Synopsis

1. The Objectives
The primary objective of this Master's thesis is to systematically evaluate the potential of low-cost, smartphone-based 3D capture technologies to update and maintain existing as-built models originally created using high-precision Terrestrial Laser Scanning (TLS). While TLS provides highly accurate data, it is a resource-intensive and costly method, leading to digital models that quickly become obsolete in dynamic industrial environments where minor changes (like relocating machines or installing new piping) are often undocumented. The study investigates whether the cost-effective and flexible acquisition provided by modern smartphone applications, leveraging built-in LiDAR and Structure-from-Motion (SfM) techniques, can generate point clouds of sufficient quality. Specifically, it focuses on determining if these low-cost point clouds can be reliably registered and integrated with high-precision TLS reference data without falling below the industry standard LOA10 (Level of Accuracy) tolerance of 0.075 m RMS.

2. The Options
The research compares six distinct mobile applications, tested on an iPhone 12 Pro, representing both active LiDAR and passive SfM methodologies. The LiDAR-based applications include Dot3D, Scaniverse, and SiteScape. These apps use the phone's LiDAR sensor for direct depth measurement. The SfM-based applications, which reconstruct geometry from overlapping photos, include PIX4Dcatch (which uses LiDAR only for camera stabilization), RealityScan, and the photo-based SfM mode of 3D Scanner App™. For comparative purposes in the field test, the thesis also evaluates the LiDAR mode of 3D Scanner App™. Each app's performance is gauged through a rigorous, two-stage evaluation process: a controlled laboratory test focusing on objective metrics (such as noise, dimensional accuracy, and density) and a practical application test conducted in a complex industrial setting (GSI Darmstadt).

3. The Results
This comprehensive analysis confirms that while all tested LiDAR solutions meet the LOA10 minimum standard, Dot3D is the unambiguously preferred solution. Dot3D consistently delivers the best performance due to its professional-grade data handling and superior output quality. It excels by achieving the highest point density, recording over $6.8 million P/m² in the field test (with a mean point spacing of 0.38 mm), which is significantly denser than all other apps. This high density is crucial for modeling fine structures and quantitative analysis. Furthermore, Dot3D demonstrates the highest dimensional accuracy and stability: its point cloud remains metrically correct from the beginning, exhibiting a near-zero scale deviation and the lowest RMS value in the ICP registration. Unlike SfM methods, which often suffer from scaling errors and local distortions despite retrospective correction, Dot3D's LiDAR-based approach is robust, providing highly homogeneous, artifact-free visual quality that seamlessly fuses with the high-precision TLS environment. Click here to download the complete study in its original form (German).