How many GCPs are needed for surveying? Placement guidelines and five items for accuracy verification

The number of GCPs for surveying affects the ease of work when few, but if too few the accuracy of deliverables becomes unstable. Conversely, arbitrarily increasing them yields diminishing returns relative to the effort and can reduce overall site efficiency. Therefore, in practice it is important not to fix a set number of points, but to decide based on the area extent, shape bias, elevation differences, required accuracy, and the methods of imaging and measurement.

Especially for tasks that create point clouds or orthoimages from photographs, the number and placement of GCPs greatly influence the stability of results. It may seem sufficient to place them at the four corners, but in reality issues such as weakness in the central area, distortion in elongated ranges, inadequate reflection of elevation differences, and error amplification at the edges often occur and cannot be addressed by a simple arrangement. Also, if you do not separate the points used as GCPs from independent check points reserved for accuracy verification, you cannot judge whether the accuracy is truly achieved.

This article organizes the way of thinking about the number of GCPs required for surveying, then clearly explains placement guidelines, practical decision criteria for the number of points, and five items for checking the accuracy of deliverables. It is intended to help personnel who will install GCPs on site plan placements that avoid rework due to insufficient points while eliminating waste.

Table of contents

• Why the number of GCPs matters in surveying

• Basic concepts for the number of GCPs needed in surveying

• Consider placement from three directions: perimeter, center, and elevation differences

• Failure examples of too many or too few GCPs

• Five items for accuracy verification

• Practical workflow for deciding point count and placement on site

• Summary

Why the number of GCPs matters in surveying

The number of points in GCP surveying is not simply about how many markers are placed on site. It carries the meaning of how many sufficient constraints can be provided to assign coordinates to deliverables, suppress shape distortion, and ensure positional reliability. In other words, the number is not simply “the more the better”; what matters is whether points are placed where they are needed and in sufficient quantity.

GCPs have two major roles. One is to align the generated point cloud, orthoimage, or 3D model to known coordinates. The other is to suppress deformation, tilt, and scaling of the entire model. If these two roles are not fully fulfilled, the results may visually appear aligned yet retain non-negligible errors in distances, areas, or positional relationships.

Commonly encountered on site are situations where results look fine in processing software but edges are shifted, the center undulates, or vertical errors are large. Such problems often arise not only from the performance of observation equipment, but also from insufficient numbers of GCPs or biased placement. Especially when the target area is elongated, has undulating terrain, or contains complex structures, a corner-and-center arrangement can be weak.

If all GCPs are clustered in similar locations, local accuracy may look good but errors expand in distant areas. This is akin to locking down only a limited area of the model, which is insufficient to stabilize the whole. Therefore, when considering the number of GCPs, it is essential not only to look at the minimum number but also to ensure the entire target area is uniformly supported.

Equally important is the concept of separating GCPs and check points. If all known points are used for alignment, you can only tell how well the result fits those points, but you cannot confirm whether the model truly achieves accuracy at other unused locations. In practice, securing the necessary number of GCPs while leaving independent points for accuracy verification is what enables a trustworthy evaluation. In other words, planning the number of required points must include points for alignment and for verification.

Basic concepts for the number of GCPs needed in surveying

There is no absolute correct number of GCPs that applies to every site. However, there is a practical way of thinking that is easy to use in the field: consider the practical minimum number that stabilizes the deliverable rather than the theoretical minimum. In theory, coordinate alignment can be achieved with few points, but to obtain stable results you need an arrangement with some margin.

First, it is a premise that constraints must be balanced both planimetrically and vertically. For simple, regularly-shaped flat ground, basic points to hold the perimeter and supplement the center are sufficient. From a practical standpoint, even for small and relatively simple areas, it is safer to include multiple points that not only cover the perimeter but also include the center or midpoints of long sides. In addition, securing separate verification points makes post-processing accuracy checks easier.

On the other hand, when the target is elongated, the required number of points tends to increase even if the area is the same. For road-like sites, along irrigation channels, linear reclamation areas, or long slopes, holding only the ends and the center weakens the constraints in intermediate sections and makes it easy to miss local distortions. In such sites it is necessary to add points at shape change points, bends, and mid-sections of long distances. Judging by area alone tends to lead to too few points, so always check aspect ratio bias.

The same applies to sites with large elevation differences. If you place GCPs only within the same elevation band in areas with slopes, steps, mixed fill and cut, or where the subject extends above and below structures, planimetric uniform placement alone is insufficient. Unless you cover both high and low positions, vertical reproducibility becomes unstable. In works that mainly use nadir photography, people tend to focus on horizontal accuracy, but in practical use vertical errors can often be problematic, so design the number of points to reflect elevation differences.

Another easily overlooked factor when considering the number of required points is visibility. If installed points cannot be clearly identified in images or point clouds, they may be sufficient in number on paper but unusable in practice. If they are buried in grass, in shadow, blend into the background, or lose shape when viewed obliquely, the number of usable points will be lower than expected. Therefore, allow some margin for points that may become unusable on site compared to the number planned on the desk.

