With drones collecting aerial data has never been easier, allowing us to create 3D point clouds and 3D models. Drones are versatile and can be equipped with different techniques. Usually, two technologies are used to measure objects and landscapes: photogrammetry or LiDAR (3D laser scanning). Photogrammetry is based on aerial photos that can be further processed and combined for mapping. LiDAR measures with laser beams. The exciting question is: Which of the two techniques is more effective? What are their respective advantages and disadvantages? And for which particular applications are photogrammetry and LiDAR suitable?
Aerial Photogrammetry with Drones
Photogrammetry works on the same principle as our sense of sight: although images arrive two-dimensionally on our retina, we have a good view of three dimensions. Similar to how the human brain uses information from both eyes to enable depth perception, photogrammetry uses 2D photos from different angles to create a 3D map. In this way, the nature, shape, and position of any objects can be determined.
More precisely: A drone with a high-resolution calibrated digital camera takes several – or rather a great number of – photos of the object or area to be measured. These images overlap in such a way that the same point on the surface is visible on several photos and from different angles to capture all details. Special software then searches the images for characteristic structures, so-called features, to compute the shape of the object in 3D. A feature is any visually distinct point in an image. Such distinctive points must be rediscovered in many photos, so that the software can estimate their position. This way, high-precision 3D coordinates can be calculated from the two-dimensional camera images in the so-called "point cloud" with several million points.
The result: A high-resolution, realistic 3D reconstruction that contains not only height information, but also texture, shape, and colour for each point on the map, allowing for easier interpretation of the 3D point cloud.
What is LiDAR?
LiDAR is similar in principle to sonar or radar, which use sound or radio waves to measure surfaces and detect objects. LiDAR stands for "Light Detection and Ranging" and usually uses infrared light. Mounted on a drone, the LiDAR sensor sends a laser beam until it hits a surface – a tree, a building, or the ground itself – that reflects the light. The surface is "scanned" line-by-line with a "sheet of light".
The time it takes for the laser light to return to the sensor is recorded, as is the intensity of the reflection. In this way, distances can be measured with high accuracy. With millions of these reflection points, a point-based model of the terrain can be reconstructed: a LiDAR point cloud. This is also the reason why this technology is sometimes referred to as 3D laser scanning – because ultimately it is used to generate a 3D point cloud. If different heights and shooting angles are combined, the result is a detailed, complete, and almost gapless 3D point cloud, which presents even the smallest details. Today's laser measuring systems record in 16-bit grayscale. The result is similar to a black and white photo, but it can be coloured either at height or intensity to facilitate interpretation.
Areas of Application and Limits of both Methods
Photogrammetry and LiDAR (3D laser scanning) both allow creation of accurate 3D maps – once using high-resolution photos and once using laser beams. Both technologies have their use cases and limitations.
Thanks to high-resolution visual data, photogrammetry scores with its photorealistic results in colour. Therefore, this method works well for applications in the field of mapping, building inspection and agriculture, for example for the 3D reconstruction and documentation of objects and systems. The surface finish can be documented particularly well with the photogrammetric method. However, the technology also has its limitations: Only points can be generated that can be detected by the camera sensor – and are therefore visible. Even with surfaces that are too featureless or in motion – such as the polished windows of a building or the sea – the process reaches its limits.
LiDAR enables a detailed representation of highly complex surfaces with the highest precision. This method is particularly suitable for complicated areas and objects such as listed buildings or factories. Narrow objects such as transmission lines, high-voltage lines, pipes, or sharp edges such as tear-off edges, roof edges or building corners can be detected with 3D laser scanning and later modelled. When it comes to topics such as factory planning, digital manufacturing, or VR (virtual reality), where the highest accuracy is required, LiDAR thus has a clear advantage. The technology is also used in the control and navigation of autonomous vehicles.
Another strength of 3D laser scanning by drone is its ability to penetrate vegetation with its laser beams to also measure and represent the terrain below. In forestry, LiDAR is therefore used for mapping, inventory, and biomass determination. And there is another advantage compared to photogrammetry: LiDAR is an active sensor that generates its own light. Weather conditions such as cloudiness and changing lighting conditions – such as absolute darkness at night – do not affect the measurement result in any way in contrast to the photogrammetric method with its passive camera sensor, which relies on a light source. However, the data set of 3D laser scanning in black and white can be difficult to evaluate in some cases. Additional photos can be helpful here.
In both photogrammetry and LiDAR, accuracy in the air is influenced not only by the sensor itself, but also by the movement of the drone. Regardless of which technology is used, a data correction is necessary.
Other points where the two technologies differ are the expenses in terms of costs and time. While photogrammetry requires an overlap of the images of 60 to 90 %, with LiDAR an overlap of the flight paths of 20 to 30 % meets requirements – the data collection with LiDAR is therefore much faster. Data processing with LiDAR is also very fast: The raw data only needs a few minutes of calibration to generate the final product – the point cloud. With the photogrammetric method, especially large areas cause an enormous amount of data. The processing of the images can take ten times longer than the survey. In comparison, scanner data can be processed much faster and better.
A LiDAR sensor set consists of scanner, IMU and GNSS that provide information on position, rotation and movement and are combined in real time to ensure high accuracy of results. Such a system costs between $US 50,000 to $US 300,000. In addition, there are costs between $US 25,000 and $US 50,000 for a suitable drone. For photogrammetry, only a professional drone with a camera is needed, which can already be purchased for $US 2,000 to $US 5,000. The cost factor should therefore not be underestimated when choosing the method.
Combination of Image and Scanner Data
The advantages of the two sensors can also be combined: The high resolution of the digital images and the fast and precise distance measurement of the laser scanner complement each other to precisely document complex objects in 3D, for example if the LiDAR point clouds have blurred structures. On the other hand, LiDAR can add (filigree) details to photogrammetry aerial data that have been "overlooked" by photogrammetry. As a result, you get a more detailed model of the respective scene.
Here are two examples: To create topographic maps of a territory with medium vegetation, a combination of photogrammetry and LiDAR is suitable for representing areas that are covered by trees or bushes. To capture the details of a railway line – such as the cables on the track or the metal posts that support it – 3D laser scanning is particularly suitable. With the aerial images of a drone, the point clouds can be provided with photorealistic textures.
What happens next with the collected Data?
Photogrammetry or LiDAR? In the end, I have to decide which technology is best suited for the respective application. How many details are needed? Do I want to be independent of the light situation? What budget do I have? How flexible do I want to be?
When choosing the measurement method, it should also be considered how the raw data – collected by the drone – should be further processed. Do I need photorealistic 3D models for visualization? The data alone is useless for companies. Scanning is the first step – the much more important question is: How do I deal with the detected data of the drone? How can I analyse and prepare the data to achieve the maximum benefit for my company?
Find out the answer to this question in our next drone blog post.