How do you verify the accuracy of drone surveying? 5 practical steps you can use in the field

Please translate the following input into English

When using drone surveying on-site, what many people in charge worry about first is, "Can this accuracy really be used?" Even if you can fly and produce point clouds and images, if the results are not at a level usable for as-built management, earthwork volume verification, construction planning, drafting, and explanations to stakeholders, you'll end up resurveying with another method. Whether it is useful on-site is determined not by whether you were able to capture the data, but by whether it meets the required accuracy.

In practice, verifying the accuracy of drone surveying is not particularly difficult. The important thing is not to judge anything just by how it looks. Even if it appears clean, it is not uncommon for coordinates to be shifted, elevations to be incorrect, or for local distortions to occur. Conversely, if you follow the verification procedures, you can quite quickly determine on site "how far these results can be used" and "where caution is needed."

Accuracy verification is not simply an exercise in finding failures. It is an important process that leads to reviewing the next flight conditions, the placement of control points, the way check points are collected, and the processing parameters. The causes of both success and failure only become visible through accuracy verification. For that reason, rather than performing it perfunctorily at the end of surveying operations, it is important to build accuracy verification into the planning stage.

In this article, we organize accuracy verification for drone surveying into a form that is easy to carry out on site and explain it as five practical steps. We explain things on a work-oriented, practical basis so they are easy to understand not only for surveyors with specialized knowledge but also for site supervisors, construction managers, construction consultants, and client-side personnel. We cover not only theoretical reasoning but also points that are easy to get confused about in the field and verification checkpoints that are easy to overlook.

Table of Contents

• Why accuracy verification is important in drone surveying

• Step 1 Decide in advance the required accuracy according to the purpose

• Step 2 Prepare reference points and checkpoints separately

• Step 3 Identify sources of error before and after flight

• Step 4 Compare the result data with the verification points and make a numerical determination

• Step 5: Make the final determination, by use case, of whether it can be used on-site.

• Common mistakes when verifying the accuracy of drone surveys

• Summary

Why Accuracy Verification Is Important in Drone Surveying

The importance of verifying accuracy in drone surveying is that the appearance of the deliverables can differ from their practical accuracy. Shooting broadly from the air lets you grasp the entire site in a short time and outputs point clouds and orthophotos in a tidy form. However, that alone does not mean accuracy is guaranteed. What is needed on site is not pretty images but that positions and elevations fall within a level usable for practical work.

For example, the required level of accuracy changes depending on whether it is used to understand site conditions before construction, to verify as-built results, to calculate earthwork volumes, or to check against the design. If the purpose is to grasp terrain trends across a large development site, some error may not be a practical problem. On the other hand, in situations where you want to evaluate at the scale of a few centimeters—such as checking the finish of slopes or the elevation of pavement surfaces—the strictness of checks can be completely different even when using the same method.

A common misconception here is the idea that using high-performance equipment will automatically result in high accuracy. In reality, accuracy is not determined by equipment alone. Flight altitude, camera angle, overlap rate, the placement of control points, how ground reference coordinates are obtained, processing conditions, terrain relief, the presence or absence of vegetation or water surfaces, the way shadows fall, and other factors all combine to determine accuracy. In other words, accuracy verification is not only a numerical check after completion but also an inspection of whether the entire operation was appropriate.

Furthermore, if accuracy checks are lax, major rework can occur in later processes. If, after slicing cross-sections from the point cloud to calculate earthwork quantities, you discover that the height reference was off, you will need to redo the quantity assessment itself. If local distortions are found after using the data for as-built verification, it will also affect the credibility of explanatory materials and reports. Skipping the initial check may seem to save time in the moment, but ultimately it makes additional corrections and re-surveys more likely.

Therefore, in drone surveying it is necessary to insert “verifying accuracy” between “flight operations” and “producing deliverables.” Moreover, that verification should not aim to produce detailed certification reports that only experts will read. It is important to be in a state where, on site and for anyone who looks, you can explain the magnitude of the errors, what purposes the data can be used for, and what cautions are required. Practical on-site accuracy checks should not end with simply producing numbers; they should enable decisions about where and how the data can be used.

Step 1 Decide the required accuracy according to the purpose

The first step in accuracy verification is to clarify the purpose of the survey before considering how to check it. If this is left vague, you cannot establish criteria for judgment even after performing the checks. The accuracy of drone surveys should not be described as simply good or bad in absolute terms, but evaluated based on whether it is sufficient for the intended use.

