Early Detection of Slope Failure Risks: 8 Comparison Points for Drone Point Cloud Analysis

Table of Contents

• Wide area coverage and rapid data acquisition

• Improved work efficiency and safety

• Comparison of photogrammetry and laser scanning (LiDAR)

• Detection of minute signs with high-resolution data

• Visualizing shape changes with 3D models

• Quantitative evaluation of deterioration by time-series comparison

• Challenges and precautions at introduction

• Integration of data utilization into maintenance management

Summary

Slope faces (nori-men) are man-made slopes constructed along roads, on residential development sites, and as cut-and-fill on hillsides, and ensuring their stability is directly linked to the safety of social infrastructure. Recently, slope failure incidents caused by heavy rain and earthquakes have occurred across the country, greatly affecting human lives and economic activity. To prevent such accidents, it is important to detect anomalies early—such as cracks, surface collapse, subsidence, bulging, erosion, rockfall signs, poor drainage, and aging changes—and take appropriate countermeasures. However, conventional slope inspections relied on workers visually inspecting wide slopes, which carried risks of oversight and judgment errors, and working on dangerous steep slopes imposed a heavy burden on personnel.

Attention has therefore focused on slope risk monitoring using three-dimensional point cloud data acquired by drones. By measuring the entire slope with cameras or laser scanners mounted on drones and obtaining high-density point cloud data, it becomes possible to capture slight changes in slope shape and abnormal locations in three dimensions. Utilizing digital 3D models makes it easier to detect subtle deteriorations that are difficult to see with the human eye alone, so it is expected that precursors to collapse can be detected without being missed. This approach also aligns with the DX (digital transformation) of infrastructure inspection promoted by the Ministry of Land, Infrastructure, Transport and Tourism, and adoption of such technologies is likely to accelerate.

This article explains eight important comparison points for early detection of slope failure risks through analysis of point cloud data obtained by drones. From the merits of wide-area coverage to data accuracy and operational precautions, we will go through the perspectives practitioners should keep in mind.

1. Wide area coverage and rapid data acquisition

One of the biggest strengths of drones in slope inspection is their overwhelming coverage range and speed. In the past, surveying and inspecting an entire slope by manpower could take more than a full day with several people on wide slopes. In contrast, drones can acquire data for the entire slope from the air in a short time. Even for long slopes in mountainous areas, aerial imaging by drone can capture the current situation in several tens of minutes to several hours. For small slopes near urban areas, drones can capture the whole picture from a distance, enabling checks without omission.

Being able to cover wide areas makes it possible to analyze all parts, including those that could previously only be spot-checked. For example, roadside slopes that are difficult for people to access—high places, steep sections, or the backs of dense vegetation—can be observed from the air by drone. This allows detection of abnormal signs hidden in places that were previously out of reach. Furthermore, by creating a full 3D model of the slope or orthophotos (mosaicked images viewed directly from above) from the acquired data, detailed slope conditions can be examined from the office.

Rapid data acquisition also makes it possible to increase inspection frequency. Whereas conventional slope inspections were conducted only once or twice a year, using drones and point cloud analysis allows inspections to be carried out as needed. It becomes realistic to fly immediately after heavy rain or an earthquake—times when collapse risk is elevated—to inspect and, if abnormalities are found, detect and respond early. Frequent monitoring allows repairs or countermeasures to be taken at the stage of small changes, reducing the risk before large-scale collapse occurs.

2. Improved work efficiency and safety

Introducing drone point cloud analysis greatly contributes to on-site work efficiency and worker safety. In terms of efficiency, the time and effort required for surveying and inspection are drastically reduced. Previously, experienced surveying staff had to climb slopes and use tapes and surveying instruments to measure slopes and displacements at key points, but with drones, an operator can complete measurements by piloting from the foot of the slope or a safe location. High-resolution photos and laser measurement results can be processed the same day, greatly advancing analysis and reporting that used to be done the next day or later. As a result, slope inspections can be carried out with fewer people and in less time, enabling efficient maintenance management even at sites with personnel shortages.

