Point cloud

from Wikipedia, the free encyclopedia
Animated point cloud model of a torus

A point cloud or a cluster of points ( English point cloud ) is a set of points in a vector space having an unorganized spatial structure ( "cloud"). A point cloud is described by the points it contains, each of which is recorded by its spatial coordinates. Georeferenced point clouds contain points in an earth-related coordinate system. In addition to the points, attributes such as B. geometric normals, color values ​​or measurement accuracy can be recorded.


Mobile, terrestrial laser scanner for the acquisition of 3D point clouds

The generation can in principle take place via scanning methods (e.g. terrestrial or aircraft-based laser scanning ) or photogrammetric methods and generally by scanning object surfaces using systems such as coordinate measuring machines or scanning 3D scanners . Optical scanners are divided into laser technology , which work according to the triangulation principle, and normal light scanners, which work according to the strip light method ("coded light"). C. Teutsch, for example, gives a comprehensive overview of the variety and performance of current optical scanning methods and the further processing of the resulting 3D data / point cloud. A four-dimensional (time-variant) discrete spatial model of an environment can be set up by recording a spatial section multiple times at different times. Each point of the cloud is localized in time and space ( XYZ coordinates ) and can subsequently also be georeferenced.


Due to the enormous volume of data, storage challenges arise in terms of storage space and efficient access to individual areas of a point cloud. Multi-resolution data structures are used to implement storage processes: "In order to be able to process the data efficiently and visualize it in real time, software implementations require out-of-core algorithms and level-of-detail structures." by Octrees . Attempts are currently being made to standardize the data storage of 3D point clouds in order to enable data management that is compatible with other disciplines. In addition to compatibility, the large amounts of data should be easier to manage and interactive research approaches should be made easier and thus promoted.


3D point cloud of a stegosaurus

To visualize massive point clouds, out-of-core algorithms are required that allow the 3D rendering system to access points of a point cloud efficiently, depending on the resolution . Point-based rendering , in particular, enables a differentiated graphic display of point clouds, e.g. B. in for points of different categories (e.g. facade points, roof points, vegetation points, etc.). Continuous surfaces can be derived from point clouds with a sufficiently high point density using 3D rendering processes in order to achieve the most closed possible visualization of surface areas.

There are many ways to create a closed 3D surface from a point cloud. Some approaches, such as Delaunay triangulation , alpha shapes or ball pivoting, build a network of triangles using the normal vectors of the individual points. Other approaches, such as B. the marching cubes algorithm extract a polygon mesh using voxel- based approaches. These play a role above all for imaging processes in medicine. The currently best-known open source visualization programs for 3D point clouds include CloudCompare and MeshLab. However, these applications are limited in certain areas. Although both programs can display point clouds and thus facilitate the exchange and communication via 3D data, processing of the point cloud is only possible to a limited extent. In addition, the amount of data to be processed is limited for both programs.

The free program library of the Point Cloud Library (PCL) has been available since 2011 . This offers numerous algorithms for processing n-dimensional point clouds and three-dimensional geometries. The modules contained therein enable z. B. the filtering , registration , segmentation , surface reconstruction or visualization. The PCL has a similar status for 3D image processing to OpenCV for 2D image processing.



For geomorphological analyzes , digital elevation models in particular are derived from 3D point clouds. This enables a large number of surface-related analyzes. Soil erosion processes were investigated using terrestrial laser scanners . In order to quantify the soil erosion, 3D point clouds were measured at different times and compared with one another. This allows statements to be made about where sediments are removed from or how they are subsequently displaced. In the field of glaciology , the movements and changes in glaciers are documented and investigated using point clouds. In addition, a number of fluvial research approaches and applications deal with the analysis of point clouds. In this way, changes in river courses can be observed over longer periods of time.


Digital terrain model generated from 3D point clouds from a laser scan flight

Many developments in applications relating to 3D point clouds come from the field of archeology . By analyzing surface shapes, conclusions can be drawn about past settlement structures. As a result, areas used anthropogenically can be identified and their arrangement and organization analyzed. In this context, point clouds make it possible, on the one hand, to visualize the earth's surface in order to give archaeologists the opportunity to search for specific structures and their location, and on the other hand, to systematically scan larger areas according to predefined patterns through automated processes. Another possible use of 3D point clouds in archeology is the modeling of historical sites.

