Minimizing distortions within panorama

Issue description

When evaluating the quality of panoramas, distortions of objects captured can be seen occasionally. This article describes the reasoning for this issue as well as best practices to manage it.

Pre-requisites

Ensure that your NavVis System Software in your NavVis device and the Processing Software are updated regularly to benefit from all improvements to the User Interface and algorithms of NavVis software products.

Analysis of Distortions

Object distortions are a well-known panorama rendering artifact for setups similar to how NavVis implemented Mobile Mapping. They are influenced by the camera placement and the algorithm used for panorama stitching and are, to a certain extent, inevitable when panoramas are constructed from images of multiple separate cameras (not 360° panoramic cameras).

They can appear in objects of different shapes and at varying distances. Nevertheless, they only appear occasionally, when the so called stitching seam of a panorama accidentally crosses a singular object, not generally within panoramas captured from NavVis devices.

When does distortion occur? 

Distortion most often occurs when:

  • Objects are close to the device head and stretch across the Field of View (FoV) of two or more cameras.

  • Objects contain thin structures like beams, frames, pipes etc. in front of “empty” space.

  • Objects appear close to each other within the panoramas, but within the 3D space have a strong. difference in actual distance to the cameras (corridor vs. door frame, outer space vs. window frame).

  • There is no laser information available due to the distance of an object expanding beyond the laser range.

More information on image stitching and potential issues can also be found on Wikipedia. The image below demonstrates particularly well how the seams for image stitching look from a technical point of view. The issues described show up in images, whenever this seam crosses a single object, builds the panorama out of several sections from differing images while at the same time the depth information within this area is not uniform.

This kind of stitching artefact can also be seen from other providers like Google Street View.

Effect on overall data quality

Overall, the effect on data quality as a whole is limited. Singular panoramas are affected, while the Point cloud structure and point cloud coloring in itself will not be affected. However, the number of distortions seen can vary based on the characteristics of the environment captured.

Best Practices

At this point in time there is no manual fix available, which can enhance results from the automatic, algorithmic creation of these panoramas. However, there are several best practices that can be used during scanning and data preparation. These best practices can minimize the impact of these distortions so that your can scanning projects and the corresponding data generated will be successful.

1 Scanning

Implement the following to limit the impact of distortions within panoramas:

  • Higher frequency of panorama triggering. When capturing a very small number of panoramas per area or room, the issue is that there are no adjacent panoramas which can help us to understand the real appearance of the environment captured.

  • Keeping an optimal distance from objects. Whenever possible keep a distance of more than one meter from objects, which should be captured within panoramas.

  • Balancing distance of surrounding objects. When walking down corridors or other narrow spaces, keep a balanced distance from all sides.

  • Facing important objects with one camera straight. When there are structures that are important to project and must be reflected with absolute accuracy, we recommended facing one of the cameras' FoVs directly at this object, with a good distance to eliminate the need of image stitching within this area.

2 Hiding panoramas in IVION

Panoramas, which do not meet the quality expectations required for a specific project after processing, can be hidden within IVION, refer to Configuring a Site for further information. 


FAQ

What is the main issue described in the article regarding panoramas?

The main issue is the occasional distortion of objects captured in panoramas, which can occur due to the camera placement and the stitching algorithm used.

How can I ensure my NavVis software is up to date?

You should regularly update the NavVis System Software on your NavVis device and the Processing Software to benefit from improvements.

When do distortions in panoramas typically occur?

Distortions most often occur when objects are close to the device head, contain thin structures, are close to each other in the panorama but far apart in 3D space, or when there is no laser information available.

Does distortion affect the overall data quality of the point cloud?

No, the overall data quality is limited; while singular panoramas may be affected, the point cloud structure and coloring remain unaffected.

What are some best practices to minimize distortion during scanning?

Best practices include increasing the frequency of panorama triggering, maintaining a distance of more than one meter from objects, balancing distances in narrow spaces, and facing important objects directly with one camera.