Streamlining Dam Displacement Monitoring: Procedures for Using Drone Point Clouds and Typical Accuracy

Table of Contents

• The Importance and Challenges of Dam Displacement Monitoring

• What Is Point Cloud Surveying Using Drones?

• Procedures for Utilizing Drone Point Clouds in Dam Displacement Monitoring

• Typical Accuracy of Drone Point Cloud Surveying

• Efficiency Gains in Displacement Monitoring from Drone Use

• Conclusion

The Importance and Challenges of Dam Displacement Monitoring

Dams are critical infrastructure that hold back vast quantities of water, and regular displacement monitoring is essential to ensure their safety. Concrete and earth dams can experience very small deformations or settlements (displacements) over years of operation or due to seasonal water level fluctuations. If abnormal displacement occurs in the dam body, it can lead to leakage or structural damage, so early detection and response are important. Therefore, managers need to monitor the behavior of the embankment systematically and assess its integrity.

Traditionally, dam displacement monitoring has commonly involved on-site staff measuring changes in the height or position of control points using levels or optical surveying instruments, or using plumb line (spirit-line) tilt and displacement measurements at in-dam observation stations. However, these conventional methods are limited to measurements at several “points” or along vertical cross-sections as “lines,” making it difficult to capture changes occurring across the entire “surface” of the embankment. As a result, there is a risk of overlooking deformations progressing away from observation points, potentially allowing damage to expand unnoticed. Inspections that require work at height or travel into mountainous areas are time- and labor-intensive, and efficient maintenance is a challenge amid worsening personnel shortages. Additionally, Japan has many dams built during the high-growth period and later; dealing with aging structures several decades after construction is a major issue. Regular inspections and observations required by law compel managers to collect behavioral data on dams routinely to confirm safety. Early detection of abnormalities may allow for small-scale repairs, whereas delayed discovery and progressed damage may necessitate large-scale construction. Appropriate monitoring directly contributes to extending dam service life and preventing major accidents. Against this backdrop, recent interest has focused on using ICT technologies and unmanned aerial vehicles (drones) to monitor dam displacement more efficiently and comprehensively.

What Is Point Cloud Surveying Using Drones?

Point cloud surveying using drones is a measurement method that acquires three-dimensional data of a target using cameras or laser scanners mounted on an unmanned aerial vehicle. The obtained “point cloud data” is a collection of many measured points in space expressed as XYZ coordinates, which can precisely reproduce the dam’s surface geometry as an aggregate of countless points. Traditionally, terrestrial laser scanners or manually measured points were used to partially understand shape, but drone surveying can cover wide areas in a short time and generate detailed three-dimensional models of an entire dam.

There are two main approaches. One is to mount a high-resolution camera on a drone to take many aerial photographs and reconstruct a 3D point cloud through software processing (photogrammetry). The other is to equip the drone with a small laser scanner and obtain the point cloud by measuring distances directly with laser light while flying (UAV laser surveying). Photogrammetry is currently widespread because equipment costs are relatively low and it is easy to handle, while laser surveying is effective in situations where cameras are challenged, such as slopes covered with trees or nighttime measurement. In either case, acquired point clouds capture fine surface undulations and shapes of the dam, and by analyzing that data one can identify locations and sizes of cracks and deformations, or extract cross-sections at arbitrary locations to measure displacement amounts. The obtained 3D models can be inspected from arbitrary viewpoints on a computer, allowing intuitive comprehension of defects that were difficult to discern by visual inspection or 2D drawings. The Ministry of Land, Infrastructure, Transport and Tourism is also promoting ICT utilization in infrastructure maintenance through initiatives like i-Construction, and three-dimensional data such as drone point clouds are becoming indispensable for advancing dam management.

It is important to choose the point cloud measurement method according to site conditions. When concrete surfaces lack patterns or features, photogrammetry may struggle to detect feature points, so placing markers on the wall surface beforehand for imaging is a common workaround. Conversely, on vegetated earth-dam slopes where the ground surface is not easily visible, laser surveying can capture ground points through gaps in the trees. By appropriately using cameras and lasers depending on material and environment, efficient and high-quality three-dimensional data can be obtained.

Procedures for Utilizing Drone Point Clouds in Dam Displacement Monitoring

Next, let us review the general procedures for monitoring dam displacement using point cloud data obtained by drones. The first step is planning. Based on the dam parts and area to be monitored and the required accuracy, plan the drone flight. For example, if both the upstream and downstream faces of the dam need monitoring, set flight routes and altitudes that can capture each face with sufficient coverage. Because high-resolution data is needed to capture displacements precisely, increase imaging density—for example, ensure photo overlap of 80% or more by setting short shooting intervals, or increase scan density in laser surveying. Check local weather and wind conditions in advance and choose a time when stable flight is possible. Prior to flight, confirm and comply with required permissions, approvals, and safety measures under aviation law and related regulations. In addition, to ensure accuracy as described below, install multiple ground control points (targets) and measure their precise coordinates as needed. There are also drone models now equipped with high-precision positioning devices, and using such aircraft can reduce the number of ground control points required.