For example, in the early assessment of a provisional plan or when sharing the overall site situation, grasping the overall shape is more important than minor vertical errors. On the other hand, for as-built management, earthwork volume calculations, and checking interfaces with existing structures, errors in the vertical direction directly affect decisions. Even if the horizontal position is correct, vertical discrepancies will interfere with the boundary between cut and fill, drainage planning, and verification of the finished surface. Conversely, if elevations are relatively stable but the plan position is off, it becomes difficult to use for checks near boundaries or for confirming installation locations.

Therefore, the first things to decide are "what purpose it will be used for" and "which aspect of accuracy to prioritize." Clarify whether you primarily want to verify the horizontal position, primarily want to verify elevation, or both. Then determine a guideline for the acceptable error tolerance on site. The important point here is not to set standards stricter than necessary. Demanding accuracy that is excessively stricter than the level required on site only increases the workload. Conversely, if the tolerance is too loose, the results may become unusable later.

Also, the required level of accuracy varies depending on the nature of the target. Results can differ even with the same flight method between a leveled, unobstructed development site and a site with many trees or shadowed slopes, or one where structures are densely clustered. Whether you verify on relatively stable surfaces such as paved or graded areas, or on surfaces with large variability such as grassland or crushed-stone coverage, the way you evaluate them needs to change. In other words, required accuracy should not be decided only on paper; it is important to set it realistically while taking site conditions into account.

What site personnel should do here is be able to state the purpose of the accuracy check in writing. For example: "If it is to be used for grasping the current ground before construction, it is sufficient if the accuracy is at a level that can confirm the overall height trend"; "If it is to be used for as-built verification, place more emphasis on vertical consistency than on planimetric position"; "If it is to be used for earthwork volume comparison, prioritize consistency of the entire surface over local distortion." With this kind of clarification, both the locations where checkpoints are placed and the methods used to judge results after processing will be less likely to vary.

Also, what must not be forgotten here is aligning expectations with the client and internal stakeholders. The less familiar stakeholders are with drone surveying, the more likely they are to expect that “if it’s a drone, everything will be accurate.” If you share the intended use and required accuracy from the start, you can reduce later misunderstandings such as “it can’t capture details as finely as we thought” or “this area needed a separate survey.” Verifying accuracy is not only a technical issue but also an operational one.

The key point of this procedure is not to retroactively set the criteria for accuracy checks. Instead of applying a convenient evaluation after the results are produced, decide before measuring "what level of accuracy will be acceptable." That extra step becomes the turning point that makes drone surveying usable in the field.

Step 2 Prepare reference points and check points separately

The next step is to prepare control points that support accuracy and validation points that verify accuracy separately. This is a part that greatly affects the success or failure of accuracy checks, but it is also something that is easily confused in practice. If you use the same points to align the results and then judge the accuracy to be good at those same points, the verification becomes weak. This is because those points have already been used to produce the results.

Control points are points used to align the position and elevation of the resulting data. In other words, they serve as supports to tie the data to real-world coordinates. On the other hand, check points are points used to independently verify whether those results are truly correct. If you align the results using only control points, the areas around the control points may look good, but distortions can occur in distant locations or in intermediate areas. That is why it is necessary to treat separately the points used to produce the results and the points used to verify them.

When setting check points, simply increasing the number is not enough. Placement is important. If you place check points biased toward one side of the site, you may miss offsets on the opposite side. It is important to distribute them as evenly as possible across differing site conditions—along the perimeter and the center, on sloping and flat areas, near structures and in open spaces. In particular, you should deliberately place check points close to the locations where the results will actually be used. For example, if you prioritize earthwork quantity assessment, place them within the earthwork area; if you prioritize checking interfaces with structures, place them around those interfaces; if you prioritize evaluating as-built heights, place them near the surface being evaluated.

Careful consideration is also needed when choosing locations for check points. In areas with tall grass, mud, puddles, strong reflections, deep shadows, or ambiguous ground boundaries, it can be difficult to read positions both on images and in point clouds. Check points should be in locations that are easy to identify on site and where the same position can be clearly identified in the deliverables. Using distinct marks, well-defined corners, or flat, stable surfaces makes comparison easier.

