Indoor 3D scanning with PIX4Dcatch

Scanning large spaces with a smartphone is a challenging task, as errors in determining the camera’s position and orientation (known as “drift”) accumulate during the capture process. This leads to inaccuracies in the final 3D model: distortions, duplicated surfaces, or geometry deformations appear, making the data unsuitable for professional use. 

Therefore, our Swiss partner Pix4D conducted research using an iPhone and the PIX4Dcatch solution, which combines the GeoFusion algorithm and Autotags technology to ensure absolute accuracy during scanning. 

Research tools

1. Autotags is an effective method for correcting drift. When the Tag Detection feature is enabled, PIX4Dcatch automatically recognises the tags during capture and uses them together with tracking data to compensate for accumulated errors immediately after the scan is complete.
2. After data collection, the GeoFusion algorithm analyses all images and calculates the precise position and angle of the camera for each frame, ensuring high spatial accuracy of the resulting model.

As a result, specialists can obtain accurate 3D models of large spaces directly during data acquisition — environments where traditional scanning methods lose their effectiveness.

iPhone with PIX4Dcatch.The phone’s acquired LiDAR data can be overlaid on the current image, providing real-time feedback on the completeness of acquisition. Autotags are detected in real-time (indicated in blue) 

Research technology

1. Place the Autotags in the scanning area.
2. Scan the space with PIX4Dcatch to cover the area and collect images and LiDAR data. The software automatically recognises Autotags in real time.
3. Automatic processing. Once data collection is complete, the GeoFusion algorithm runs on your smartphone and performs drift correction and camera position optimisation within seconds.
4. Merging scans (for large projects). In large-scale scenarios, you can merge multiple datasets to obtain a complete 3D model of the entire space.

Practical application

Comparative test of the GeoFusion algorithm with and without the Autotags feature.

Two smartphones with identical settings and PIX4Dcatch enabled were used for this test. The Tag Detection feature was activated on one device and disabled on the other. In addition, the second smartphone relied on the app’s minimalist mode, which uses real-time video streaming and smartphone sensors (GNSS, compass) — a technology similar to Simultaneous Localisation and Mapping (SLAM).

After walking indoors along a closed trajectory, the results showed a significant difference. The smartphone with Tag Detection enabled demonstrated a much lower average error (7 pixels), while the smartphone without Tag Detection had an error of 41 pixels.

This indicates that the GeoFusion method tracked the camera position far more accurately and effectively reduced drift. This was especially noticeable at the beginning of the test, when drift is usually most pronounced. 

Combining multiple scans to obtain an accurate 3D map of an office space.

In such scenarios, several separate scans can be collected and merged during processing to produce an accurate 3D map. To ensure a single, precise result, specialists placed fixed reference targets throughout the scanning area using the Autotags function.

Screenshots from PIX4Dmatic show the 2000 m2 office space from a top view across five different scans. In each image, the Autotags are marked in green and the iPhone LiDAR point cloud, which is colored with the images 

This solution enables PIX4Dmatic to automatically align and merge multiple projects, correctly combining all independent scans into a single large-scale 3D model. As a result, the final test produced a detailed 3D reconstruction of a 2,000 m² office space, created from 5,000 photos and LiDAR scans captured with an iPhone. 

Screenshots from PIX4Dmatic of the merged datasets from the image above are shown. One can observe the Autotags in green, the 5000 camera positions as small green dots, the iPhone LiDAR point cloud colored with the images, and the dense photogrammetry point cloud 

Scanning the interior and exterior of a private house to obtain a single 3D model.

For this scenario, data from both the exterior of the house and its different floors were combined. As a result, the complete model was created from several PIX4Dcatch projects, each processed separately and then merged in PIX4Dmatic. 

Screenshots from PIX4Dmatic of a house from different PIX4Dcatch scans that have been merged with Autotags, which facilitate the modeling for both indoor and outdoor environments 

To obtain accurate 3D models, specialists have long had to choose between two solutions: expensive specialised laser scanners that offer high accuracy, or simple mobile applications that work quickly and inexpensively but produce imprecise results. 

Pix4D has changed this situation by combining an accessible solution — a smartphone — with software that delivers professional-level accuracy. This has made it possible to capture high-quality 3D data even in challenging working conditions. 

Futurology is the official distributor of Pix4D software. The brands' products are available through the Futurology dealer network in the United States of America. For more information, please contact info@futurology.tech.