Five Procedures for Point Cloud Acquisition Using RTK Positioning: How to Avoid Failures on Site

When you consider acquiring point clouds using RTK positioning, many practitioners are less interested in the theory of positioning itself than in one practical question: how should you organize the workflow to minimize rework on site? RTK is indeed a powerful method that enables real-time, several-centimeter-level positioning, but point cloud quality is not determined by RTK equipment performance alone. Alignment with known control points, placement of control and validation points, planning of imaging or traversal, on-site inspections, and post-generation verification must all connect in order to produce usable point clouds. The Geospatial Information Authority of Japan (GSI) also states that RTK methods can achieve centimeter-level positioning through simultaneous observation and correction using a reference station, and that network RTK is an efficient way to achieve cm-level positioning in real time. In addition, because point cloud creation methods differ between photogrammetry-derived and laser-derived approaches, proceeding without a clear objective often leads to later issues with accuracy or missing data.

In practice, RTK in the field functions less as a standalone technology and more as the foundation for correctly placing point clouds in a coordinate system. Official operational guidelines also indicate that network RTK can be used not only for terrain and object measurement but also for setting control points for aerial photogrammetry, supplemental terrain surveys, and on-site supplementary surveys. Moreover, recent amendments to working procedures have expanded the scope of three-dimensional point cloud surveys to include UAV point cloud surveys, terrestrial laser point cloud surveys, UAV LiDAR surveys, and vehicle-mounted photogrammetry-LiDAR surveys, so opportunities to combine RTK with point cloud acquisition are certainly increasing. In other words, measuring points with RTK is not the ultimate goal; the real objective is to produce point cloud deliverables that can be used downstream for comparison, cross-sectioning, quantity calculation, and maintenance management.

Table of Contents

• Why point cloud acquisition using RTK positioning is drawing attention

• Step 1 Decide measurement method based on objectives and deliverables

• Step 2 Finalize coordinate system, known control points, and validation point plan

• Step 3 Check field conditions and create acquisition plan

• Step 4 Carefully acquire control and supplementary points with RTK positioning

• Step 5 Perform verification and correction after point cloud generation

• Practical notes to avoid failures on site

• Summary

Why point cloud acquisition using RTK positioning is drawing attention

The main reason RTK positioning is attracting attention for point cloud acquisition is that it can turn point clouds from mere visual 3D data into deliverables with reliable coordinates. If a point cloud is not correctly referenced to a coordinate system, it becomes difficult to use for practical tasks such as overlaying with existing drawings, comparing with design models, as-built verification, earthwork quantity calculation, or creating cross-sections. GSI’s guidance also defines terrain surveying using network RTK as the task of measuring terrain and features based on known control points, indicating that alignment with known points is a premise for point cloud utilization.

RTK’s role also varies depending on where it is used in the point cloud workflow. One role is measuring control and validation points on site; another is positioning photos or laser scans taken while moving; and a further role is for on-site supplementary surveys to fill gaps. Official guidance states that network RTK can be applied to establishing control points for aerial photogrammetry, supplemental terrain surveys, and on-site supplementary surveys, showing that it is deeply involved in pre- and post-acquisition processes. Focusing only on point cloud creation software often overlooks the importance of those surrounding processes, and many on-site failures actually occur in these steps.

Furthermore, the methods for acquiring point clouds themselves differ. Photogrammetry-derived point clouds reconstruct elevations of objects captured in images, so ground sample distance and overlap are fundamental to accuracy. In contrast, laser-derived point clouds rely on pulse density as the basis of accuracy and may be able to observe terrain through gaps in vegetation. However, both methods include non-terrain objects and therefore require filtering. Without understanding these differences and assuming that “RTK will solve everything,” you risk choosing an acquisition method unsuitable for your objectives.

Therefore, the truly important thing when acquiring point clouds using RTK is not chasing device names or trendy methods. Decide first what deliverables you want, at what accuracy, where, and within what time frame, and then incorporate RTK only to the extent necessary to achieve those goals. Following this order stabilizes decision-making on site and simplifies post-processing and verification. Considering the differences between photogrammetric and laser point clouds and RTK’s usefulness for control points and supplementary surveys, the least failure-prone approach is to plan the workflow by working backward from the desired deliverables.