In general, smaller, regularly-shaped flat areas tend to require fewer points, whereas elongated sites, sites with large undulation, sites with many occlusions, or sites requiring high accuracy should be approached with increasing numbers of points. The required number cannot be determined by area alone; shape, topography, imaging conditions, and intended use must be judged together. This approach helps avoid both the waste of placing too many points and the need to redo work due to too few.

Consider placement from three directions: perimeter, center, and elevation differences

When deciding GCP placement, think first to secure the perimeter, then to supplement weakness in the center, and finally to pick up areas with elevation differences—these three directions make it less likely to fail in practice. Focusing only on the count tends to concentrate placements in biased locations; but if placement quality is poor, increasing the count does not stabilize accuracy.

First secure the perimeter. Corners and edges of the target area are prone to model deformation. Especially if one side of the perimeter has few points, that side retains degrees of freedom, causing stretching or twisting. Therefore, it is important to evenly encircle the outline of the area. Note that it is not enough to simply place points at the four corners. If the corners are extremely far apart or a long side exists, auxiliary points along that side are necessary.

Next consider the center. Relying only on the corners can leave the center weak. In flat sites, a solid perimeter may appear sufficient, but slight undulations or rises and falls in the center are easily overlooked. Thus placing a representative point at or near the center improves overall model stability. For large areas, using multiple intermediate points to even out the interior is effective rather than a single central point.

The third consideration is elevation differences. If there are steps or slopes and all GCPs are placed in the same elevation band, vertical constraints will be lacking. As a result, the planimetry may match but vertical position may be offset. For sites with clear elevation differences, place points both at high and low locations, and preferably also at slope inflection points or terrain transitions to improve stability. Planimetric and vertical placements should not be treated separately but combined to satisfy both simultaneously.

Moreover, on elongated sites you must pay attention to spacing in the distance direction in addition to perimeter, center, and elevation differences. If points are spaced too far apart, the interval becomes weak. For linear sites, supplement not only the endpoints but also intermediate sections, curves, and sections where width changes to stabilize results. Practically, it is more important to imagine where degrees of freedom might remain and place points accordingly than to treat the whole site as a single surface.

Also, selection of installation locations directly affects accuracy. Avoid pavement joints, movable covers, muddy areas, places that will be covered by vegetation, and locations with strong shadows. Choose locations that are easy to find, stable during observation, and easy to identify during processing. In short, good GCP placement is not only evenly distributed on a map but also resilient for actual observation and image interpretation.

In practice, it is easy to first envision a perimeter-holding placement, then fill the central gaps, and finally add points according to elevation differences and shape changes. Thinking in this order clarifies the reason for increasing the number of points and helps avoid arbitrary additions.

Failure examples of too many or too few GCPs

Too few GCPs is problematic, but too many also creates different problems. When too few, poor accuracy is easy to see; when too many, it is often thought only the workload increases, but in fact it can lead to management errors and operational disorder that reduce quality. The important thing is to keep the necessary points within a range that can be reliably managed.

A typical example of too few is relying only on the four corners. For a small site this may seem acceptable, but central area accuracy verification becomes weak and errors can appear in areas far from the edges. Also, on elongated sites holding only both ends makes twisting or scale irregularities likely in the middle. Placing GCPs only within the same elevation band on undulating terrain is another classic case causing unstable vertical reproduction.

Another failure is using all points in processing and leaving no check points. In this case, results are fitted to the known points and may appear well aligned. However, without independent verification you cannot evaluate accuracy externally. When explaining deliverables in practice, saying they fit internally is insufficient; it is important to show how well they agree at unused points.

On the other hand, having too many points can lead to variable quality in installation and observation. Increasing the number too much increases the effort for marking, observation, recording, photo organization, and matching. This leads to unevenness where some points are carefully observed while others have poor visibility, ambiguous records, or missing coordinate notes. A large number of points is not inherently bad, but if it exceeds manageable levels quality deteriorates.

Placing many points clustered in similar locations is also inefficient. For example, concentrating many points on one side does not solve weakness on the opposite side or in the center. What matters is distribution balance, not density. When adding points, be clear about which degree of freedom each addition is intended to constrain. Additions without reasons increase workload rather than improving accuracy.

To prevent failures, start with a hypothesis of the required number of points and make minor adjustments on site based on visibility and placement balance. Also separate and manage GCPs and check points, and rather than demanding identical quality for every point, ensure each role is reliably executed. The practical correct approach is to avoid too few and too many and choose a number you can explain the purpose of.

Five items for accuracy verification

Even with appropriate GCP placement, deliverable reliability cannot be guaranteed without accuracy verification. Here are five minimum checks to keep in mind in field practice. The important thing is not to judge by appearance alone on the processing screen but to look in order at coordinates, distribution, vertical dimension, independent verification, and suitability for use.

The first is observation quality of the GCPs themselves. No matter how good the placement, if the original coordinates are unstable the deliverable will not be stable. During observation check for basic issues such as coordinate system confusion, mixing of elevation datums, transcription errors, instability of fixed solutions, and poor observation environments. Especially when multiple people work on site, failure to standardize point naming and record formats often causes mix-ups later. Accuracy verification should be considered to begin at the observation stage, not only after processing.

