In drone survey missions, the choice between photogrammetry and LIDAR depends heavily on the exact application. You also need to consider operational factors, such as cost and complexity. Knowing what outputs you really need will help you make the right decision.
We get a lot of questions about LIDAR sensors and their application to 3D surveys using drones. What is LIDAR and how does its output compare with results obtained with high-resolution RGB cameras and photogrammetry? In this article, we’ll explore the ways photogrammetry and LIDAR are actually quite different from each other, even if their three-dimensional (3D) outputs look similar. We’ll then dig deeper into specific applications and how photogrammetry can provide exceptional results for most missions at a fraction of the cost and complexity of LIDAR.
Photogrammetry and professional, high-resolution cameras can cost-effectively generate 2D and 3D surveys like this one, with absolute accuracies down to 1 cm (0.4 in) RMS horizontal and 3 cm (1.6 in) RMS vertical.
How does photogrammetry work?
In photogrammetry, a drone captures a large number of high-resolution photos over an area. These images overlap such that the same point on the ground is visible in multiple photos and from different vantage points. In a similar way that the human brain uses information from both eyes to provide depth perception, photogrammetry uses these multiple vantage points in images to generate a 3D map.
The result: a high-resolution 3D reconstruction that contains not only elevation/height information, but also texture, shape, and color for every point on the map, enabling easier interpretation of the resulting 3D point cloud.
Drone systems that use photogrammetry are cost effective and provide outstanding flexibility in terms of where, when, and how you capture 2D and 3D data.
How does LIDAR work?
The sensor itself is only one part of a LIDAR system. Critically important for capturing usable data, you’ll also need a high-precision satellite positioning system (GNSS) as well as high-accuracy sensors to determine the orientation of the LIDAR sensor in space—an inertial measurement unit (IMU). All of these high-end subsystems must work in perfect orchestration to enable processing of the raw data into usable information, a process called direct geo-referencing.
Classical airborne LIDAR surveys are conducted from a manned airplane and are less accurate but capable of covering more ground than lightweight UAV LIDAR operations. Specifically, you can cover between 10 and 1000 km2 (4 and 400 mi2) in one flight. The absolute accuracy depends on the flight height and sensor choice. At a typical flight height of 2000 m (6600 ft) above ground level (AGL), you can expect an absolute accuracy limit of about 20 cm (8 in) horizontal and 10 cm (4 in) vertical.
Lightweight drone LIDAR systems cover as much as the drone allows per flight. As we will discuss in detail in below sections, these systems can be more accurate than those carried by manned aircraft. Specifically, fixed-wing drones carrying a LIDAR payload can cover up to 10 km2 (4 mi2) in a flight, with absolute accuracy limits right around 10 cm (4 in) horizontal and 5 cm (2 in) vertical.
In both cases of manned aircraft and lightweight drone LIDAR, the accuracy is significantly less than photogrammetry avails. Plus the post-processing for LIDAR absolutely requires expertise beyond a quick training or reading of a manual, as we’ll discuss below.
What about accuracy?
As we have seen, photogrammetry and aerial LIDAR differ in the way points on the ground are registered. This directly affects the final point cloud accuracy and we will see that, especially for horizontal accuracy of areas free from dense forest canopy, photogrammetry clearly outperforms aerial LIDAR.
In the case of photogrammetry, a quality, high-resolution, full-frame sensor camera like WingtraOne’s Sony RX1R II can yield outputs with horizontal (x-y) accuracies in the range of 1 cm (0.4 in) and elevation (z) accuracies in the range of 2 to 3 cm (0.8 to 1.2 in) over hard surfaces, enabling precise volumetric analysis.
Note, however, that in order to achieve such performance the payload used for photogrammetry must be a professional one, with the right image sensor and lens to capture more detail. It’s not just about the number of pixels. In fact, two cameras with the same number of megapixels and different size sensors provide different image quality and accuracy.
Proper mission planning and post-processing are also important for achieving optimal accuracy: good overlap among images increases accuracy and provides better error correction compared to complete reliance on the direct geo-referencing method used in LIDAR. A high-end drone system with professional mission planning and post-processing workflow helps ensure that you capture quality data that generates accurate results.
As for aerial LIDAR methods, the sensor does not target specific features on ground but instead shoots the beams at a set frequency in a defined pattern. Even if the horizontal accuracy of the single point might be higher, the best horizontal accuracy of a point of interest on the ground is limited by the point density.
Manned aerial LIDAR can provide a point density of up to 50 pts/m2 and offers a typical absolute accuracy of 20 cm horizontal and 10 cm vertical if flown at a standard height of 2000 m (6600 ft) AGL.
By flying lower, lightweight UAV LIDAR provides a higher point density than manned aerial LIDAR and can achieve better accuracy even though the laser is less powerful. Mounted on a multi-copter, point density and the resulting point cloud accuracy can be improved by flying low and slow at the expense of reduced efficiency.
In the case of LIDAR on fixed-wing drones, a point density between 50 and 200 pts/m2 is possible. This means a measurement every ~ 10 cm, so an absolute horizontal accuracy of about 10 cm can be achieved.