Improved safety is another key benefit. Work on steep slopes or locations at risk of collapse has always carried the risk of falls and secondary disasters. Using drones allows data acquisition without personnel entering dangerous slopes, ensuring worker safety. This is especially important for slopes with increased collapse risk or ground loosened after heavy rain; close-up ground inspection by personnel is extremely hazardous. In such situations, drones can remotely capture the condition of damaged areas and assist in early emergency measures. This enables rapid response while preventing secondary human casualties. Eliminating the need to carry heavy surveying equipment across slopes also reduces the physical burden on workers and lowers risks such as heatstroke.

3. Comparison of photogrammetry and laser scanning (LiDAR)

There are two major methods for acquiring drone point cloud data: photogrammetry and laser scanning (LiDAR). Each has different characteristics, so it is important to select or combine the appropriate method according to site conditions.

Photogrammetry generates 3D models from numerous aerial photographs taken by a camera mounted on a drone. By using high-resolution images, it is possible to obtain point cloud data of the slope surface with accuracy of a few centimeters or less (a few cm or less (about 1 in or less)). The advantage of photogrammetry is that it provides detailed 3D models with color texture and orthophotos. Visual information such as cracks, discoloration, and seepage marks on the slope surface is captured, making it useful for detecting subtle signs described later. However, photogrammetry struggles where the ground is obscured by trees or dense vegetation. Photographic methods cannot capture terrain under dense vegetation, and this will not be reflected in the point cloud. Photogrammetry is also affected by lighting and exposure conditions, making it difficult to obtain clear data on cloudy days or at dusk.

By contrast, laser scanning (LiDAR) mounts a laser scanner on a drone to directly measure distances to the ground. Laser beams can penetrate gaps in trees, and by capturing reflections that reach the ground, point clouds of the ground surface can be obtained even on slopes covered by forest. LiDAR’s characteristics include obtaining stable, high-precision position information over wide areas (on the order of a few centimeters (a few cm (about 1 in))) by using GNSS and IMU for positioning, and acquiring very dense point clouds because it can capture several hundred thousand points per second. As a result, topographic undulations and steps that are difficult to capture with photogrammetry can be represented in detail. However, LiDAR equipment is more expensive than cameras and heavier to mount on drones, presenting hurdles in operating cost and expertise. Also, LiDAR point clouds mainly contain shape information and do not include surface color or texture (although material differences can be inferred from laser intensity, intuitive visual information like photos is not available).

In practice, it is essential to use photogrammetry and LiDAR appropriately according to slope conditions. For example, high-resolution photographic point clouds are suitable when crack detection is prioritized on sparsely vegetated rock slopes. In forested mountainous slopes, combining LiDAR measurements helps grasp ground surface shape. Recently, hybrid approaches have become common—using smartphone-mounted LiDAR or ground-based laser scanners for detailed close-range measurement and drone photogrammetry for wide-area coverage. By planning the optimal measurement strategy according to site conditions, sufficient point cloud data for collapse risk assessment can be obtained in any environment.

4. Detection of minute signs with high-resolution data

Signs that precede collapse can appear as very small changes in the initial stages. Examples include cracks a few millimeters wide (a few mm (about 0.1 in)), slight peeling or uplift of a surface part, seepage from gaps in sandbags, or fine sediment flow patterns (erosion traces) on the slope—subtle signs that can be easily overlooked. High-resolution point cloud data and orthophotos obtained by drone are powerful tools for capturing such minute changes.

With orthophotos generated from photogrammetry, high and hard-to-see portions of slopes can be enlarged and examined in detail from the office. This allows careful checks of where small cracks occur and whether there are any discolorations on the surface (wet areas, rust stains, etc.). If the point cloud data itself is captured at very high density, it can indicate abnormal locations as slight surface irregularities. For example, parts that are beginning to delaminate and float, or voids beneath concrete-faced slopes, may appear in the point cloud model as slightly distorted surfaces or irregularities compared to the surroundings.

