Over the past few years, photogrammetry-based drone platforms have been exploding in popularity across the surveying industry. These flying cameras have improved all three points of the “engineering triangle.” They have made surveying faster, cheaper, and provided more data outputs. Many companies have emerged that are taking advantage of these developments; from small businesses that perform drone flight and photogrammetry, to high tech software companies that are pushing the limits of what is possible with photogrammetry. However, this technology has limitations, especially when trying to see the ground through vegetation. In “Assessing the Value of UAV Photogrammetry for Characterizing Terrain in Complex Peatlands”, Julie Lovitt performs a survey in the Peatlands of Canada to study the errors in the results of photogrammetry. The study identifies:
“vegetation cover negatively affect [photogrammetry] results. Vertical accuracies fell from 17 cm (RMSE) in bare areas to 78 cm in areas of vegetation, due to the inconsistent ability of passive photography to penetrate the vegetation canopy . Similarly, elevated wind speeds have been shown to decrease UAV data quality in vegetated areas due to movement of the canopy surface (increased photo blurriness), and interference with UAV positioning and orientation during image capture”
MDPI Open Access Source
Near a meter of ground error does not work for even the most basic surveying jobs. In areas of dense vegetation this has prevented surveying companies from adopting the technology. For example, Submar, a company that provides solutions for soil erosion, has run into this issue:
"In my business we have been able to successfully implement drones using photogrammetry in states west of the Mississippi River. However, in my operations east of the Mississippi we have not invested in a drone fleet due to the thick vegetation making photogrammetry for ground topographies impossible."
- Kevin Foley CEO Submar
Submar is just one example of a company implementing photogrammetry technology being limited by the technology itself. Looking at a map of the US Geological Survey of Vegetation it can be seen that this problem is keeping nearly a third of the land area in the United States from using drones for surveying. A third (or more) of the country not being able to adopt and implement the efficiency of drone surveying platforms means surveying companies are spending more time and more money getting the data they need.
Ultimately, the problem comes from limitations to how much information you can draw from collection by camera alone. Since a camera is a 2D sensor, what it can see is limited to a pixel and whether it can then match that pixel to a pixel in another photo. What the industry requires is a sensor that collects in 3D, so as to not run into this problem with vegetation; specifically, Lidar. Lidar sensors send out pulses of light that can penetrate a canopy, similar to how sunlight penetrates a canopy, it can then derive depth by reading the light pulse’s time of flight. The light pulses are sent out millions of times per second and measured, a 3D collection of measured points is then generated, this is called a point cloud. When point clouds are collected over vegetation, they include both the ground and the bush. This is superior to a photograph, where only the bush would be visible.
At Apollo Robotics, we take the point clouds collected on our aerial platform and, using software onboard the drone, we remove points that are brush, leaving just the ground, as seen below. The new results are accurate ground topography, in record time, that the surveying industry can take advantage of in almost any environment. We are glad to bring this fantastic technology to an industry that provides data for civil design and is always looking to get a technological edge in data, speed, and results.