Verification of Slope 3D Measurement Accuracy: Achieving Centimeter-Level Precision with GNSS × Point Clouds

Introduction: Advanced Slope Management and Growing Demand for 3D Surveying

In civil engineering sites, slope management has increasingly demanded higher levels of precision in recent years from the perspectives of safety and quality assurance. Traditionally, slope shapes were checked using total stations (TS) or levels, but these methods have limits in measurement coverage and point density, and they impose heavy workloads. Meanwhile, advances in 3D technologies such as laser scanners and drone photogrammetry have increased demand for three-dimensional surveying that captures terrain, including slopes, as surfaces. With the i-Construction initiative promoted by the Ministry of Land, Infrastructure, Transport and Tourism, efforts to use point cloud data for as-built management are expanding. However, when transitioning to new technologies, concerns often arise: "Can accuracy be maintained as before?" This article examines the accuracy of the latest method combining real-time kinematic GNSS positioning (RTK) and smartphone 3D point cloud measurement. From a technical perspective, we outline the surveying requirements for slopes, present demonstration data and comparisons with conventional methods, and explain the basis for achieving centimeter-level accuracy.

Technical Perspective: Slope Survey Items and Accuracy Requirements

When surveying slopes, several important items must be measured. First, the coordinates of the upper edge (slope shoulder) and lower edge (slope toe) need to be known accurately to determine boundaries with land parcels and structures. It is also crucial to check whether the slope gradient (inclination angle) matches the design. For example, if a 1:1.5 slope is specified in the design drawings, you must confirm that this gradient is maintained uniformly along the entire length and that it does not suddenly flatten or steepen at any point. In addition, surface irregularities and finishing thickness of the slope are inspection targets. Especially when slopes are protected with sprayed concrete or vegetation mats, it is necessary to confirm that the surface is smooth and the specified thickness is achieved.

It is also important to understand the accuracy levels required for such slope shape measurements. In general, tolerance on as-built dimensions for civil works is set at the order of several centimeters. For example, finishing positions of slopes are often required to be within approximately ±5 cm of design, and inspections check whether this requirement is met. In other words, surveying must be capable of reliably detecting offsets of several centimeters. Single-point observations with a total station can deliver millimeter-level accuracy, but the number of points that can be collected at once is limited. Therefore, point cloud surveying, which can capture dense surface data, becomes effective. The Ministry of Land, Infrastructure, Transport and Tourism’s "3D As-Built Management Guidelines (draft)" indicates a measurement density standard of at least 100 points per 1 m² (1 point per 0.01 m²) when using TLS (terrestrial laser scanner) for earthworks. The intent is to cover the entire slope surface with points so that even small bumps and hollows are not missed and the shape can be accurately captured.

In short, both "point accuracy" and "point cloud density" are important in slope surveying. While point clouds can capture countless points at once, the positional error of each individual point can be larger than that of traditional TS surveying. However, the large number of points provides a statistical advantage: errors can cancel out and the surface shape can be reproduced with high accuracy. For example, fitting a cross-section to numerous point cloud points will average out local scatter and yield a smooth surface shape. Taking these characteristics into account, using 3D point cloud technology can be expected to reconcile the need for cm-level accuracy and comprehensive shape capture in slope management.

Performance and Comparison: Accuracy Differences between GNSS × Smartphone Point Clouds and Conventional Methods (TS, Drone)

Here we present a case study that actually verified the accuracy of GNSS × smartphone point cloud measurement in the field. On a road slope about 10 m high and roughly 50 m long, we scanned the entire slope using a smartphone equipped with LRTK and compared the resulting point cloud data with measurements from a total station survey and drone photogrammetry.