The important point here is the reliability of the measurements of the check points themselves. If the coordinates of the check points are ambiguous, the comparison results will also be ambiguous, because it becomes impossible to distinguish whether the error stems from the drone results or from measurement errors on the ground. For accuracy verification, it is desirable that the check points be obtained by a method more reliable than the results. At a minimum, they must have sufficient certainty with respect to the ground-side reference values.

Also, checking only the plan (horizontal) position is insufficient. On site, vertical errors have a large impact on practical work, so you need to choose points where height can be verified. In particular, it is important to confirm that elevations match at locations that affect practical decisions, such as the top of slopes, the toes of slopes, graded surfaces, areas around existing structures, and transition zones between cuts and fills. Height errors are difficult to notice visually yet readily affect quantities and construction judgments, so they should be checked more carefully than plan positions.

Moreover, separating control points and check points also serves an on-site training purpose. If operators understand the difference between "points for alignment" and "points for verification," it prevents overconfidence in the results. Because drone surveying relies heavily on automated processing, there is a tendency to be reassured by appearances. However, when you actually compare the numerical values at check points, biases that are not visible to the eye and localized errors due to site conditions become apparent. This helps improve the reproducibility of on-site work.

If you want to ensure accuracy verification, don't be satisfied with only placing control points; always secure independent check points. Simply adhering to this basic practice will make the accuracy assessment of drone surveying considerably reliable.

Step 3: Identify error sources before and after flight

The third step is not just to check errors after the deliverables are produced, but to identify error sources before and after the flight. When people think of accuracy verification, they tend to imagine only comparing the finished point cloud or orthophoto to ground coordinates, but in reality the quality of the results is largely determined in the earlier stages. Rather than inferring causes later from numbers alone, recording error factors at the time of the operation makes diagnosing and improving far easier.

First, what you should check before flight is the site conditions. Strong winds, sudden weather changes, ground surface reflections, bodies of water, abundant shadows, tall vegetation, and the movement of heavy machinery or vehicles—all of these affect the stability of the results. Strong winds make the shooting attitude more prone to disturbance and reduce the quality of overlap. Deep shadows make it difficult to extract feature points, and water surfaces or uniform ground surfaces offer few cues for alignment. Dense vegetation makes it harder to accurately represent the ground surface itself. These conditions provide material for considering “why only this area turned out poorly” after processing.

At the flight planning stage, considerations of flight altitude and overlap rate are also important. Increasing altitude lets you capture a wider area, but the resolution of fine details decreases. Conversely, flying too low is inefficient and makes you more susceptible to terrain undulations and obstacles. If the overlap rate is insufficient, alignment becomes unstable and local distortions or gaps are more likely to occur. Depending on the site, it may be better to take additional photos from auxiliary directions rather than photographing only from a single direction. Especially on slopes, in areas with large elevation differences, and around structures, capturing the shape sufficiently from directly above alone may not be possible.

After a flight, it is important to check the condition of the acquired data as soon as possible. Check whether the number of photos is sufficient, whether there are any missing areas, whether any extremely blurred images are mixed in, and whether there is sufficient overlap on the target surface. If an anomaly is found at this point, it may be possible to capture supplemental images while still on site. If a deficiency is discovered later, a revisit or reflight may be necessary, significantly increasing the workload. Accuracy checks are not just something to be done after processing; they should also include an initial on-site check.

Also, the placement condition of control points and check points needs to be reviewed. Check whether the marks are blending into the surroundings and hard to distinguish, hidden in shadows, placed too close to the edges of the image frame, or at risk of being moved or of sinking.

Even if a control point itself has the correct coordinates, if it cannot be accurately picked up in the images, it will be difficult to reflect it in the results.

In particular, temporarily placed markers or marks on unstable ground may appear fine at the time of installation but can change over time or be affected by work activities.

Additionally, something to be aware of before processing is to make note of locations on site that are prone to producing errors. For example: one side having many trees, shadows falling on slope faces, highly reflective materials in the central area, or temporary structures having moved. With this information, when distortions or errors are found in the results, it becomes easier to judge whether they are due to a simple processing mistake or are caused by site conditions. To improve the precision of accuracy checks, not only the values being compared but also records of working conditions are important.

Practical on-site accuracy verification is not simply reporting "the error was X centimeters." It is important to be able to explain "under what conditions the images were captured, what factors were involved, and that, as a result, the error fell within this range." For that, observations and records before and after the flight are indispensable. Simply carrying out this procedure carefully can greatly improve repeatability at the next site.