Step 1 Decide measurement method based on objectives and deliverables

The first step is to clarify what you want to produce. This may seem obvious, but it is frequently omitted in practice. For example, the required point density, imaging conditions, and validation methods differ depending on whether the final deliverable is an orthoimage, a 3D terrain model, slope deformation checks, pre/post development comparisons, or earthwork quantity calculations. Official manuals state that imaging plans for 3D point cloud creation should be established considering position accuracy, ground sample distance, flying height above ground, equipment used, terrain shape, land cover, and weather conditions; the underlying assumption is to determine deliverables and accuracy up front.

Next, decide whether to center the workflow on photogrammetry, on LiDAR, or to use both. Photogrammetry-derived point clouds efficiently capture wide areas as surfaces and are easy to interpret visually, but they cannot reconstruct surfaces that do not appear in images. LiDAR-derived point clouds directly measure terrain surfaces and can sometimes capture ground through vegetation gaps, but points are discrete and post-processing for shaping and classification is important. In short, the choice of acquisition method depends on whether the focus is on building exteriors and pavement surfaces or on capturing terrain under vegetation and slope toes.

Importantly, for projects with strict accuracy requirements, do not rely on RTK alone. Public manuals on creating 3D point clouds from UAV photos state that when the required positional accuracy of the 3D point cloud is strict—such as within 0.05 m (0.16 ft)—surveying using a total station (TS) alone may be allowed from the perspective of securing the positional accuracy of control and validation points. In other words, RTK is convenient, but for high-accuracy projects it is safer to plan on using RTK only as an auxiliary method or in combination with other techniques. Deciding on the assumption “RTK is sufficient” on site increases the risk of failing to meet accuracy requirements and needing to re-survey later.

Also, decide at this stage where RTK will be used. Will you use RTK only to measure control and validation points, for positioning the onboard sensors, or also for on-site supplementary surveys? GSI’s guidance indicates that network RTK can be applied to terrain surveying itself, to terrain surveying combined with TS, or to only some stages of a terrain survey. In other words, RTK is not an all-or-nothing choice; it is a technique that should be applied optimally by stage.

If you choose a photogrammetry-centered acquisition, finalize imaging conditions early. Official standards recommend planning to ensure image overlap of at least 80% within the same flight line and 60% between adjacent lines if post-flight verification of actual image overlap is possible. If such post-flight verification is difficult, the plan should assume at least 90% overlap within the same line and 60% between adjacent lines. Considering both deliverables and verification methods and building a margin into the initial acquisition plan is the quickest way to avoid re-flights or re-runs.

Step 2 Finalize coordinate system, known control points, and validation point plan

The second step is to confirm in advance which coordinate system you will use to produce deliverables. Entering the site without clarity here risks RTK-measured points not matching other drawings or models later. GSI’s manual specifies that known control points used in terrain surveying with network RTK should have accuracy equivalent to at least Class 4 reference points. In other words, the coordinate foundation used on site must be aligned with reliable known points; not just any reference will do. This principle applies in practice even when the work is not a public survey. If control points are ambiguous, the entire point cloud anchored to them becomes ambiguous.

You should also unify how heights will be handled. Even if horizontal positions align, differing vertical datums will make earthwork and cross-section judgments unusable. GSI materials recommend performing vertical datum transformations in terms of orthometric heights and using at least three control points for such transformations. In practice, decide before starting work whether to use the Plane Rectangular Coordinate System (zone number), ellipsoidal heights or orthometric heights, how to align with existing control point results, and whether conversion to a site-local coordinate system is necessary. Trying to fix mismatches later in post-processing often leaves unexplained vertical offsets.

At this stage, it is also important to distinguish between control (ground) points and validation points. Public manuals for generating point clouds from UAV photos clearly define control points as reference points necessary for 3D shape reconstruction calculations and validation points as points used to verify the 3D point cloud. The guidance also suggests that the number of validation points should be at least half the number of control points and that they should be distributed evenly within the measurement area. A frequent on-site mistake is to use the same points for both control and validation and claim those points as verification, but that does not provide independent accuracy confirmation. To truly assess point cloud quality, you must separate control points and independent validation points.

