Coded Targets vs. AprilTags in RealityCapture / RealityScan

Photogrammetry relies on accurate scaling to produce correctly dimensioned 3D reconstructions. In this study, we tested two types of scale bars: circular coded-target scale bars and AprilTag scale bars. The objective was to compare their detection speed, usability and scale accuracy.

The test was performed in RealityCapture, now known as RealityScan. Both marker systems were evaluated using the same image set so that the conditions remained as consistent as possible.

Experimental Setup

We captured a surface measuring approximately 1 × 1 metre using:

• Camera: Sony Alpha 7R IV
• Lens: 35 mm prime lens
• Photos: 74 images
• Reconstruction software: RealityCapture / RealityScan

The image set was designed to provide consistent coverage and sufficient detail across the scanned area. Coded-target and AprilTag scale bars were included in the same scene, allowing both systems to be evaluated under identical capture conditions.

A physical one-metre folding ruler was included as an independent reference for evaluating the resulting scale.

3D Reconstruction Workflow

The reconstruction and evaluation followed four main steps:

1. Image alignment: The 74 images were aligned to generate the initial camera poses and sparse point cloud.
   
   
2. Marker detection: RealityCapture’s Detect Markers tool was run separately for each marker type.
   
   
3. Scale definition: The detected targets were used to define the known distances represented by each scale bar.
   
   
4. Scale verification: Measurements from the scaled reconstruction were compared with the physical one-metre folding ruler.

Running separate detection passes was necessary because the marker type is selected before starting automatic marker detection.

Results

Marker detection time:

• Circular coded targets: 174 seconds
• AprilTags: Approximately 10 seconds

In this dataset, AprilTag detection was therefore roughly 17 times faster than coded-target detection.

Scale accuracy:

Both scale-bar systems produced measurements corresponding to approximately 99.2–99.8% accuracy when compared with the one-metre folding ruler. This represents a deviation of approximately 0.2–0.8% from the reference length.

Observations

1. Detection speed
   AprilTags were detected substantially faster, requiring approximately 10 seconds compared with 174 seconds for the circular coded targets. This can provide a meaningful advantage in workflows where processing time or dataset throughput is important.
   
   
2. Scale accuracy
   Both systems produced results within the same accuracy range in this test. No meaningful accuracy advantage was observed for either marker type under the tested conditions.
   
   
3. Marker detection
   AprilTags combine a distinctive square pattern with encoded identification, making them easy for the software to distinguish and identify. RealityScan also describes AprilTags as having a high detection rate, including under more challenging lighting conditions or when photographed from greater distances.
   
   
4. Workflow selection
   Detection speed is not the only consideration when selecting a scale-bar system. Marker size, visibility, mounting surface, environmental conditions and compatibility with an established workflow can also influence the most suitable option.

Additional Insights

• Marker placement: Markers should be placed on flat, clearly visible surfaces and distributed across the area being captured. They should not be covered, distorted or positioned where they are visible in only a small number of images.
  
• Marker image size: Markers must occupy enough pixels in the source photographs for their encoded information to be resolved reliably. RealityScan recommends that photographed markers be at least approximately 100 pixels wide.
  
  
• Viewing angle: More direct camera views generally improve detection. Very shallow viewing angles can reduce the apparent size of a marker and make its pattern more difficult to identify.
  
  
• Image quality: The Sony Alpha 7R IV provided sufficient resolution and detail for reliable marker detection in this setup. Sharp focus, controlled exposure and adequate coverage remain important regardless of the camera used.
  
  
• Test conditions: Detection time can vary depending on image resolution, number of photographs, computer hardware, marker size, lighting and software settings. The recorded timings should therefore be understood as results from this specific dataset rather than universal performance figures.

Conclusion

Both circular coded-target and AprilTag scale bars delivered strong and closely comparable scale accuracy in this test. The primary difference was detection speed. AprilTags were detected in approximately 10 seconds, while the circular coded targets required 174 seconds.

For workflows where rapid processing is important, AprilTag scale bars may offer a clear practical advantage. Circular coded targets remain a reliable option and may still be preferable where they are already integrated into an established capture process.

The results demonstrate that both systems can provide dependable scaling for photogrammetry. The most suitable choice depends on the requirements of the project, including processing speed, marker visibility, environmental conditions and compatibility with the wider capture workflow.