Background
The setting for this project is a water treatment plant located in Southern California. The project setting is very “photogrammetry-friendly.” Meaning that the project area-of-interest (AOI) lends itself well to photogrammetric processing. LiDAR is, and almost always will be, a much more robust survey tool. This is especially true in vegetated areas and when trying to detect finite objects like wires, conductors, poles, comm lines, etc. However, the AOI for this project has lots of hard top, structures, and completely lacks vegetation. The project does not require any finite objects to be accurately detected either. Both LiDAR and photogrammetry should produce fantastic results.
For this observation, we performed both LiDAR and photogrammetry in the way we deemed was most economical, effective, and practical. This means that there are some extraordinary circumstances where the results of the LiDAR and/or photogrammetry could potentially have been improved. For example, there definitely are better photogrammetry sensors to deploy. In this case, we deployed a simple Phantom 4 Pro, which houses a 1 inch sensor. This is a very common platform for average service providers. Similarly, we used flying techniques, altitudes, and speeds for LiDAR that are conducive to efficiency, rather than methodologies that favored unrealistic results.
What Makes Accurate Data?
There is no secret formula in creating high quality accurate data, regardless of whether it is LiDAR or photogrammetry. Both systems require an in-depth working knowledge of the acquisition and processing workflows.
You can send acquired data to two separate processing teams, and come back with two wildly different products. LiDAR processing requires an intricit knowledge of boresighting, calibration, and geodesy/kinematics that isn’t necessarily automated. Similarly, survey methodology and implementation requires an understanding that doesn’t make either workflow necessarily “plug and play.” An aptitude for datums, geoid corrections, projections, transformations, and how they affect each other is essential.
Understanding the processing methodology is a big difference when comparing remote sensing service providers. Acquisition isn’t necessarily difficult, but processing (especially in LiDAR), can show the true colors of a someone providing a survey grade product.
Equipment:
LiDAR
- UAV: DJI M600 Pro
- Scanner: Riegl VUX 1 LR
- IMU: KVH 1750 Fiber Optic Gyro
- Integration: Phoenix LiDAR “Ranger”
- Reference Specs: L1C/A, L1C, L1P, L2 P(Y), L2C GLONASS: L1, L2 BeiDou:
- Usable Flight Time: 10 minutes
- Unit Cost: $230,000
Photogrammetry
- UAV: DJI Phantom 4 Pro
- Sensor size: 1 inch
- Effective pixels: 20M
- Lens: FOV 84° 8.8 mm/24 mm (35 mm format equivalent) f/2.8 – f/11 auto focus at 1 m – ∞
- Usable Flight Time: 18 minutes
- Photo Mode: JPEG, DNG (RAW), JPEG + DNG
- Unit Cost: $1,500
The biggest difference between the two setups is cost. The total in the air cost for the LiDAR “Ranger” system is roughly $230,000. On the other hand, the DJI Phantom 4 Pro is an off-the-shelf item with a sticker price of $1,500.
Acquisition
A mission of this size is light logistically and only required a two-man crew. Prior to arriving on-site, our survey partners established a control network with control, correction, and check (truthing) points. Correction points were dual purpose, serving as visual photogrammetry targets, as well as z-axis LiDAR corrections. Targets were also reflective, allowing us to use them as “LiDAR-identifiable” corrections for LiDAR horizontals.
For LiDAR, GNSS corrections were applied with a PPK (post processing kinematics) solution. The other methodology is RTK (real-time kinematics), which communicates these live corrections during flight. A major goal of this project was high accuracy. For this reason, RTK was not used.
Photogrammetry flights were flown with double nadir grids with 90% front overlap, and 80% side overlap. This resulted in a flying height of 200 feet.