There are also strategies for placing points. Control points should not be limited to the periphery; interior points are necessary as well. Official explanations advise placing interior control points in addition to exterior ones, and placing control points in areas with large relative height changes or areas with little surface texture to help secure measurement accuracy. This matches practical experience: flat pavement, long slopes, shadow-prone areas, and bare ground that is hard to distinguish from surroundings tend to make automatic feature extraction and point cloud alignment unstable, so strategically placing RTK-acquired points in such areas is valuable.

Also decide on a manageable recording method in advance. Standardize point naming rules, how to photograph installed marks, whether a sketch of current conditions is needed, mark materials, removal policy, and whether there are permanent reference features for later rechecks. GSI’s standard forms include control point installation accuracy management tables, transformation accuracy management tables to the Plane Rectangular Coordinate System, inspection survey result accuracy management tables, and 3D point cloud data file accuracy management tables, showing that recordkeeping by stage is expected. Neglecting on-site notes makes tracing causes difficult if coordinate transformations or re-computations are required later.

Step 3 Check field conditions and create acquisition plan

The third step is to perform checks before entering the site and immediately upon arrival. RTK errors do not arise solely from equipment performance shortfalls. GSI research materials list tropospheric delay, ionospheric delay, and multipath errors from reflections off surrounding features as major error sources in GNSS observation data. This means that even with the same equipment, solution stability varies greatly between open sky locations and places near trees, slopes, buildings, metal fences, vehicles, or water surfaces. On-site checks must consider not only the working area but also the surrounding environment.

Pay special attention to multipath and satellite geometry. GSI manuals advise considering other methods for poor signal environments, performing re-observations at different times for quality checking, and avoiding low-elevation satellites that are prone to multipath, so choose appropriate observation times based on satellite geometry. This translates directly to practice: in partially open sites or areas with many reflectors, do not prematurely conclude “it’s fixed now so it’s fine”—it is safer to confirm again at another time.

Understanding PDOP is also helpful when assessing satellite geometry. U.S. geodetic agency guidelines explain that DOP is an indicator of the influence of satellite geometry on positioning error, and that PDOP is a representative index for three-dimensional position uncertainty. Smaller values are better; for example, a PDOP around 2 is favorable for positioning, while higher values are associated with degraded solutions. On site, check not only the satellite count but whether satellites are widely distributed across the sky and whether DOP indices are worsening.

For photogrammetry-centered acquisitions, on-site inspection of captured images is as important as RTK verification. Public manuals recommend checking image results on site immediately after capture: confirm coverage, image quality, overlap, occlusions, and that all control and validation points are appropriately imaged. Check image quality for blur, motion blur, and noise; if there are areas with missing coverage, occlusions that cannot be compensated for, improper images, or if the actual flying height deviated more than 10% from the planned flying height, re-shooting is required. Even if RTK yields accurate point positions, missing or blurred imagery will prevent creating a good point cloud.

When using network corrections for RTK, verifying communications is essential. GSI guidance states that if satellite signals are lost during observation, reinitialization should be performed; if cellular communication is interrupted or if you move beyond a certain distance from the point for which a virtual station was requested, re-request correction data or network correction parameters. In practice, don’t be complacent just because correction data appears to be available—plan for movement ranges, communication-prone areas, and time required for reinitialization. Ignoring this in valleys, beneath slopes, in machine shadows, or in unstable-development sites can halt work.

Step 4 Carefully acquire control and supplementary points with RTK positioning

The fourth step is to handle RTK-measured points with care. On site, there is a temptation to adopt a measurement the moment a FIX indication appears, but that is insufficient. GSI manuals note that although a single set of 10 epochs at 1-second intervals may complete quickly, observations are part of the initialization and convergence of integer bias, so confirm that a FIX solution is obtained and that positioning quality indicators are satisfactory before accepting values. In other words, RTK practice is not just about “FIX or not FIX”; it includes assessing the stability of the FIX and the quality indicators.

When measuring control and validation points, make it routine to input the pole height, check the bubble level, center the instrument over the point, and check for surrounding occlusions for each point. Public manuals also recommend performing two sets of observations for control and validation points using RTK or network RTK, designating the first set as the adopted value and the second as a check value, with acceptable set-to-set discrepancies of 20 mm (0.79 in) for X/Y components and 30 mm (1.18 in) for Z. When pressured by time on site, the temptation is to record a single observation, but these small redundant checks greatly reduce rework. Proceeding with only the initial observation makes it hard to determine later whether a detected offset is due to point cloud processing or to the field measurement.

