Visual Alignment

Introduction

Visual Alignment is a feature that makes it easier for you to scan places with empty or repetitive architecture.

Examples of such places include:

  • Open areas such as empty factories, empty storerooms, empty convention centers, ballrooms for dancing, etc
  • Long uniform hallways such as in schools, hotels, and rental storage units
  • Areas with uniform and repeating architecture, such as very evenly spaced columns in old-school factories
  • Architecture, engineering, and construction models
  • Commercial real estate spaces that are not built-out or without furniture

Previously these places were very difficult to capture — the lack of 3D data or the fact that a new scan was too similar to a previous scan often led to misalignments and alignment errors.

empty-nola-factory-medium.jpg

 

How it Works

To use Visual Alignment, you first print out markers such as this one below and tape them to key locations on-site.

marker-19.png

While you are scanning, the Matterport Capture app looks at your 2D panorama images and recognizes these markers. The Capture app then uses visual information such as the size and location of the marker to help place the scan in roughly the correct location, and then uses depth information for a fine-tune alignment.

Markers will appear in the final virtual tour -- they will not be edited out. If you are capturing mainly for 3D data such as the point cloud, then this is not an issue. If you care strongly about the visual quality of the virtual tour, we suggest you continue to use appropriate 3D objects such as plants, chairs, boxes, etc that blend in to the space yet still give the camera something to latch onto. You can also create your own branded markers.

Visual Alignment with Matterport is based on April Tags from the APRIL Lab at the University of Michigan. Special thanks to Professor Olson and his team.

 

Download

First, download the set of markers based on your paper size:

Print them at normal size (do not expand to poster-size) and take them with you on-site.

We’ve included a set of 100 tags. However, typically you do not need to use all 100. The exact amount you use is up to you, as it depends on the size of the space and its unique characteristics. Take more markers than you think are necessary, since you cannot reuse markers within a single Matterport model.

If you plan to use these markers a lot, we suggest you laminate them so they last longer.

These scan markers are based on AprilTags by the APRIL Lab at the University of Michigan. Visual markers from other systems or companies will not work with Matterport.

 

Place the Markers

When you arrive on-site, place the markers as part of your preparation to scan.

Place the markers at about eye-level height and about 15 ft (4.5 m) apart. Exact distance between markers is not necessary. For reference, 15 ft is about the length of a typical car (four-door passenger sedan).

 with_apriltags_instead_of_objects.png

Apply the markers to a flat surface such as a wall or a column and tape on all four corners so it is flush with the surface.

Markers that are not completely flush with the surface of the wall or column (curling up or flapping about) may not be recognized or will be incorrectly recognized as 3D data. This can interfere with the point cloud or 3D mesh you wish to export later on.

Place markers on the wall so the camera can see them 'head-on'. You can place markers on the floor or ceiling, for example if the hallway is particularly narrow. However, in general a marker on the floor is difficult to recognize because it might show at an oblique angle or in the camera's blind spot.

Finally, once you place the markers, do not move them. For example, suppose you only printed out 10 markers. If after using all ten you pick up marker #1 and move it to a new spot, then this can confuse the Capture app. Instead, print out new markers #11 to #50 and place them in the new locations.

 

Scanning

Open the Capture app and create a new model like normal.

Read a 10 step walkthrough for scanning for beginners. Learn our normal best practices for scanning.

Go into Settings and enable Visual Alignment.

Visual Alignment can add two to three seconds to the time it takes to align a scan. If you will not use alignment markers in your next model, turn it off so you can scan a little faster.

Connect to the camera and scan like normal. Visual Alignment works for both the Pro and the Pro2 camera. Each scan position should be about 15 ft (4.5 m) apart and have a line-of-sight to a previous scan and a marker.

 with_apriltags_instead_of_objects.png

While your markers will not explicitly appear on the minimap, you can tap on the scan number and then tap on Preview to visually identify the marker.

first-scan-preview-button.png

When you are done, just upload your model for processing like normal. When processing is complete, open my.matterport.com and find your 3D Space. From here, you can view and share as normal.

Have more questions? Submit a request