When preparations are complete, fly the drone to acquire data. Use an automatic flight program to follow the planned route while imaging or scanning the dam from above and oblique angles. Even large dams can be recorded in a short time, and data can be safely acquired for steep or high areas that are inaccessible on foot. After shooting, process the images or laser measurement data with dedicated software to generate three-dimensional point clouds. In photogrammetry, the software matches feature points across numerous photos to reconstruct spatial coordinates and outputs colorized point clouds or orthophotos (distortion-free nadir views). In laser surveying, combine the acquired laser points with the drone’s position and attitude data during flight to convert them into three-dimensional coordinates. In either method, use high-accuracy surveyed control points (known points) to correct coordinate offsets and align the entire point cloud to a real-world coordinate system. This allows accurate comparison between point clouds acquired at different times.

Save the resulting whole-dam point cloud model as baseline data, and thereafter periodically acquire additional point clouds following the same procedure. The monitoring frequency depends on dam type and importance: from annual measurements aligned with annual inspections, to multiple measurements per year for large dams to track seasonal variations, and even ad-hoc surveys immediately after earthquakes or heavy rainfall. When new point cloud data are obtained, analyze differences by overlaying them with previous data. Specifically, use point cloud processing software to compare corresponding positions of two 3D models and calculate how much the surface has moved. You can visualize displacement amounts across the dam surface as a color map, or plot changes in height or position along critical cross-sections (for example, the central vertical section of the embankment or the crest horizontal line). As a result, you may detect specific displacements such as “a bulge of +20 mm compared to the previous measurement at the central downstream side of the dam.” If detected displacements are within allowable ranges, continue observation; if abnormal variations exceeding expectations are seen, conduct detailed on-site investigations and use the results for appropriate maintenance decision-making.

Typical Accuracy of Drone Point Cloud Surveying

The accuracy attainable with drone-based point cloud surveying has reached practically useful levels due to recent technological improvements. Generally, when proper procedures are followed, positional accuracy of point clouds generated from aerial photographs can be on the order of several centimeters in both horizontal and vertical directions. For example, in one field demonstration, measurements of structure positions on a 3D model generated by drone photogrammetry showed errors of about 3 cm (1.2 in) compared to conventional ground surveying. Also, if the ground sampling distance (GSD) of the captured images is 1 cm (0.4 in)/pixel, point cloud errors are often about 1–3 cm (0.4–1.2 in) (within 1–3 times the image resolution). Drones equipped with RTK-GNSS high-precision positioning can tag photos during flight with positional errors of a few centimeters, enabling high-accuracy results even with significantly fewer ground control points. Conversely, without such measures, for example when modeling with a consumer GPS drone and no control points, errors on the order of meters can occur, making the data unsuitable for displacement monitoring. To reliably achieve centimeter-level accuracy, it is important to use RTK-capable drones or a sufficient number of known points and to perform accuracy verification during processing.

Point cloud accuracy has two aspects: absolute accuracy (correctness relative to a coordinate system) and relative accuracy (consistency within the point cloud). Dam displacement monitoring requires both consistency of absolute coordinates for temporal comparisons and relative accuracy to capture shapes within a single dataset. The former can be kept within a few centimeters with appropriate positioning, and the latter benefits from low internal noise so that very small changes can be read. In practice, even if individual points contain millimeter-level errors, fitting a surface from many surrounding points can average out noise and allow detection of surface tilt or movement on the millimeter order in some cases. In other words, while point-by-point precision of point clouds may be inferior to that of conventional precision surveying, the large number of points can statistically improve effective accuracy. However, when discussing minute displacements, removing measurement noise and performing strict coordinate alignment are indispensable. Field conditions such as wind and lighting during imaging or reflectivity effects in laser measurement greatly affect accuracy, so acquire data under stable conditions where possible and apply noise filtering and careful registration between paired point clouds (e.g., fine adjustment using ICP algorithms) during post-processing.

For example, analytical methods are being studied that fit a plane to the embankment surface from the acquired point cloud and track changes in its normal-direction inclination angle or position to detect slight tilts or bulges of the entire embankment with high accuracy. Considering the above, data obtained from drone point cloud surveying, when processed appropriately, can be expected to have sufficient accuracy for dam displacement monitoring. In practical terms, it is possible to reliably detect displacements of several centimeters, and under favorable conditions differences on the order of several millimeters may be distinguishable. On the other hand, directly detecting extremely small changes such as a crack width of 1 mm (0.04 in) is difficult due to the resolution limits of photography or laser, and in such cases it is necessary to combine high-magnification camera inspections or sensor measurements. Nevertheless, for the typical aims of dam maintenance—to avoid overlooking signs of deterioration—the accuracy of drone point clouds is generally sufficient.