On top of limited horizontal accuracy, LIDAR-derived point cloud accuracy depends on the precision of the LIDAR itself and the quality of the INS (IMU and GNSS) system. Considering all technological advancements and system variables at this time, the typical absolute accuracy that you can expect from a lightweight LIDAR system on a fixed-wing drone is approximately 10 cm (4 in) horizontal and 5 cm (2 in) vertical.
Bottom line: if your applications depend on high absolute accuracy, you will want to go with photogrammetry.
Other key specs
Photogrammetry offers photorealistic mapping results in the form of orthomosaics, point clouds and textured mesh. A true, life-like digital twin. This can be extremely helpful for identifying and measuring features, especially when the accuracy is so good. LIDAR offers a sparse laser point cloud that gives a general sense of shapes and contours but does not offer contextual detail. You can also colorize LIDAR data with RGB data, but this is a more complicated process, and you will still be lacking the details you get with photogrammetry.
Photogrammetry (top) and LIDAR (bottom) both allow creation of accurate 3D maps. Photogrammetry outputs also include high-resolution visual data in full color for every point on the map to aid in the interpretation. LIDAR outputs are often colorized based on intensity of reflection, such that soft objects (leaves of trees, grass) can be more easily differentiated from hard objects (walls, roofs, cars, etc.). LIDAR image credit: USGS (2018a): Kilauea Volcano, HI June 2018
In the cases of both LIDAR and photogrammetry, the sensors do not penetrate through the leaves. However the nature of the two methods allows for LIDAR to offer information on the ground in densely wooded areas that feature a tight canopy. Specifically, the laser beams shower down over a forested area, and some can reach the floor and bounce back up. Photogrammetry, which relies on photographs, will not give as much information on canopy-covered grounds due to the darkness and shadows that dense tree stands create. Note that photogrammetry will offer topographical information in lightly wooded areas or forested areas that do not feature heavy canopy.
Workflow and support
In the case of photogrammetry, companies like Pix4D, Agisoft, Bentley CC, Propeller and Dronedeploy have optimized workflows over years of experience with much data. In some cases, the workflow is turnkey—one software suite offers clients processing and accuracy verification within 24 hours when they simply upload the data. Cloud/server solutions allow for large-scale data processing with minimal hardware investment. And dedicated support teams are available to help.
Point clouds from manned aerial LIDAR require trained professionals to oversee post processing. While the processing is well-established, it requires expertise that cannot be picked up along the way and is not an inherent part of the software available.
In the case of lightweight drone LIDAR, Companies like Phoenix LIDAR Systems and Yellowscan are pioneering a streamline data processing workflow. Yet, still, today, the process typically involves multiple manual steps and different tools and expert knowhow to develop an accurate point cloud.
Photogrammetry processing for full resolution takes several hours (or days) depending on the project size. If you only need a sparse set of accurate tie points (like from a LIDAR source), photogrammetry tools offer downsampled processing options.
LIDAR point clouds are directly geo-referenced with real-time kinematic (RTK) positioning during flight, or via post-processed kinematic (PPK) positioning after the flight. Typical post-processing steps include trajectory adjustments, strip adjustments and noise filtering, which are computationally less expensive than the photo adjustments of photogrammetry. Depending on the project size, the process takes from less than an hour (smaller projects with lightweight drone LIDAR) to several days (large projects with manned aerial LIDAR).
In most cases, a LIDAR sensor alone will rival the cost of an entire photogrammetry data gathering system. Precise geo-located lasers are more expensive than cameras. For this reason, it’s critical to assess your current and future applications (next section) to make sure the investment in LIDAR is the best decision. In today’s market, an aerial data gathering system can set you up to take on a lot of new projects, so the last thing you want is to invest in something that limits your opportunities.
Photogrammetry vs. LIDAR cost breakdown
All above prices in US dollars. To note: an entire high-end photogrammetry system costs between $US 20,000 – 30,000 whereas just the sensor for manned LIDAR typically costs $US 100,000. Lightweight drone LIDAR payloads by themselves run between $US 65,000 and $US 100,000.
Photogrammetry and LIDAR in specific applications
For most missions, 3D results achievable with photogrammetry are similar to those obtained with LIDAR, but with better accuracy and greater versatility, e.g., photorealistic outputs, thanks to the high-resolution visual data. There are some applications—specifically featuring power lines or large areas of dense forest canopy—where the higher expense of LIDAR for airborne missions is justified, however. Let’s look at the evidence for this across a range of actual applications as follows.
Topographical maps featuring light vegetation (sparse tree stands or open canopy) are best captured with high-resolution RGB data capture available through payloads like the RX1R II with PPK. The resolution and photorealistic results are useful in cases like wildfire management in residential areas, and have been used by some of the world’s largest urban fire and rescue services, since the information serves many stakeholders who need a real view of what’s happened. This payload is also highly accurate, offering dependable, survey-grade results to government agencies, as in this Indiana Port Authority survey case. Finally, and not the least important is price and ease of workflow. For businesses like this vineyard, which would benefit greatly from detailed and accurate information without extensive training and overhead.