First, we selected several known control points and distinctive features on the slope and compared coordinates read from the smartphone point cloud with values measured by the TS. The horizontal and vertical errors were generally within 2–3 cm. For example, at a checkpoint placed on the slope shoulder, the planar position difference was 2.1 cm and the height difference was 2.8 cm, showing excellent agreement with TS measurements. Regarding the overall slope shape, when longitudinal and cross-sectional profiles derived from the smartphone point cloud were compared with profiles formed by connecting TS survey points, the overall shapes largely overlapped and even local bumps and hollows matched.

Next, the same slope was measured by drone photogrammetry (point clouds generated from aerial photos) and compared with the smartphone point cloud. After aligning both point clouds to control points and analyzing elevation differences on the slope surface, the average error was below a few centimeters, with maximum differences around 5 cm occurring only in areas with dense vegetation. This indicates that smartphone point cloud surveying captures slope shapes at an accuracy level comparable to drone photogrammetry. Visual overlays of the point clouds also showed that the slope surfaces do not appear doubled but almost coincide as a single surface, indicating high consistency between the two measurement results.

From these verifications, GNSS × smartphone point cloud slope measurement can be said to achieve accuracy comparable to conventional methods (TS and drone). The results agree with TS reference values within a few centimeters and match well with drone point clouds, demonstrating that the method provides sufficient accuracy for as-built management purposes. Furthermore, smartphone measurement can be completed by a single operator in a short time, offering significant efficiency benefits. The combination of such accuracy and operational convenience is highly useful on-site.

Reliability of Procedure: Field Conditions and Measurement Steps to Ensure Accuracy

To achieve high and stable accuracy, it is important to prepare measurement conditions on site and follow appropriate procedures. First, for GNSS positioning, securing an RTK fixed solution (Fix) is a prerequisite. With a three-frequency RTK-GNSS receiver such as LRTK, you can capture many satellites in open sky and obtain centimeter-level positioning as a rover. Before starting measurements, power on the equipment and confirm that correction information from a networked RTK service (Ntrip) or the QZSS "Michibiki" CLAS signal is being received correctly. The device will indicate when a fixed solution is obtained, so always check that status. If no Fix is achieved, check for tall structures or trees that may block signals, change your position, or wait a bit before retrying. Even in out-of-coverage areas, CLAS-capable devices can receive augmentation signals directly from satellites, but initial convergence may take longer. By remaining stationary and averaging for a few minutes until positioning stabilizes, you can obtain reference point coordinates with sub-centimeter horizontal accuracy and vertical accuracy on the order of a few centimeters.

There are also points to note in the smartphone point cloud measurement procedure. Basically, you hold a smartphone equipped with camera and LiDAR and walk while pointing it at the slope, but certain practices are needed to ensure accuracy. First, scan evenly from bottom to top so the entire slope is covered. For long slopes, split the scan into sections and include overlap zones to merge datasets. Move the smartphone slowly and capture each area from various angles to reduce shadows and uneven density in point cloud generation. The upper slope near the slope shoulder may be hard to see from below, so, if possible, move around above and scan from the shoulder side as well. Conversely, the lower slope toe can be missed if too far away, so approach and shoot as needed. Smartphone LiDAR can miss points outdoors due to sunlight, so rely primarily on photo-based point cloud reconstruction and use LiDAR to supplement fine details at close range.

Verification and adjustment of the acquired point cloud data are also essential. After smartphone scanning, check for offsets against known points in software. Many apps and analysis tools allow you to georeference point clouds using several reference points. For example, identify existing survey stakes or structural datum points with known coordinates around the slope, and translate/rotate the entire point cloud so those points align—this corrects small positional shifts or tilt errors. It is also effective to perform quick on-site checks by measuring dimensions directly in the point cloud. Comparing measured values in the point cloud to known distances (e.g., distance between toes or slope length) verifies that scale is represented correctly. By providing rapid feedback during measurement, you can maintain high reliability of the point cloud data.

Implementation and Reproducibility: Repeatability and Quality Assessment in Field Cases

When introducing a new surveying method on site, it is essential to confirm reproducibility and whether data quality varies. Examples show that smartphone point clouds combined with GNSS can produce very stable results under appropriate conditions.