Efficiency Gains in Displacement Monitoring from Drone Use

Introducing drone point cloud measurement dramatically improves the efficiency of dam displacement monitoring work. First, it greatly reduces on-site time and manpower. Conventionally, survey teams had to set up instruments at each designated observation point and measure sequentially, and for large dams with many measurement locations this could take more than a full day. With drone surveying, the entire dam can be captured by simply flying the aircraft, and imaging itself can be completed in tens of minutes to about an hour (including setup and teardown, typically less than a half-day). This shortens on-site work time and reduces the number of personnel required, lowering the burden and cost of monitoring operations. Eliminating the need for work at height or access to dangerous steep slopes also improves worker safety. In the event of an emergency inspection immediately after a major earthquake or heavy rain, drones can quickly capture dam conditions even in areas where people cannot approach, speeding initial response.

In practice, sites that adopted drone inspections have reported remarkable efficiency improvements. There are cases where dam displacement observations that previously required a five-person survey team working a full day were completed after drone introduction by just two staff conducting a few hours of flight and imaging, with acquired data accuracy comparable to conventional methods. Moreover, from the point cloud model obtained, inspectors could identify not only displacement amounts but also surface damage locations, yielding multiple outputs from a single field operation.

Furthermore, using drone point cloud data improves not only speed but also the quality of monitoring. Abnormalities missed by limited point measurements are easier to discover with comprehensive 3D data. Point clouds and orthophotos are stored as digital data, enabling repeated comparisons and detailed analyses over time. For example, long-term monitoring—quantitatively tracking displacement progression by comparing data from several years ago with the latest dataset—is straightforward. Point cloud data are also easy to share among stakeholders, allowing remote consultations where specialists unable to visit the site can review models in the office and provide advice. However, point cloud files are large and handling them can be challenging, so using cloud services and setting up data management systems are important. Recently, attempts to automatically detect concrete surface cracks and deterioration by applying AI to high-resolution photos taken by drones are progressing, and combining such technologies can further streamline inspection workflows. Overall, drone-based displacement monitoring offers significant advantages over traditional methods: rapid, safe data collection and comprehensive information that supports accurate decision-making. Drone point cloud measurement can also work in tandem with in-dam continuous monitoring instruments such as inclinometers and extensometers. By cross-referencing numerical data from fixed sensors with the drone-acquired full-coverage visualization data, a more reliable integrity assessment is possible.

Moreover, combined with advances in communication network technology, an era in which dam conditions can be monitored remotely is approaching. Drone-captured data can be transmitted to the cloud via wireless communication in real time, allowing remote specialists to review them from distant offices. Cutting-edge demonstrations have even integrated high-precision sensors and satellite communications to remotely monitor fine displacements of 1–2 mm (0.04–0.08 in) continuously for 24 hours. These are still experimental, but they have the potential to evolve into systems that detect abnormalities without requiring on-site presence.

Conclusion

Surface-wide displacement monitoring of dams, which was difficult with conventional methods, can now be realized efficiently and with high precision through drone-based point cloud surveying. For dam managers, drone point clouds offer a powerful tool to quickly and safely understand current conditions. Although initial investment and training are required, once introduced the workflow from data acquisition to analysis can be completed in-house, making continuous monitoring easier. Being able to objectively determine the presence or absence of displacement from comprehensive data improves decision-making on repair timing and the need for further detailed investigation. Drone point cloud technology is being applied not only to dams but to various infrastructure inspections and is expected to become increasingly widespread. This technological innovation aligns with DX (digital transformation) in infrastructure management and greatly contributes to both labor reduction and sophistication of field work.

Looking ahead, combining drones with new measurement tools that support on-site accuracy control will also be effective. For example, using an iPhone-mountable high-precision GNSS positioning device called LRTK allows on-site measurement of control point coordinates with a few centimeters of accuracy (cm level accuracy, half-inch accuracy) without specialized surveying equipment. Because a palm-sized receiver and a smartphone alone can achieve positioning with a few centimeters of accuracy (cm level accuracy, half-inch accuracy), the mobility of field work is dramatically improved. By combining such easy high-precision positioning technologies, it becomes possible to quickly verify and correct point cloud data acquired by drones on site or measure additional checkpoints, thereby improving on-site responsiveness. LRTK is one example, and in the future, tools that support maintenance—such as smartphone-based high-precision positioning technologies and autonomous-flight drones—will continue to evolve. By effectively leveraging drone point cloud surveying together with smartphone-linked high-precision positioning devices, dam displacement monitoring operations can be made more efficient and reliable.