Effects of Working with Camera and Project Instability

Project Instability is when something changes between photos that PhotoModeler expects not to. PhotoModeler expects the camera and marked points not to change between photos.

When a camera’s internal parts or your project targets move between photos, PhotoModeler can no longer line everything up perfectly. This can reduce measurement accuracy or cause some photos to fail. The most important step is to keep your camera settings and your targets as fixed as possible during the whole photo session.

Summary

  • Instability in the camera (zoom changes, image stabilization) or in the scene (moving targets) can reduce accuracy or cause PhotoModeler to fail to solve some photos.
  • Other photograph quality factors such as blur (focus,  motion), sharpening, reflections, and fringing can make targets look different photo to photo leading to instability.
  • Improve stability by using a non‑zoom lens or locking the zoom so it cannot move and avoiding or disabling image stabilization when possible.
  • If using targets, make sure they cannot move between photos (for example, tape paper targets at multiple corners).
  • Improve image quality with lighting and turning camera features off.

 

What is Instability?

Instability means something that should stay the same between photos is changing without you realizing it. PhotoModeler assumes your camera’s internal settings stay constant between photos. These include the zoom level (focal length), where the image sits on the sensor (principal point), and how the lens bends light (lens distortion). When these change unexpectedly, the math no longer matches reality, and accuracy drops or some photos cannot be solved. You can think of this like measuring with a tape measure that is stretching and shrinking slightly as you use it.

Cameras with Image Stabilization (IS/OIS) add to the internal instability and that can be confusing. IS/OIS makes photos and videos look steadier to your eye by moving lens parts or the sensor, but for PhotoModeler those hidden internal movements mean the camera is no longer a perfectly repeatable measuring device.

For projects with Coded Targets, another possible unexpected instability is when the targets move (even a small amount) between photos. Photograph quality can cause targets to look different between photos as well. Camera sharpening, fringing, blur, and reflections can affect how the dots are marked on each photo and shift their center.

This instability can be subtle, and the camera or targets could be changing even when you do not think they are.

How Instability Manifests

Instability usually shows up as high Total Error, photos that will not orient, or obviously wrong 3D results. Total Error is a single number that describes how well all the data fits together. The higher this number, the worse the fit.

Here are ranges of Total Error values:

  • 1.0 or less: Very good, suitable for high-accuracy industrial work.
  • Up to 2.0: Good for most projects.
  • Over 4.0: There is likely an instability you should investigate.​
  • Over 9.0: Serious concern – expect poor results or failed photos.

The Total Error is shown in several places in PhotoModeler:

  • In a manual project, the final Total Error is shown in the Total Error bar graph and then in the Processing Finished dialog.
  • In an automated project, the Total Error may be shown at the end of processing in the Automated Coded Target Project dialog if the error value is too high.
  • For all projects, Total Error can be seen in the Project Status Report / Information from most recent processing / Total Error / Last Error.

Other signs of instability include:

  • Error indicators in reports (high maximum point residual, poor scale fits, lens distortion solving to 0.0 during Automated Calibration).
  • Photos that refuse to orient (red X icons) even though they share enough points.
  • Photos that do orient but look obviously wrong in the 3D Viewer, leading to large errors in 3D measurements.

The end result of instability is that your project data could be less accurate, or you may have photos not orient that you need in your modeling.

 

How to Improve Stability

Use the following steps in order – start with the easiest changes before considering new equipment.

Camera Changes

  • Use a non‑zoom (prime) lens if possible.
  • If you must use a zoom, lock the barrel with strong elastic bands so the zoom cannot move, and avoid bumping the zoom ring.
  • Avoid cameras or lenses with Image Stabilization (IS/OIS) for high‑accuracy work; if your camera or lens has IS, turn it off where possible.
  • Be especially cautious with mobile phone cameras – their stabilization is usually always-on and cannot be fully disabled.
  • Improve image quality by having good lighting (reduces blur), and reduce or turn off image sharpening in the camera settings.
  • If you consistently see high Total Error and you are already following these steps, consider using a camera with a prime (non‑zoom) lens and no image stabilization for your PhotoModeler projects.

Project / Target Changes

  • Tape paper targets securely with at least two pieces of tape at different corners.
  • Rubber-backed targets should be placed in secure locations (usually do not need tape on flat surfaces).
  • Avoid placing targets where wind, vibration, or walking traffic can move them.
  • For surfaces that can vibrate (equipment, machines), avoid using them as reference points if you expect motion between shots.
  • Ensure the surface of any targets is very matte to avoid specular reflections of the light source that move between photos.

Further Background on Instability (optional read)

The following sections provide extra technical detail for users who need very high accuracy or want to understand the underlying photogrammetry.

Camera Instability

Instability in a camera is usually caused by one of these:

  • Zoom lens that allow for small inadvertent changes in focal length, or
  • Lens or camera body that have IS/OIS (Image Stabilization).

This can be confusing because ‘Image Stabilization’ sounds like something we would want, but in photogrammetry it can introduce internal movement. These IS systems are typically used to reduce motion during video and sometimes reduce motion blur with stills. These IS systems use one of three methods to move the image around in the camera so that the video seems more stable to the viewer: moving the image sensor around, moving elements in the lens, or with digital-post-processing, shifting the image. In photogrammetry we want the internals of the camera not to move – the internals (lens elements, and imaging area) should be stable as much as possible.

Even when a camera has a way to turn off IS, if the IS is done with floating lens elements or floating imaging chip, then there is inherent instability (shake the camera hard enough and these elements will move) even if the IS mode is ‘off’. Ideally one uses a camera and lens with no IS, and if it does have IS, turn it off (although not perfect). The effect of IS is to change the principal point in the geometry (this is where the central axes of the lens intersects the imaging area/chip). For standard projects, PhotoModeler assumes the principal point is constant across all photos in a project. With self-calibration this point can be allowed to vary across photographs, but this is generally not a good solution as it adds some other mathematical instability to the solution – this is not PhotoModeler’s default action.

Zoom lenses have floating lens elements that can move in the lens body. Slight changes in zoom (focal length) can affect the internal stability of the camera. Even if you are not zooming the lens on purpose. A lens/camera without zoom is best, if not, second best is trying to stabilize the lens so the zoom does not change.

High Accuracy Industrial Measurement

Most PhotoModeler users do not need to worry about the following – they apply only to very high accuracy (around 1 part in 20,000 and better). You would normally not need to worry about these in bespoke fabrication tasks such as marine decks, marine canvas, 2D template digitizing, and similar. If you are working in another field and aiming for very high accuracy (using a high-resolution camera with high-end lens) you might also consider these instabilities:

  • change in focus, and
  • unstable lens mount for removable lenses.

The latter is harder to correct – other than buying the correct camera. Changes in focus can sometimes be managed the same way zoom is handled in a lens with a barrel (use rubber bands). Also turn off auto-focus (after your scene has good focus and depth of field). The more you can lock the camera down so nothing changes internally during photography, the less the stability issue will negatively impact your accuracy. Note for most projects without extreme accuracy demands, it is better to leave auto-focus on.

Project Instability

There are a few ways that photogrammetry projects can be unstable outside of the camera instability issue. With ‘instability’ we are looking for things that can change that we do not expect or want to change between photos. The most common of these would be when the camera internal parameters change. And a secondary cause, when using dot or coded targets, is they somehow move between photographs (even slightly). The movement can be small too (a target sheet of paper not taped down that is buffeted by the wind). Ensure any points being used in PhotoModeler that one expects to be stationary, are actually stationary across all photos (even small movements).