The second is the distribution of residuals and offsets. Rather than looking only at individual point numbers, check where large offsets occur. If residuals are large only at a particular edge, biased placement or insufficient imaging conditions may be suspected. If only the center is disturbed, the configuration may be overly perimeter-weighted. If errors are large only at high locations, reassess elevation constraint insufficiency or vertical observation conditions. In other words, not only the magnitude but also the spatial pattern of errors is important.

The third is checking the vertical dimension. It is not uncommon for planimetric positions to match well while height is poor. Particularly in undulating sites, slopes, fills, and around structures, vertical errors can hinder practical use. For height checks, avoid being distracted by planimetric distribution and verify consistency at locations of differing elevation. If low areas match but high areas shift, or vice versa, reconsider placement or observation conditions.

The fourth is evaluation using independent check points. This is extremely important. Small errors at GCPs used in processing are, in a way, expected. What you really need to know is whether unused points also align. Spread check points not only on the perimeter but also in the center and at varying elevations to assess the model’s overall reliability. If check point errors are large in some areas, insufficiency in imaging or placement at those locations is suspected. Omitting independent verification invites mistaken judgments based on apparent goodness of processing results.

The fifth is whether the accuracy is sufficient for the intended use. The purpose of accuracy verification is not to minimize numbers for their own sake. What matters is what the deliverable will be used for. Required accuracy differs if it is for general overview, explaining positional relationships, or for cross-section and as-built evaluation. Setting excessively strict targets increases workload, while insufficient accuracy for the intended use renders the deliverable unusable. Therefore, judge accuracy not only by absolute values but also in light of the intended use.

These five items are interconnected rather than separate checks. Poor observation quality tends to increase residual scatter; insufficient vertical checking leads to unusable deliverables; without independent verification you cannot judge overall reliability. In practice, don’t decide pass/fail by a single number—check these five together to more easily isolate causes and plan improvements.

Practical workflow for deciding point count and placement on site

Deciding GCP count and placement should not be done by feel after arriving on site; it stabilizes if you create a provisional plan in advance and adjust on site. Practically, first understand the shape of the target area, then confirm the required accuracy, and based on that create a provisional placement plan from the three directions of perimeter, center, and elevation differences.

In the initial stage, confirm the outline and elevation differences of the target area from drawings, current photos, and existing map information. At this stage, determine whether the area is regular or irregular, elongated, undulating, or heavily occluded, as that indicates the direction for required point counts. For example, regular flat areas can start from a basic arrangement, while linear or highly variable elevation sites should initially assume additional points.

Next confirm the deliverable’s intended use. Even at the same site, whether the purpose is rough positional confirmation or dimensional/cross-section usage changes the thinking on point count. The stricter the purpose, the more you must reserve check points and ensure unbiased placement. Deciding point count without clarifying this leads to later discoveries of insufficient accuracy.

On site, check whether the provisional plan can be implemented. Even if ideal on paper, vegetation, traffic, work flow, shadows, and access restrictions may prevent placement. In such cases, do not simply move points nearby; rethink the overall balance of the arrangement. Especially if one corner of the perimeter cannot be obtained, do not compensate by merely adding points on the opposite side—instead consider whether you can constrain that degree of freedom from another position.

After observation, organize point names, coordinates, installation photos, and role classification. Clearly separating GCPs and check points here prevents confusion during processing. Also note any concerns observed on site, such as low visibility points, points likely to blend into the background, or nearby reflections or occlusions—these notes help interpret outliers later. Causes of poor accuracy are not always in the processing stage; preserving site information is important.

Thus, deciding point count and placement is a design exercise based on site conditions and intended use, not a simple per-area point calculation. With experience judgment becomes quicker, but the basics remain the same: survey the target area, confirm required accuracy, assemble perimeter/center/elevation placements, reserve check points, and manage observation quality. Rigorously following this flow is the fastest path to stable GCP surveying.

Summary

The number of GCPs required for surveying is neither “the fewer the better” nor “the more the safer.” What matters is to balance perimeter, center, and elevation differences according to the target area’s shape, elevation variation, required accuracy, and observation conditions. Regular flat sites can often work with relatively few points, while elongated or undulating sites require additional points to cover intermediate sections and elevation transition points.

Equally important as placement is leaving independent check points to verify accuracy. Evaluating results only with points used as GCPs can be misleading; apparent goodness may not mean the deliverable is truly usable. By checking observation quality, residual distribution, vertical consistency, independent point verification, and suitability for use—the five perspectives—you can determine the optimal point design for each site.

When in doubt in practice, don’t try to force a small number of points; instead choose a placement you can explain the purpose of. If you want to streamline everything from on-site observation to record organization and coordinate management, using smartphone-mounted high-precision GNSS positioning devices such as LRTK can be effective. They allow you to confirm GCP installation positions on the spot and reduce missing records or mix-ups, making it easier to implement point count and placement plans in actual operations. If you want to stabilize GCP surveying quality, reviewing not only how to determine the required number of points but also ensuring a reliable on-site positioning system is ultimately the most effective approach.