One example involved scanning the same slope twice on different days and comparing the resulting point cloud models. Despite different operators on each day, the generated slope point clouds showed almost no difference. Distance analysis between point clouds indicated average differences of 1–2 cm or less, with maximum differences around 3 cm. This demonstrates that results from this method converge to a consistent accuracy level regardless of operator or timing. Provided field conditions do not change significantly, the method can reproducibly deliver centimeter-level accuracy.

For quality assessment, using objective metrics to confirm accuracy is important. For instance, obtain residuals (differences) in height and position between point cloud data and TS-checked verification points and compute the RMSE (root mean square error) to grasp overall accuracy tendencies. In measured cases, RMSE for smartphone point cloud verification points was about 2 cm in planar position and about 3 cm in elevation, confirming an accuracy level comparable to conventional photogrammetry and mobile mapping systems. Also, as-built inspection using TS measurements corroborated that cross-sectional dimensions calculated from the smartphone point cloud fell within specification limits. In other words, data produced by this method are reliable enough to withstand third-party verification.

Currently, pilot ICT construction projects by the national and local governments are also adopting simple 3D measurement using smartphones × GNSS. In practical projects where the method was applied to as-built measurement, results met inspection quality standards while significantly reducing work time compared with conventional methods. As such achievements accumulate, on-site technicians are increasingly gaining confidence in this approach.

Analysis of Error Sources in Measurement Results (Satellite Corrections, Vegetation, Sunlight, LiDAR Resolution, etc.)

Even with new technology, it is necessary to understand various error sources that affect measurement results. Below is a concise summary of main factors, their impacts, and countermeasures.

Satellite correction data and GNSS positioning: GNSS accuracy strongly depends on the presence and quality of correction data. If corrections from an RTK reference station cannot be received or a fixed solution is not achieved and only a float solution is available, errors of tens of centimeters can occur. Fortunately, devices like LRTK can generally maintain corrected positioning via network RTK or satellite-delivered CLAS augmentation, but unfavorable satellite geometry or ionospheric disturbances at certain times can still cause small positioning errors. Near slopes, one side may be blocked by mountains or structures so that only half the sky is visible. In such cases the number of visible satellites decreases and multipath (signal reflection) is more likely, degrading accuracy. Countermeasures include taking measurements from as open a location as possible, mounting the GNSS receiver on a pole to raise it toward the sky, and averaging positioning over a longer period to reduce transient fluctuations. Always monitor satellite corrections and positioning status, and be cautious to exclude or remeasure data segments showing anomalies.

Vegetation and ground conditions: Vegetation or dense grass on a slope can introduce measurement errors. Point clouds capture the surfaces seen by laser or photos, and on grassy slopes the sensor may capture leaf tips instead of the true ground, making the slope appear raised relative to the actual terrain. For example, if 10 cm tall grass is dense, the point cloud will appear to show the slope raised by about 10 cm. Ideally, mow grass beforehand to expose the ground. If that is impractical, avoid windy days when leaves move, conduct imaging when vegetation is still, or remove obvious floating points (noise) in post-processing. Wet or highly reflective surfaces or very dark surfaces can also be harder for lasers and cameras to capture, increasing errors. Avoid measuring immediately after rain or under extreme contrast between sun and shade; perform measurements under favourable conditions whenever possible.

Sunlight and illumination conditions: Smartphone LiDAR (infrared laser) can suffer from noise outdoors under strong sunlight. In bright conditions, LiDAR effective range and accuracy tend to degrade, so outdoor daytime measurement often relies primarily on photo-based point cloud reconstruction, with LiDAR supplementing short-range detail. Photogrammetry is also affected by lighting: strong noon sunlight can create harsh shadows or overexposed areas that obscure detail, while late-afternoon low light can produce noisy images and hinder feature detection. Ideally, measure under light cloud cover for uniform illumination, or schedule imaging in the morning or late afternoon with gentle oblique light. In tunnels or under bridges where sunlight does not reach, illuminate the area with the smartphone light or portable lamps to prevent missing data.

