From RTK Point Clouds to Surface Models: A Practical Guide

In recent years at surveying sites, RTK (real-time kinematic) positioning technology and the use of point cloud data have been transforming traditional surveying methods. This article is a practical guide aimed at beginner to intermediate surveying professionals that explains the procedures and key points for generating surface models (ground surface models) from point cloud data obtained by RTK surveying, and how to utilize them in practice. It covers basics to applied examples, tips for improving accuracy, and examples of introducing the latest surveying devices.

Contents

• Basic knowledge of RTK surveying

• How to acquire RTK point cloud data

• Post-processing and shaping of point clouds

• Generating surface models

• Practical application examples

• Practical advice to improve accuracy

• Comparison with mobile simple surveying and advantages

• Introduction of surveying with LRTK

• FAQ

Basic knowledge of RTK surveying

RTK surveying is one high-precision positioning method using GNSS that achieves centimeter-level positioning by having two receivers—the base station and the rover—communicate in real time. Standalone positioning (standalone GPS/GNSS) typically produces errors on the order of several meters (several ft) due to atmospheric errors and satellite orbit errors, but RTK has the base station send correction information from a known accurate coordinate position which the rover receives to correct positioning errors in real time. As a result, position accuracy can be improved from meter-level to on the order of a few centimeters (a few in).

There are various RTK schemes. The basic form is one base station to one rover, but by using a network RTK (Ntrip) service that leverages nationwide continuous reference station networks without deploying a local base, you can obtain stable correction information over a wide area. Also, satellite positioning systems are not limited to GPS; they can use multiple constellations such as GLONASS, Galileo, and QZSS (Michibiki), which increases the number of receivable satellites and improves positioning reliability.

RTK positioning has solution status distinctions called Fix solution and Float solution. When the rover succeeds in resolving the integer ambiguities of the carrier phase, the solution becomes Fix, yielding horizontal and vertical accuracies typically within a few centimeters (a few in). Conversely, when the solution has not yet converged it is displayed as Float or Single, and may include errors on the order of tens of centimeters (tens of in) or more. In practice, it is important to record points only after obtaining a Fix solution. With sufficient satellite visibility and a stable communication environment, Float typically converges to Fix within tens of seconds to a few minutes. If a Fix cannot be obtained, take countermeasures such as avoiding nearby obstructions, adjusting the receiver position or antenna height, or checking the reception status of correction data from the base station.

The key to RTK’s high accuracy is the combination of real-time processing and relative positioning. By making relative measurements to the base station to cancel error factors and computing solutions sequentially to obtain immediate results, RTK offers the advantage of confirming positions on the spot. Compared with conventional surveying using total stations, RTK provides significant benefits in terms of broad-area positioning without line-of-sight constraints and labor savings.

How to acquire RTK point cloud data

RTK provides high-precision positioning data, but to collect this as large amounts of point cloud data you need to combine it with measurement methods such as laser scanning or photogrammetry. Recently, combining RTK-GNSS with various sensors allows immediate assignment of absolute coordinates to acquired point clouds, greatly reducing post-processing. Here we describe representative point cloud acquisition methods and how to choose based on field conditions.

• Drone photogrammetry (UAV photogrammetry): Use RTK-capable drones to generate point clouds (3D models) from aerial photos. This covers wide areas efficiently from above and can simultaneously produce orthophotos and DSMs (digital surface models). With RTK, shooting positions for each photo are determined with high accuracy, allowing the reduction or elimination of many traditional ground control points (GCPs), simplifying processing while maintaining accuracy. However, it is difficult to apply in areas where drone flight is problematic, such as dense forests or urban canyons.

• Terrestrial laser scanner (TLS): A tripod-mounted laser scanner directly measures the ground and structures. Point clouds at each scan position are acquired in relative coordinates, but by measuring the scanner setup position with RTK you can assign geodetic coordinates to the acquired point clouds. When performing multiple scans, setting RTK coordinates at each position reduces the burden of registration. TLS yields high-density, detailed point clouds but the equipment is large and setup is time-consuming.

• Mobile mapping (mobile measurement): Use vehicle-mounted or handheld mobile scanners to acquire point clouds while driving or walking. SLAM (simultaneous localization and mapping) can estimate relative positions even where GNSS is interrupted, but long-distance movement may accumulate errors that distort the point cloud. If equipment that can use RTK is available, self-position can be corrected with high accuracy during movement, suppressing point cloud drift and distortion. In urban areas where GNSS signals are unstable, consider also using ground reference points for correction or post-processing adjustments.