Step 4 Compare results data with checkpoints and make a quantitative judgment

The fourth step is to actually compare the result data with the check points and make a numerical judgment. This is the core of accuracy verification. However, the purpose is not to pile up complicated formulas. What really matters on site is understanding how large the differences are in horizontal position and elevation, and interpreting their variability and any bias.

The first thing to do is compare, at each check point, the coordinates in the deliverable with the values measured on the ground. Examine the differences in the horizontal plane and in elevation separately. Here it is important not only how much they deviate on average but also whether there are large deviations at specific locations. Even if the overall result is good, large errors in some areas may make it unusable for certain applications. For example, it may be suitable for materials describing the overall terrain but not for localized construction decisions.

When making a judgment, it is important not to be reassured by the average value alone. Even if the average appears small, positive and negative errors may simply be canceling each other out. For that reason, check the difference at each point individually, and also examine the direction of any bias and the maximum error. If heights across the entire site are uniformly higher or lower, there may be problems with how the reference was taken or with the consistency of the height datum. If the planar offset is large on only one side, distortion at the edges of the imaging area or an uneven distribution of control points may be suspected. By reading these tendencies, you can make a deeper assessment than a simple pass/fail judgment.

Also, you should check not only the numerical agreement at control points but also the consistency of the surface as a whole. Drone surveys can be accurate at individual points yet exhibit undulations or warping across the entire surface. Especially on large sites or sites with significant elevation differences, what may be acceptable when viewed point by point can appear unnatural in cross-section. It is important to cross-check practically important surfaces—such as graded surfaces, paved surfaces, slopes, and areas around structures—and confirm that there are no obvious anomalies.

In height verification, it is effective not only to compare check points point by point but also to assess shape stability using known flat surfaces or linear structures. For example, check whether a surface that should be smooth is undulating, or whether a location that should have no step shows an unnatural difference. Such checks help detect local distortions that numerical tables alone may not reveal. Because people are the ones who use the results on site, shapes that look odd to the eye should be treated cautiously, even if the average error is small.

Furthermore, the verification results should be interpreted as closely as possible in relation to their intended use. If planimetric differences are small but height differences are somewhat large, the results may be suitable for positional verification but may require caution for quantitative purposes. Conversely, even if the vertical direction is stable, large planimetric shifts at the edges may mean the data can be used for terrain checks in the central area but may not be suitable near boundaries or for equipment layout. Accuracy verification is not a binary judgment of good or bad, but a process of determining “what it can be used for.”

At this stage, field personnel should avoid limiting their perspective to merely whether something is a "pass" or "fail." It is important to identify and organize where, in which direction, and by how much differences occur, and what possible causes might explain them. By doing so, you can turn those findings into improvements such as adjusting flight conditions for the next mission, repositioning control points, or adding supplementary imaging. Numeric judgment is not the goal of accuracy verification, but the entry point for improvement.

Step 5 Make the final use-case-specific judgment on whether it can be used on-site

The final step is to make a definitive, use-case-specific judgment—based on the verified figures—about whether they can actually be used in the field. If this is left vague, the effort spent on accuracy checking cannot be applied in operations. The results of accuracy verification are not sufficient if merely left as a table of numbers; they only become meaningful when organized to clarify "what these results may be used for" and "what they should not be used for."

On-site, it is common to try to repurpose a single output for various uses. For example, data originally acquired for progress monitoring may be used as-is for earthwork quantity assessments or as-built verification. However, because required accuracies differ, the same output may not be suitable for all purposes as-is. Therefore, for final decisions it is necessary to separate and organize by intended use.

For example, while it can be used for overall terrain assessment, sharing the construction extent, visualizations for stakeholder explanations, and rough quantity comparisons, it is entirely reasonable to conclude that auxiliary surveying should be used in combination for fine height evaluations and final decisions on installation positions. By distinguishing between appropriate uses and those that require caution in this way, misuse on site can be prevented. To avoid diminishing the value of drone surveying, it is important neither to overestimate nor to underestimate it.

Also, for the final judgment, it is desirable to share not only the results themselves but also the site conditions. For example, adding operational cautions such as "be careful when interpreting the ground surface in areas with dense vegetation," "recheck when using cross sections in areas with strong slope shadows," and "prioritize using the central area since edges tend to be slightly distorted" will make it easier for users to make decisions. The results of accuracy verification should be prepared in a form that can be shared across the entire site, not something understood only by specialist personnel.