Time-shifted re-observations are also effective at critical points. U.S. geodetic agency guidelines explain that RT positioning of important points cannot be fully trusted without redundancy; averaging positions obtained three to four hours apart—thus under different satellite and multipath conditions—tends to be more accurate. GSI also advises conducting re-observations at different times for quality checks. It is not necessary to do this for every point, but for control points that form the basis of a work area, for points that underpin downstream processes, and for points whose rework costs are high, scheduling re-surveys at different times is worthwhile.

Also, do not be complacent with control points alone. Independent validation points are required for point cloud verification. Public manuals define validation points as points used to verify the 3D point cloud. The workflow for checking 3D reconstruction results specifies comparing the coordinates of pre-determined validation points with those obtained from reconstruction calculations; if they are out of the required positional accuracy, remove bad images, correct feature points, re-run calculations, and if necessary perform additional imaging. On site, what matters is not “it seems to have processed well” but “it meets accuracy even when checked at independent points.”

Moreover, in areas with strong occlusion or proximity to structures, do not insist on RTK alone. GSI guidance states that poor signal environments may require consideration of other methods, and other clauses outline terrain surveys that combine network RTK and total station measurements. Therefore, under eaves, in narrow passages, near stands of trees, adjacent to retaining walls, or under bridges, it is usually faster and more accurate to split points between RTK and other methods. The more convenient a technique appears, the more valuable the judgment not to use it can be for protecting quality.

Step 5 Perform verification and correction after point cloud generation

The fifth step is not to finish at converting acquired data into a point cloud. Point clouds are not immediately usable deliverables upon acquisition. Official materials explain that both photogrammetric and laser point clouds include non-terrain objects and therefore require filtering, which can lead to partial data gaps or interpolation. Standard forms include 3D reconstruction accuracy management tables, point density inspection accuracy management tables, inspection survey result accuracy management tables, point cloud correction accuracy management tables, and 3D point cloud data file accuracy management tables, indicating that multiple aspects of inspection are expected after point cloud generation.

Priority in verification should be comparing against independent validation points. Public manuals for 3D reconstruction from UAV photos stipulate checking that the coordinates of validation points match within required positional accuracy the coordinates obtained from reconstruction calculations; if not, remove bad photos or correct feature points and recompute, and if still not satisfied perform additional imaging. This approach applies beyond photogrammetry to other point cloud processing as well. When errors occur, do not immediately shift the entire dataset manually to match; instead, sequentially trace whether the cause lies in acquisition, control points, image quality, insufficient overlap, or processing parameters.

For practical reuse of deliverables, coordinate transformations and metadata management are also indispensable. GSI standard forms include a separate transformation accuracy management table for the Plane Rectangular Coordinate System, treating correct remapping into a coordinate system as a process step in 3D point cloud surveying. Practically, before delivery you should at minimum document the adopted coordinate system, how heights were handled, known control points used, lists of control and validation points, acquisition dates, acquisition methods, point density rationale, whether filtering was applied, and areas prone to data gaps. If this is left ambiguous, you will waste time interpreting coordinate differences months later when overlaying with other zones or comparing with re-surveys.

For terrestrial laser-based point clouds, integrating multiple scans is also important. GSI’s published notes on terrestrial laser point cloud composition describe processing individual scans in software to combine them, and then transforming the entire combined point cloud into the Plane Rectangular Coordinate System to create 3D point cloud data. This implies that on sites with many scans, it is more efficient to manage and transform the whole set collectively rather than performing coordinate transformations for each scan individually. The same concept applies to mobile or walking surveys: be mindful of loop closures and redundant paths on site, and finish by aligning to the coordinate system and confirming validation points.

Ultimately, the criterion is not whether the point cloud looks clean but whether it can support required decisions. Check whether heights are stable when cutting cross-sections, whether there are systematic errors when overlaying with existing drawings or design models, whether the ground extraction is sufficient for quantity calculations, and whether comparisons with future re-surveys can be made in the same coordinate system. Point cloud acquisition does not conclude with data capture; it only becomes a deliverable after verification and correction. Adopting this mindset significantly changes both on-site acquisition and post-processing approaches.

