Clean orthomosaic imagery and 3D Mesh - Point clouds are visual data that not only enable clear communication during meetings but also deliver excellent results in terms of surveying accuracy. Today, we introduce how to obtain the high-quality 3D point clouds that form the foundation of drone surveying analysis.
First, you should fly the drone in a good shooting environment. Drone surveying can also be described as photogrammetry. The principle of photogrammetry is to extract feature points based on the overlap of photos taken from various angles and build a 3D model from them. Naturally, therefore, the clearer the information that can be obtained from the captured photos, the higher the accuracy of the resulting values.
The more the information of the area to be surveyed is obscured—such as terrain densely covered with trees and vegetation, or terrain with many shadows—the higher the probability that errors will occur in the analysis results. Likewise, places with many reflective surfaces, such as snow-covered ground, lakes, or puddles, can also cause errors for the same reason.
Therefore, capturing well-organized terrain on a clear day, during a time when shadows are minimal, can be a good approach.
You have probably experienced subjects appearing shaky and stretched when moving the camera to capture a panorama or photo. Drones, too (although this varies by model), experience distortion in photos taken during high-speed flight. This is called the jello effect. As explained earlier, if the captured photos contain noise, blur, or distortion, errors can occur in data analysis.
There are a few ways to solve this kind of problem. The first is a hardware approach, changing the drone to one with a camera sensor or mechanical shutter type. The second is a software approach, analyzing by utilizing the metadata acquired during drone capture and correcting displacement values according to the camera specifications.
Meissa can help with the latter approach. Meissa has its own proprietary drone analysis engine, enabling flexible solutions tailored to the data acquired by the drone. You can find more details on this in this article.
Lowering the flight altitude can also be one approach, because it is a way to improve the GSD.
GSD (Ground Sample Distance) refers to the ground sampling distance. Here, the distance is the distance between the center points of two consecutive pixels. For example, if there is an orthomosaic image with a GSD of 1cm/px, then 1 pixel of this orthomosaic image represents an actual size of 1cm. The smaller the GSD, the more accurately the details are revealed in the image.
Unlike pixel resolution, GSD varies depending on the distance between the camera and the subject. GSD improves as the camera gets closer to the object.
Just as we prefer maps with a small scale for finding our way, a small GSD interval is preferred for accurate quantity calculation during surveying. In a previous article, we introduced a general-purpose rotary-wing aircraft surveying experiment conducted at an actual DL E&C site. As a result, in the case of the Z coordinate, we could confirm that the lower the flight altitude, the more the mean error and standard error decreased. This means that the lower the flight altitude, the higher the surveying accuracy.
Of course, if you capture a wide area at a low altitude, the number of photos the drone takes and the time required increase dramatically. Drone batteries at a price that is not burdensome for field use last about 20-25 minutes, and it can be difficult to obtain the desired data within that time.
At Meissa, we set the drone path by focusing on the best altitude and overlap achievable within the limited time. After data analysis, the GSD is fixed at 5cm/px, which is sufficient for practical use. However, you may want a narrower GSD for reporting documentation or other reasons. In that case, denser modeling is possible through a separate analysis via a technical support request.