Furthermore, documenting improvement points for the next time during the final assessment will raise the quality of operations. If you record this session’s issues—such as too few checkpoints, errors occurring at the edges, insufficient supplementary photography in undulating areas, or significant effects from shadows—you can create an improved plan from the outset for the next time. In this way, drone surveying will develop not as a one-off trial but as a standard on-site procedure.

What matters in practice is not the fact that accuracy checks were performed, but that the way the data will be used has been decided based on their results. Rather than making a simplistic judgment—use everything because the numbers looked good, or discard everything because they were slightly off—organize usability by intended application. That is the final step to deploy drone surveying on-site without undue effort and effectively.

Common Mistakes When Verifying the Accuracy of Drone Surveys

So far we've explained five steps, but in practice there are some common mistakes. Simply being aware of these can significantly improve the quality of accuracy verification.

The first is judging based on appearance. There are cases where people assume accuracy is good simply because the orthophoto looks clean or the point cloud displays smoothly. However, how neat something looks and the correctness of its coordinates are separate matters. Especially with large-area datasets, local differences of several centimeters or more can occur even when the overall view seems fine. You must always cross-check with control/check points and verify the results numerically.

The second issue is not separating control points and check points. If you use the same points that you used to match the results for accuracy checking, you are likely to get favorable results. However, that does not constitute an independent verification. Even if things align in the middle of the site, you may overlook offsets at the edges or on undulating terrain. It is important not to abandon the basic principle of establishing separate check points.

The third issue is that checks in the vertical direction are weak. Horizontal positions are easier to verify because they are easy to see on drawings, while heights tend to be overlooked. However, in practice, errors in height often have a greater impact. Earthwork quantities, as-built shape, drainage, and interfaces with existing structures—many decisions depend on height. Don’t assume everything is fine just because the plan matches; you need to verify height independently.

The fourth is that the causes of errors are not being recorded. If you only record the numbers and do not note factors such as the wind on the day, shadows, vegetation, imaging conditions, or whether supplementary photography was used, you cannot make improvements. If you look only at the results and stop there, you are likely to repeat the same problems next time. Accuracy checks should be regarded not only as an evaluation but also as a record for improvement.

The fifth is treating results uniformly without differentiating by intended use. What may be sufficient for one purpose can be insufficient for another. If you overconfidently assume something can be used for everything, problems will arise in downstream processes. Conversely, if you decide it is unusable because of even a small error, you cannot leverage the efficiency of drone surveying. What is needed in the field is differentiation, not black-and-white thinking.

Many of these failures arise not so much from a lack of specialized technical skill as from an unstructured approach to verification. That is why it is important to standardize the workflow—setting objectives, preparing points, checking error factors, cross-checking numerical values, and making use-specific judgments—as in the five steps introduced here.

Summary

Verifying the accuracy of drone surveying is not something that requires understanding difficult theory. What is needed on site is to first decide why you are measuring, separate the points that support the results from the points used for verification, record flight conditions and site conditions, compare the values at the check points, and finally determine whether the results are usable for each intended purpose. If this workflow is established, you can use drone survey results in a practical way—without overtrusting them and without being unduly skeptical.

What's particularly important is not to treat accuracy verification as cleanup after producing the deliverables. If you assume verification from the planning stage, you can consistently organize the placement of control points and check points, flight methods, the need for supplemental imaging, and how to record on-site data. As a result, you can shift from an operation that simply flies and finishes to a reproducible surveying operation.

On-site, a major attraction of drone surveying is that it enables you to cover a large area in a short time. However, to truly realize that value, it is essential to operate in a way that includes accuracy verification. Rather than focusing on visual appeal, aiming to be able to explain which applications it suits and to what extent is the quickest way to avoid failures in practical work.

And to make acquisition of ground-based references and measurement of checkpoints more reliable and to make drone survey results easier to apply on site, building a foundation for positioning information is also important. For example, at sites where you want to acquire checkpoints and auxiliary points more flexibly, using an iPhone-mounted high-precision GNSS positioning device such as LRTK can make it easier to link drone surveying with ground verification. If you want to establish a workflow of capturing broadly with drones and verifying on the ground, it can be effective to combine these kinds of systems and review the overall on-site surveying operations.