Practical notes to avoid failures on site

A common failure when using RTK for point cloud acquisition is equating RTK’s high-precision positioning with high-precision point cloud deliverables. RTK is powerful, but GNSS observations are subject to multiple error sources—tropospheric delay, ionospheric delay, multipath—and positioning reliability is greatly influenced by the field environment. U.S. geodetic agency guidelines also note that real-time positioning is harder to validate than static observations and requires understanding many variables and careful attention by field operators. In short, if you use RTK, operate it as part of a workflow that includes reading observation conditions rather than relying solely on the equipment.

The second failure is focusing only on control points and neglecting validation points. Public manuals define validation points as independent points for verifying 3D point clouds and prescribe even distribution and required counts. They also specify that transformed point clouds are checked against these validation points and, if accuracy is not met, re-computation or additional imaging follows. On site, you may be tempted to accept “it looks roughly right” from control points, but that is not true quality assurance. If you want to trust the resulting point cloud, you must critically examine it using independent validation points.

The third failure is skipping on-site checks of images or traversal logs. The idea of checking image quality, overlap, occlusions, and the capture of control/validation points immediately after shooting is not mere formality. Discovering missing data or blur after entering post-processing can make re-visits very costly. Especially at slope tops, behind structures, near trees, backlit areas, and monotonous pavement, problems that look acceptable on site can later produce feature shortages or excessive noise in post-processing. Checking and re-shooting as needed on the spot is ultimately the most efficient approach.

The fourth failure is judging workability solely by satellite count. Even with many visible satellites, poor geometry can destabilize solutions, and a bad reflection environment can induce multipath that causes shifts of several centimeters or, in some cases, larger errors. U.S. guidelines emphasize PDOP as an important index for satellite geometry influence and describe how multipath from nearby structures can lengthen apparent distances and lead to centimeter-level noise or larger errors in integer ambiguity resolution in RT solutions. GSI likewise recommends avoiding low-elevation satellites and re-observing at different times. Make a habit of considering satellite count, geometry, and surrounding reflectors together.

The fifth failure is postponing management of coordinate and vertical datum decisions. Horizontal positions may appear to match, but differing vertical datums will cause problems in cross-sections, earthwork, and connecting adjacent zones. Public standards formally articulate alignment with known control points, the number of points used for vertical transformations, and transformations to the Plane Rectangular Coordinate System because these are fundamental to point cloud utilization. Before measuring on site, make sure everyone shares which coordinates will be used for deliverables and record this in file names and attributes to prevent subtle mistakes.

The final failure is using methods that are overkill or insufficient for the required deliverables. Photogrammetric and laser point clouds each have strengths and weaknesses; strict accuracy requirements often call for decisions not to rely solely on RTK. Conversely, adding excessive steps where high precision is unnecessary wastes time and manpower. The strongest practitioners in the field are not always those who use the most advanced methods but those who can appropriately combine RTK, photogrammetry, LiDAR, and supplementary surveys to meet required accuracy, coverage, and deadlines. Being able to subtract techniques to match objectives, rather than simply adding technologies, leads to failure-free field operations.

Summary

To avoid failures in point cloud acquisition using RTK positioning, first define the deliverables, then finalize the coordinate system and plan for known control and validation points, check field conditions and create an acquisition plan, carefully acquire points using RTK, and finally take responsibility for verification and correction after point cloud generation. In short, the key to success is not merely using RTK but deciding where and how to integrate RTK into the overall point cloud workflow. Differences arise more from alignment and verification than from acquisition alone. Following this order greatly reduces the risk of re-surveys and reprocessing.

If you want to operate the RTK–point cloud linkage more simply on site, using an iPhone-mounted high-precision GNSS device like LRTK can also be effective. Making the entry points of coordinates—control point acquisition, on-site supplementary surveys, and positional checks—easier to handle simplifies overall accuracy management for point cloud utilization. The quality of point cloud deliverables depends greatly on how you acquire the very first point. For those who want to establish RTK on site and build a system that reliably handles even simple surveys, reviewing operations starting with LRTK can help reduce failures in point cloud acquisition.