Topographical maps with medium vegetation can be obtained via a combination of photogrammetry and a method to capture the ground below the vegetation. To capture the additional information below the vegetation, ground survey methods or aerial LIDAR can be used. The combination with ground survey methods keeps the price down while guaranteeing high accuracy plus the resolution and photorealistic results available through photogrammetry. Detailed tutorials offering a reasonable learning curve on this approach are available.
Large-scale topographical maps featuring heavy vegetation are best acquired via manned airborne LIDAR. A digital terrain model (DTM) of the forest ground provides useful information for project planning in construction (e.g., the planning of new roads), forest biomass or detailed information on vegetation and habitats via topography and underlying terrain, Applications falling under these circumstances will always require LIDAR at least in part to normalize topographical data, as is shown in research that examines the strengths and limits of photogrammetry in such cases.
Typically state agencies try to maintain reasonably accurate digital terrain models (DTMs) of the forest grounds. For these kinds of large-scale projects with low resolution requirements, manned airborne LIDAR is the most cost-effective option available. If a more accurate or up-to-date DTM of a small forest is needed, a traditional ground survey will be the cheapest option available, yet lightweight drone LIDAR might fill a niche in-between.
Bare-Earth mining, volumetric and natural resource surveys are best handled by high-end RGB payloads like the RX1R II. Even massive surveys, like those performed by an energy firm in Finland and the US, are ideal with the right drone and RGB camera. Established mining firms like Jellinbah and Westmoreland have offered examples of how they’ve incorporated photogrammetry into their workflows because of the accuracy, resolution and photorealistic results that enable efficient mine management. On top of this, photogrammetry is cost effective and saves time not only to capture and process data related to cut and fill volumes, stockpile assessments and status reports, but also to share this information and reconcile with contractors and stakeholders.
Professional drones like the WingtraOne can capture up to 400 hectares (988 acres) in a single flight at resolutions of 2.5 cm/pixel (1 in/pixel). In this example above, WingtraOne was used to efficiently create surveys over tens of thousands of hectares (2D RGB map on top, 3D digital surface model at the bottom).
Power line surveys for vegetation control can be done with LIDAR or high-resolution photogrammetry and powerline extraction features on software like Pix4Dsurvey. For the sake of photorealism, price and workflow, we recommend the later option. A good example is this one from Poland, where FlyTech UAV used photogrammetry to revolutionize its powerline vegetation management. Research is ongoing around photogrammetry as a go-to, cost-effective solution that is even incorporated into a management update to the largest European power grid operator.
Power line pole tower inspection benefits from live video inspection with a multicopter carrying an RGB or thermal payload. These are usually relatively small areas that multicopters can maneuver around and take oblique shots of easily and safely, as this overview demonstrates. With this method, you get all information within a very short amount of time. Zoom cameras allow detailed inspection that can not be offered by photogrammetry or LIDAR.
Rail track inspection is still most often carried out from the ground—by a train equipped with ultrasonic, LIDAR, and visual sensors. Inspection from the air with either photogrammetry or aerial LIDAR is gaining more and more interest but both methods are in early stages. High-resolution photogrammetry offers data that avails outputs with all of the essential details accurately and autonomously while saving time. Plus the photorealism adds an element of easy identification and versatility that can answer to a range of questions. In the end, more and more firms are making the case for this methodology.
City mapping with vertical structures requiring 3D vantage points has been widely demonstrated with photogrammetry based on imagery captured with a payload featuring oblique capabilities. For cityscapes with many high-rises and intense levels of vertical detail, multicopters work well, although their ability to cover wide-spread areas per flight is compromised. VTOL drones carrying oblique payloads can still capture wide areas and achieve impressive vertical accuracy. In fact, more and more cases of city mapping are being reported with VTOL drones.
The difference between photogrammetry and LIDAR grows when considering operational and logistical factors. In order to generate quality results, a LIDAR system requires all of its components to work perfectly in sync. Small gaps or errors in sensor measurements can lead to significant errors in outputs. Or worse, outputs that “look” right but are not. Techniques like ground control points (GCPs), which are useful in photogrammetry to correct issues, are harder to implement with LIDAR. Most of the time, the only solution for erroneous LIDAR data is to repeat flights.
We have explored the differences between how photogrammetry and LIDAR work and the similarities in their outputs and learned about situations where each technology can be best applied. And while some specific applications might justify the cost and complexity of LIDAR, photogrammetry can meet most of the everyday challenges presented across a range of projects and industries, providing exceptional accuracy and stunningly detailed maps, available on demand and with minimal expertise overhead.
So if you don’t need what LIDAR uniquely provides—specifically to mid- or large-scale forests with heavy but penetrable canopy—you can do more using photogrammetry coupled with a professional drone for significantly less money and complexity. Learn more about the use of 3D models and photogrammetry in surveying and mining and how the WingtraOne system enables efficient, accurate, and cost-effective mapping missions.