• Direct RTK surveying (sampling survey): Measure many representative points on the ground with RTK and treat them as point cloud data to create a terrain model. For example, measuring ground elevations at regular grid intervals by RTK yields a coarse point cloud dataset. This method requires no special scanner, so equipment costs are low, but manual measurement limits point density and coverage. For wide-area and detailed terrain mapping, drones or scanners are practical, but combining direct RTK surveying as a supplement can provide spot accuracy checks and comparisons.

The optimal acquisition method depends on field conditions. Drone photogrammetry is effective for open development sites and farmland, while terrestrial laser scanning or ground surveying is indispensable in forested mountainous areas. In urban areas where buildings cause GNSS positioning errors, acquire details with ground-based scanners and perform RTK positioning where the sky is open to correct global coordinates. The important point is to combine RTK accuracy and the advantages of point cloud measurement to efficiently obtain required data.

Post-processing and shaping of point clouds

Raw point cloud data often contains measurement noise and unnecessary points, so post-processing for data cleaning and shaping is essential. First, remove isolated or spurious points that are clearly measurement errors using noise removal filters. Laser scans can generate erroneous points due to sunlight reflection or weak returns at long ranges, and photogrammetry can produce floating points due to matching errors. Removing such noise points improves the accuracy of subsequent processing.

Next, perform density adjustment of the point cloud. If point density is uneven across the acquisition area, analyses may be biased by region. It is effective to resample at regular intervals in software or extract representative points in a grid to create a uniform point spacing. If the data volume is excessively large, thin out points to reduce data while retaining analytical capability. However, retain points in important terrain-change areas (edges and steep slopes) as much as possible.

When integrating multiple measurement datasets, registration and coordinate transformation are also required. Even if absolute coordinates are assigned to each point cloud by RTK, small instrument errors or residual GNSS errors can cause discrepancies of several centimeters between datasets. To correct this, adjust translations and rotations between point clouds using overlapping correspondences or known reference points. Additionally, classify features by type (ground, buildings, vegetation, etc.) and remove or layer non-target points as appropriate. For example, for creating a terrain surface model, remove vegetation and vehicles that obscure the ground surface.

Through these post-processing steps you obtain clean point cloud data suitable for analysis and modeling. At this stage, confirm whether transformation to the surveying coordinate system has been applied (apply corrections to plane rectangular coordinate systems or vertical datums as needed). With prepared point clouds, proceed to generating the surface model.

Generating surface models

By creating surface models that represent the terrain or structure surfaces from point cloud data, you can produce deliverables directly useful for volume calculations and drawing production. Typical surface model types are TIN (triangulated irregular network) represented by triangular meshes and DEM (digital elevation model) represented by a gridded elevation dataset.

The TIN method connects neighboring points in the point cloud as triangle vertices to represent the current terrain as a polygonal mesh. It can express local details even with uneven point distribution and faithfully reflects features of the source data. On steep or complex terrain, triangles can subdivide to follow fine changes, producing a high-accuracy model. However, when generated from large numbers of points the triangle count can become huge and data heavy.

The DEM method divides the area into regular grid cells and stores a representative elevation for each cell. For example, a 1 m (3.3 ft) grid DEM has the image of one representative height per square meter. DEMs have simple data structures, making contour generation and volume calculations straightforward. On the other hand, terrain features smaller than the grid spacing cannot be represented and details are smoothed, so selecting an appropriate resolution according to analysis objectives is necessary.

Beyond terrain, when dealing with detailed shapes of concrete structures or plant piping, you can also directly generate 3D mesh models from high-density point clouds. These are 3D shapes closer to CAD models, suitable for reverse engineering and visualization. However, in general civil surveying work, surface models representing ground undulations are primarily used.

Various specialized software and CAD tools are available for surface model generation, and many can automatically generate models with button operations if the point cloud is well prepared. After generation, check the model and, if there are holes (missing data), perform interpolation filling from surrounding areas, or smooth obvious unnatural protrusions. The completed surface model can be used for downstream analyses such as comparing with design data and calculating volumes and cross sections.

