Revopoint MetroX Review – Good and affordable 3D scanner for the right target group

1 - Introduction, unboxing and features 2 - The principle of the structured light method 3 - The functional principle of point clouds 4 - Triangulation of the point clouds in a mesh 5 - 3D for Dummies: One-Click-Solution 6 - Limits of the system and the effective accuracy 7 - Summary and conclusion 8 - Appendix: Datasheet 9 - Appendix: Quick Start Quide

What is a point cloud?

A point cloud is a three-dimensional data structure consisting of a large number of individual points, where each point is described by its coordinates in space – typically in a Cartesian coordinate system (x, y, z). In metrology, especially in 3D scans using structured light, such a point cloud is created by the projection-supported reconstruction of the surface of a real object. The point cloud is therefore not a continuous surface, but a discrete scan that approximates the geometry of the surface.

Generation begins with the projection of structured light patterns, as already described. Triangulation is used to calculate a 3D point for each pixel on which the pattern was recognized. These points are clearly positioned in three-dimensional space and their density depends, among other things, on the resolution of the sensors, the quality of the pattern recognition and the distance to the object. Each point is a spatial measurement value, but without direct surface information or topological links to neighboring points.

In addition to the spatial coordinates, the point cloud often also contains other information, such as color values (RGB) or intensity information, provided the corresponding sensors are available. This additional data is helpful for subsequent classification, texturing or differentiating between materials.

The point cloud itself is initially unorganized. This means that there is no explicit connection between neighboring points. Point clouds are therefore not directly suitable for classic CAD processes or physical simulations, as they do not have a topological or surface-like structure. In order to obtain a closed surface, a subsequent surface reconstruction is necessary, for example by triangulation (e.g. Poisson Surface Reconstruction or Delaunay method), in which neighboring points are connected to form a coherent mesh.

A central characteristic of point clouds is their non-uniformity. The point density varies depending on the surface orientation, the distance to the sensor and the quality of the pattern recognition. This inhomogeneity requires adaptive methods for smoothing, interpolation or standardization during further processing. Point clouds are also susceptible to measurement errors: noise, missing points and holes are caused by shadows, reflections or inaccessible areas. These errors must be corrected or segmented by the software.

In practical use, the point cloud forms the basis for almost all further steps in digital geometry reconstruction. It can be used for direct measurements (e.g. distance, curvature, surface analysis) or converted into a closed, analyzable model. Despite its simplicity, the point cloud is therefore a central tool in optical 3D measurement technology, as it provides a largely loss-free and precise representation of the real object geometry – albeit always subject to the physical and algorithmic limits of the acquisition method.

Processing point clouds in the software

In its pre-processing stage, the Revopoint MetroX software provides several targeted tools that start before the actual triangulation of the point cloud and aim to optimize the raw data in terms of completeness, low noise and geometric coherence. These processing steps are crucial in order to be able to carry out the subsequent surface generation (triangulation) on a consistent and cleansed database.

The first step is usually geometric alignment, especially when several individual images (scans) are merged. The software offers methods for coarse and fine alignment, typically in the form of ICP-based algorithms (Iterative Closest Point), in which overlapping point cloud sections are registered and geometrically aligned. In addition, manual reference points or geometric features can be used to support registration.

An elementary processing step is the filtering of outliers and noise points. As structured light systems generate point noise even under optimum conditions – for example due to stray light, measurement errors or insufficient pattern projection – the software allows the use of statistical filters. These remove individual points or small-scale anomalies based on criteria such as point density, distance to neighboring points or deviation from local surface gradients.

In addition, smoothing or surface normalization is used, in which neighbouring points are adjusted in such a way that local height differences are reduced without distorting the global shape. This improves the subsequent mesh generation, as very noisy surfaces can lead to incorrect triangulation. Another preparatory step is the completion of incomplete areas. The software can close small holes by local interpolation or extrapolation. This is usually not done on the basis of the final mesh structure, but directly on the point cloud level. The detection of such gaps is based on the local point distribution and the neighborhood graph.

The reduction of the number of points (decimation) is also provided for, particularly in preparation for computationally intensive processing steps. Excess points are removed without significantly affecting the geometry. The software uses voxel-based methods, for example, in which a representative point is retained within a spatial grid. Last but not least, the software allows semantic segmentation or grouping of point areas. This is used less for technical cleanup than for better handling of complex scenes, for example by separating object and carpet pad or removing background information.

The processing of point clouds in the Revo Scan 5 MetroX software, which was specially developed for the Revopoint MetroX 3D scanner, enables comprehensive processing of 3D scan data. This process comprises several steps, ranging from raw data fusion to the creation of a final 3D model.

• Fusion of the point cloud:
  After scanning, the captured raw data is merged into a uniform point cloud. The average distance between neighboring points, known as the “point distance”, can be adjusted. For detailed objects, it is advisable to select a smaller point distance, while a larger distance is suitable for flat, less structured surfaces. The software also offers the option to automatically remove marker points that were used during scanning.
• Editing the point cloud
  After the fusion, various tools are available to optimize the point cloud:
  – Remove Isolation: Isolated point groups that are not connected to the main model can be recognized and removed.
  – Detect overlaps: Overlapping data points are identified and eliminated to ensure a consistent model.
  – Smoothing: Noise is reduced to achieve a smoother surface.
  – Simplification: Point density can be reduced to reduce the amount of data and facilitate processing.