LiDAR resolution and distance attenuation: Smartphone LiDAR sensors are limited in output and scan range compared with professional laser scanners used in civil engineering. The point density obtainable is not very high, and when scanning targets several meters away, point spacing may exceed several centimeters. The further the distance, the more the light disperses and reflections weaken, so effective ranging is typically limited to about 5–10 m. Therefore, attempting to scan a wide slope from a distant position will not yield sufficient point density or accuracy. Smartphone point cloud surveying requires getting as close as possible to the target and walking to cover the slope with successive close-range measurements. When combining with photogrammetry, smartphone camera pixel resolution also affects accuracy: distant objects occupy fewer pixels and become coarse in the point cloud. While modern smartphone cameras are high performance, they have limits for capturing distant detail. Aim to fill the frame with the target at appropriate focal length/distance, and use zoom if necessary. The key is that, given equipment-specific resolution and range limitations, the more you take data "close-up and detailed," the better the accuracy.

By understanding these factors and applying countermeasures, the reliability of smartphone × GNSS slope measurement is further enhanced. Modern software increasingly offers AI-based noise removal and positional correction features, and appropriate post-processing can reduce residual errors to within a few centimeters. In the field, keeping these error sources in mind and performing additional measurements and calibrations as needed will ensure the final deliverables meet accuracy requirements.

Summary: Rationality and Value of Introducing Smartphone × GNSS

As discussed above, combining smartphone-based point cloud measurement with GNSS positioning can be a rational and useful solution for slope surveying. The main points of its rationality are summarized below.

• Balancing accuracy and coverage: It is groundbreaking in that it achieves centimeter-level accuracy while remaining easy to use and enabling a single operator to cover wide areas. It records the entire slope as a high-density point cloud, contributing both to quality control and efficiency.

• Improved work efficiency and safety: With only a smartphone and a small GNSS receiver, there is no need to carry heavy equipment up slopes. Work that formerly required multiple people can be completed by one person in a short time. Being able to acquire data remotely reduces high-altitude work risks and makes frequent progress checks and as-built confirmations easy, offering major safety benefits.

• Cost-effectiveness and low adoption barrier: Initial investment is lower than purchasing dedicated 3D scanners or expensive surveying instruments, and existing smartphones can be leveraged, simplifying equipment management. App operation is intuitive and training costs are low; in-house 3D measurement capability reduces outsourcing expenses and enables rapid decision-making through immediate data sharing.

Therefore, smartphone × GNSS slope measurement achieves high accuracy, safety, and low cost simultaneously, and can strongly support on-site DX (digital transformation). Enabling in-house, easy 3D measurement—previously outsourced to specialist firms—can reduce costs, improve responsiveness, and represent a first step toward data-driven advanced construction management.

Conclusion: Moving Slope Management to the Next Phase with LRTK-Based Simple Surveying

Traditionally, slope surveying and management have required significant effort and expertise. However, the approach of smartphone surveying using LRTK is ushering in an era in which anyone can perform precise 3D measurement with ease. Centimeter-level slope point cloud data can be directly used for design comparisons, volume calculations, and even deformation monitoring. For example, periodic scans of a slope after construction allow detection of slight landslides or deformations in digital data, enabling early maintenance decisions. This represents a new management method that was not possible with conventional visual inspections or spot measurements.

The spread of simple GNSS × point cloud surveying will elevate slope management to the next phase. When site personnel themselves can instantly acquire and share high-accuracy terrain data, data-driven decision-making can be applied consistently from construction management through maintenance. Slope management that once relied on "intuition and experience" will evolve into a scientific process based on evidence. By appropriately adopting advanced technologies such as LRTK, accelerate your slope management DX and step toward next-generation safe and secure infrastructure maintenance.