Practical application examples

Once you can generate surface models, the data can be applied to various civil and surveying tasks. Here are representative practical application examples.

• Earthwork quantity calculation (cut-and-fill volume calculation): By comparing the on-site terrain surface model with the design model, you can accurately calculate required fill and cut volumes. For example, from pre-construction ground and planned ground models, total earthwork and location-specific excavation/fill can be automatically calculated. Point cloud–based models reflect detailed undulations, enabling more reliable quantity control than calculations based on a few traditional survey points.

• Section drawing and longitudinal/transverse profile creation: Create longitudinal and transverse profiles from the terrain model along arbitrary lines. In road and river design, understanding existing cross-section shapes is important. Surface models generated from point clouds allow arbitrary section extraction, enabling accurate drafting of the current situation. This is useful for creating transverse sections at specified intervals for river embankments or roadworks and comparing as-built shapes to design shapes.

• Comparison and verification with design data: Overlay the completed surface model with the design 3D data (design surface) to visualize differences between as-built and planned conditions. Create elevation difference maps to show finish errors in color distributions, or calculate excess/deficit earth volumes by area. In construction management, high-accuracy point cloud–based comparative analysis is useful for inspecting whether as-built conditions match design and for estimating additional work quantities.

These applications link information obtained from RTK point clouds directly to project quality and schedule management. Especially for earthwork management, regularly scanning the site with drones or LiDAR and performing volume calculations streamlines progress monitoring and earned-value assessment. RTK’s high-accuracy assurance allows these measurements to be relied upon for decision-making.

Practical advice to improve accuracy

When handling RTK point clouds, always keep in mind the importance of accuracy management for positioning and measurement data. Below are practical tips you can apply in the field to improve accuracy.

• Optimize satellite reception environment: During positioning, choose locations with as much open sky as possible. If sky visibility is obstructed, the number of satellites decreases, delaying Fix acquisition and reducing accuracy. Near high-rise buildings or in forests, multipath (reflected signals) has a large effect, so raise the antenna or temporarily move away from obstructions.

• Check device settings and confirmations: Pre-check RTK receiver settings (enable multi-GNSS, input base station coordinates accurately, set antenna height, etc.). On-site, monitor the solution type (Fix/Float/Single) and DOP values (position dilution of precision) on the display and address abnormalities immediately. For example, if Float persists, pause measurement and investigate the cause.

• Always record on Fix solutions: As noted above, record point coordinates only when a Fix solution is obtained. Points recorded while in Float may produce critical errors in elevation differences or distance calculations. If Fix cannot be obtained, wait a few minutes or recheck base station correction reception (for Ntrip, verify internet connectivity; for radio, verify signal line-of-sight). Restarting the app or receiver can also resolve some cases.

• Multiple measurements and averaging: For static point measurements, measuring the same point multiple times and taking the average improves accuracy. Some receivers like LRTK have automatic averaging functions that compute coordinates from statistical processing of dozens of observations. If time allows, remeasure at different times and compare results to confirm that no large differences occur.

• Compare with known points: If there are public control points or existing survey marks near the site, measure them with RTK and compare to their known coordinates. If errors are within 1–2 cm (about 0.4–0.8 in), it indicates the day’s surveying is being conducted at a reliable accuracy. If large discrepancies appear, immediately investigate possible datum setting errors or positioning environment problems.

• Check point cloud data quality: After point cloud acquisition, quickly inspect point renderings for outliers or abnormal distortions. Even with high-precision RTK, sensor malfunctions or processing errors can introduce anomalies. Perform a coarse 3D view check of the point cloud and, if obvious anomalies exist, perform additional measurement or data processing corrections.

By following these points you can further improve the accuracy and reliability of RTK point cloud surveying. Small improvements to maintain high accuracy directly affect the quality of final deliverables.

Comparison with mobile simple surveying and advantages

Simple surveying apps using smartphone or tablet GPS functions have become widespread, allowing anyone to easily obtain position information. They are convenient for preliminary field surveys and approximate location checks, but they are far less reliable than surveying tasks requiring high precision.

The biggest difference is positioning accuracy. Typical built-in smartphone GPS has errors of several meters and can deviate more than 10 m (over 30 ft) under building or tree cover. In contrast, RTK surveying, as noted above, confines errors to around a few centimeters (a few in). Vertical accuracy is also within a few centimeters with RTK, making it suitable for earthworks and foundation construction control.