AI image analysis technologies have also advanced in recent years. Software that automatically detects deterioration signs such as cracks, spalling, and leakage marks from drone-captured images has emerged, helping prevent human oversight. However, complete automatic anomaly detection has limitations, and at present it is indispensable for human experts to check high-resolution data. Drone point cloud data serve as a highly accurate “on-site record,” allowing inspections equivalent to observing the actual slope from the office. Because dangerous locations that people cannot approach can be examined in detail, the risk of missing abnormalities can be greatly reduced.

5. Visualizing shape changes with 3D models

By creating a 3D model of the entire slope, slight distortions and deformations of the terrain become intuitively recognizable. Traditional visual inspections mainly involved planar observations from the front of the slope, making it difficult to grasp changes in the depth direction or overall undulations. However, 3D models generated from point cloud data can be viewed from any angle and arbitrary cross-sections can be created, enabling a tangible understanding of the three-dimensional nature of deterioration.

For example, consider a case where the mid-slope is slightly bulging outward. Such subtle bulging that is hard to notice when looking up from below becomes obvious when viewed from the side on a 3D model. While healthy slopes generally form smooth curved surfaces, localized depressions (subsidence) or bulges (heave) can be highlighted by coloring those areas. Overlaying the current point cloud with the design section or past healthy-state data and displaying the differences as a heat map quantitatively shows areas deviating from the standard. On a color-coded heat map, protruding or sunken parts stand out intuitively, enabling field personnel to instantly grasp the presence and extent of deterioration.

Additionally, 3D models are useful as tools for sharing information among stakeholders. In the past, pointing on-site and saying “this area is slightly bulging” could lead to inconsistent understanding among listeners. Indicating deformation points on a 3D model ensures everyone shares the same information. It is easy to scale up or down like a model or attach cross-section prints to documents as needed. This facilitates common understanding of slope conditions among site supervisors, engineers, and clients. Visualizing shape changes plays an important role in rapidly reaching consensus on early-detected anomalies and implementing countermeasures.

6. Quantitative evaluation of deterioration by time-series comparison

The power of drone point cloud analysis is most evident in tracking slope changes over time. By repeatedly measuring the same slope at regular intervals with drones, point cloud data from each time can be compared to quantitatively measure slight displacements. Even subsidence or inclination progression on the order of several centimeters (several cm (about 1-2 in)) that are not noticeable by human visual inspection can be detected with high precision by comparing point clouds.

For example, suppose a slope moves slightly each rainy season. A single inspection might conclude there is no anomaly, but comparing this year’s point cloud model with last year’s may reveal that the shoulder has subsided by a few centimeters and the toe has slightly bulged. Such minute displacements are difficult to detect with rulers or levels on site, but point clouds capture them over an area and do not miss them. Quantitative evaluation can calculate specific rates such as “centimeters per year,” providing material to predict future collapse risk.

Early detection of anomaly signs through time-series comparison allows repair schedules to be advanced to prevent damage. Sites that continually monitor with point cloud analysis can set rules such as “if the change since last year exceeds a threshold, issue an alert” and implement automatic monitoring. In civil engineering, systems combining IoT and sensors for real-time deformation detection have emerged, and combining these with drone point clouds can build a more reliable early-warning system. Comparing multiple time-point datasets reveals “trends of aging changes” that are invisible in one-off inspections and is highly useful for preventive maintenance planning.

7. Challenges and precautions at introduction

Although drone point cloud analysis is highly useful, there are challenges to be noted during introduction and operation. First are the costs and technical hurdles related to equipment and software preparation. Initial investment is required for drones, high-performance cameras, and, in some cases, laser scanners, and certifications and knowledge are needed to operate them. In addition, specialized software or cloud services are required to process and analyze acquired point cloud data. Operating these tools requires a certain level of expertise, and if such skills are not available in-house, outsourcing may be necessary. When introducing new technologies to the field, sufficient prior training is desirable, and it may be beneficial to enlist experienced surveying companies or technicians.

