and reputation.

For the entire processing workflow, it is important to keep datums, projections, geoids, and units consistent. One error in conversion or in an output setting can cause a big headache down the road! At some point you’ll notice your data isn’t accurate to the level that was required, and it won’t be exactly obvious where the mistake was made. Going slow and double checking settings was critical in our processing procedures.

Inertial Explorer Sample Data

Once the trajectory was calculated and it’s information confirmed, we were able to combine it with the raw LiDAR scanner data to output point cloud data. In this step we also added in our ground points and parameters for what points are included and excluded. There are lots of options for what information to keep and throw out. During this step it was important for us not to add excess noise to the data, especially in the far ends of the swath.

Creating a point cloud from raw scanner data and Inertial Explorer trajectory

Point Cloud

The point cloud data below is unfiltered, unclassified, and not yet cropped. For a final deliverable, most often a client will require a fully classified point cloud with noise and erroneous points removed. Fully classified point clouds can be imported with little complication into programs like PLS-CADD for further analysis. They can also be filtered and manipulated to create bare earth models like DEMs and contours.

Optical imagery was also acquired for a digital orthomosaic. The orthomosaic can be viewed alone, or can be used to colorize the point cloud and add RGB attributes. At times, LiDAR data can be difficult to visualize. RGB attributes are great for spatial awareness and visualizing data.

The video below shows the process of colorizing LiDAR point cloud data with RGB attributes. This can be done most effectively with a digital orthomosaic. However, Google Earth can also be used for situations in which RGB precision isn’t as important.

Conclusion

Our work flow was fairly straight forward, and the project went as planned. Often times, you can mitigate headaches and errors in the field just by having proper project planning and clearly defined objectives. In this case, we knew the exact accuracy and density requirements as well as the final deliverable. That meant we were able to plan our flights, gather the data, and process it all quickly and efficiently. You can see our final results/numbers below.

Project Results

The final deliverables were tiled LAZ point cloud files that were fully classified. Ground control was added for a Z correction, and exported in the State Plane Coordinate System.

Specifications Summary

Length of Corridor: 2 miles
Number of Points: 108,936,112
Point Density: 179.58 points/meter^2
Point Spacing: 0.07 meters
Relative Accuracy: +/- 2 cm
Z Absolute Accuracy: 0.0876 meters