There is also a difference in data quality. Simple surveying yields single point coordinates or track logs, whereas combining RTK with point cloud measurement acquires detailed 3D information. Point cloud data digitizes the existing terrain itself, enabling later extraction of arbitrary sections and measurements—offering more value than mere position logs.

Until recently, RTK surveying required expensive dedicated equipment and skilled operation, hindering ease of use. But now, solutions that combine mobile devices with RTK to achieve both ease and accuracy have emerged. Using small RTK receivers that attach to smartphones enables centimeter-level positioning and point cloud scanning in the field without carrying heavy traditional equipment. The next section introduces the device “LRTK” developed by Lefixea as a representative example.

Introduction of surveying with LRTK

LRTK embodies the mobile×RTK concept as an all-purpose surveying device. Lefixea’s [LRTK Phone](https://www.lrtk.lefixea.com/) is an ultra-compact RTK-GNSS receiver attachable to smartphones and tablets, realizing RTK positioning, point cloud measurement, photogrammetry, and AR visualization in one unit. Despite its pocketable size weighing only approximately 125 g, it runs on an internal battery and is easy to carry on site.

The strength of LRTK is combining ease of use with high functionality. Using a smartphone and a dedicated app, you can start single-point positioning or continuous scanning with one button, and all acquired point clouds and coordinate data are assigned global coordinates in real time. Problems that plagued smartphone-only scans—“point clouds lacking absolute coordinates” and “walking scans producing distorted data”—are solved by LRTK because it continuously corrects self-position with cm level accuracy (half-inch accuracy) during measurement. Anyone can intuitively operate it to obtain high-accuracy point-cloud-attached as-built data.

Acquired data can be uploaded to the cloud instantly for sharing, and you can view point clouds in a browser and measure distances, areas, and volumes. For example, you can upload terrain point clouds surveyed on site with LRTK, overlay the design 3D model, and compute cut-and-fill differences on the spot. Tasks that previously required returning to the office and using specialized software can now be completed in real time on site.

LRTK is also easier to introduce price-wise compared to traditional surveying equipment, enabling a true “one-person-one-device” era. With a simple UI and ample guidance features usable by non-experts, field staff themselves can incorporate high-precision surveying into daily work. The entire workflow from RTK point clouds to surface models can be dramatically streamlined using LRTK. As a representative of increasingly proliferating mobile RTK devices, LRTK is advancing the practice of surveying significantly.

FAQ

Q: What is the difference between RTK surveying and ordinary GPS surveying? A: Ordinary standalone GPS surveying has errors on the order of several meters, whereas RTK uses correction information from a base station to reduce errors to a few centimeters (a few in). RTK also provides real-time results so positioning can be confirmed immediately. Standalone GPS is adequate for rough location awareness, but RTK’s high precision is essential for design and construction management.

Q: What are Fix and Float solutions? A: They are status indicators of solution quality during RTK positioning. A Fix solution means integer ambiguities are resolved and centimeter-level accuracy is achieved. A Float solution means the solution has not yet stabilized and has large errors, so it is not reliable. In surveying, record coordinates only when Fix is obtained and wait for Fix if the solution is Float or Single.

Q: Is a base station always necessary for RTK surveying? A: You do not always need your own base station. By subscribing to a network RTK service called Ntrip that uses public reference station networks, you can survey with only a rover. In Japan, systems are also being developed that use correction signals from the QZSS (Michibiki) to achieve centimeter-level positioning without a base station. However, for aligning to a local high-precision reference frame or where communication is unstable in mountainous areas, deploying a local base station is still more reliable.

Q: What are the benefits of making a surface model from point clouds? A: Point clouds are huge collections of coordinates and can be difficult to handle as-is. Converting them into surface models represents terrain undulation as surfaces, making practical analyses like volume calculation and section drawing easy. It also reduces data size, improving display and computation performance in software. In short, surface modeling organizes point clouds into “usable” data.

Q: Is RTK surveying possible with only a smartphone or tablet? A: Smartphone GPS alone is insufficiently accurate, but combining it with an external RTK-capable receiver makes it possible. For example, devices like Lefixea’s LRTK Phone attach to commercial smartphones and enable centimeter-level positioning and point cloud scanning. Using dedicated apps, operation is simple and a smartphone can become a high-precision surveying tool.