There are also cases where drone point clouds alone cannot provide sufficient evaluation depending on slope conditions. For slopes covered with thick vegetation, as previously mentioned, photogrammetry is ineffective and some vegetation clearing in advance may be necessary. In cases where rock is fragile and drone vibration or noise could trigger rockfall, cautious measures such as photographing from a distance or switching to ground-based laser scanning may be needed. Drone flight regulations must also be observed. Even in mountainous areas, aviation law restrictions, radio reach, and obtaining flight permission in populated areas are regulatory hurdles to clear. Thorough planning and communication are indispensable for safe flights.

On the data management side, point cloud data volumes are enormous, so systems for storage and sharing must be established. High-resolution point clouds and photos for a single project can reach tens of GB, so consider corporate networks, PC specifications, and cloud storage capacity. Moreover, no matter how good the data is, humans must interpret and make decisions. Locations flagged by drone inspections should, when necessary, be followed up by on-site detailed checks or additional nondestructive testing, so combining with conventional methods remains important. Do not over-rely on new technology; position it as a complementary tool to enhance risk management accuracy and use it alongside traditional inspection techniques and experience for success.

8. Integration of data utilization into maintenance management

Point cloud data acquired by drones becomes an information asset that helps long-term maintenance management, not just detecting anomalies on the spot. Accumulating 3D models and image data over time acts like a “medical record” for slopes. Keeping digital records of conditions for each year makes it easy to verify questions such as “around what year did deterioration begin and how rapidly is it progressing?” or “has a previously repaired area deteriorated again?” This provides quantitative evidence that was difficult with paper reports or photo ledgers, improving the accuracy of maintenance plans.

Point cloud data can also be integrated into GIS (Geographic Information Systems) and CIM (Construction Information Modeling). Incorporating slope 3D models into a system that centrally manages multiple infrastructure assets enables comprehensive maintenance management that considers positional relationships with other structures and terrain conditions. For example, if data are integrated, it becomes efficient to prioritize and intensify monitoring of particularly high-risk slopes within a road network.

For field use, coordination when repairing or inspecting anomalies found by drone point cloud analysis is important. Accurately locating anomalies identified in the data on site requires surveying-based positioning. In the past, field crews searched using drawings and landmarks, but now high-precision GNSS allows pinpointing anomaly locations. For example, by using LRTK GNSS positioning devices attachable to smartphones like iPhones, a smartphone receiving correction information in advance can measure on-site coordinates to centimeter-level accuracy. If drone point clouds are accurately georeferenced to control points and the on-site coordinates indicated by LRTK are compared, cracks or subsidence identified in the data can be found on site without difficulty. This can also be applied to control point surveying and as-built verification; precise on-site positioning can be performed with only a smartphone, greatly improving field response efficiency.

By integrating acquired data into comprehensive maintenance management as described above, the value of drone point cloud analysis is maximized. Utilizing digital data across the cycle from inspection and record-keeping to repair planning and implementation makes it possible to protect slope safety with unprecedented speed and certainty.

Summary

Drone-based slope inspection and point cloud data analysis enable wide-area, high-frequency monitoring, precise anomaly detection, and safe, efficient operations. This article reviewed eight comparison points. Subtle signs and gradual changes that were often overlooked by conventional manpower-centered inspections can be understood three-dimensionally and quantitatively with drones, greatly increasing the possibility of detecting collapse risks in advance.

On the other hand, introduction requires addressing equipment and skill challenges and building an operational framework to effectively use the technology. Nevertheless, the value of adopting drone point cloud analysis is significant. In the context of national resilience and countermeasures for aging infrastructure, the importance of preventive maintenance using digital technology is increasing. Adding drones as a new tool for slope management allows inspections to evolve more smartly, promising both risk reduction and cost optimization.

Finally, positioning technologies that bridge drone data and field action—such as high-precision GNSS units like LRTK—is indispensable. By leveraging such solutions, insights from data can be accurately and quickly reflected in on-site responses. Making full use of digital data from the air to the ground will contribute to ensuring slope health and reducing collapse risk.