Agriculture and forestry

Point cloud model of a forest

In agriculture and forestry, 3D point clouds are used primarily for monitoring applications. The use of LiDAR data makes it possible to monitor large areas of agricultural land without having to be on site. This prevents the crops and their surroundings from being disturbed or destroyed by the presence of humans. In addition, the workload can be significantly reduced, since the agricultural areas are monitored over a large area and targeted influence can be exercised. Point cloud-based methods are used particularly in the field of precision farming . Statements can be made about plant growth by analyzing recorded 3D point clouds. The aim of the analysis is to identify areas of an agricultural area that show special growth patterns in order to be able to fertilize them individually. In the field of forestry z. B. investigated the relationship between tree health and bark beetle infestation . Differences in the tree canopy structures in healthy or infested trees can be derived from the 3D point cloud data. In this way, specifically infected trees are identified and subsequently treated.

Urban geography

3D city model

Laser scanning data is increasingly used in the area of urban development and spatial planning processes. Using computer algorithms , it is possible to segment 3D point clouds of an area into different areas. A distinction can be made between vegetation, buildings and undeveloped areas. Information from such analyzes can subsequently be taken into account in urban planning decisions . The individual areas are recognized based on the arrangement of the scanned points within the cloud. Recorded tree tops are characterized, for example, by an irregular arrangement of the points, whereas buildings have clear linear structures. Another algorithm that filters regular areas within a point cloud was developed to identify potential locations for photovoltaic systems . In addition, point clouds make it possible to generate detailed 3D models of an urban area. In order to be able to record structural changes within an urban area, point clouds are recorded at different times and the point distances between the individual surveys are compared.

Natural hazards

In natural hazard and risk management , 3D point clouds enable a detailed analysis of natural events and targeted monitoring of potential danger areas. This enables the optimization of early warning systems. Gravitational mass movements are monitored using time series of 3D point clouds in order to recognize dynamics at an early stage and to be able to warn affected persons. In areas at risk of falling rocks , areas with infrastructure can be obtained through regular scanning and analysis of the 3D point clouds generated from them, important information for local risk management . The roughness of a surface is an important parameter in the analysis of natural hazards. Statements about the structure of surfaces can be made using 3D point clouds. This makes it possible to draw conclusions about possible natural hazards, such as floods , falling rocks or avalanches , based on the nature of the soil and its potential properties . For risk management measures it is important to know the exact extent of the event in order to react correctly, in good time and sufficiently with the implementation of protective measures. The acquisition of 3D point clouds by airborne laser scanning has the advantage over traditional methods of remote sensing (e.g. photogrammetry ) that the active measuring system can penetrate vegetation and thus ground points can be recorded.

Computer graphics

A point cloud is used:

  • for visualization in order to display scanned objects and surfaces on the computer;
  • for modeling ; the original, (mostly) closed surface is restored by means of surface reconstruction. The resulting surfaces mostly consist of polygons (see also meshing );
  • as a basis for geometric calculations, e.g. B. for measuring people (body scanning) or objects;
  • to identify people or objects.
  • in medicine (especially forensics, criminalistics); to create a surface model to document injuries and to reconstruct them. A further method is the combination of laser scanning and radiological procedures. A virtual 3D model of a body can thus be created. In this model all injuries (internal and external) become visible.


Body shop

In the CAD area , point clouds are used to import scanned design objects into the CAD system. In the case of demanding forms (e.g. automobile body), a scale model of clay is often created. With draw knives and other hand tools, the shapes are created from a modeling clay and then scanned. In the case of symmetrical components (e.g. engine hood), only one side is modeled. This is then scanned with tactile, measuring or optical scanners. The resulting 3D geometry initially only consists of points in space (XYZ coordinates). This point cloud is either read into special software for reverse engineering or, in some cases, read directly into the CAD software. The often common conversion of points to simple surface networks, which is usually sufficient in computer games, is not sufficient in automotive engineering. Here, Bézier and NURBS surfaces are placed through the points and possible measurement errors are compensated for using filter methods. The surfaces created in this way have exact surface boundaries and precisely defined transitions to the adjacent surfaces. The CAD model created from the clay model can be mirrored in CAD. An exactly described model of a car body (or parts thereof) is then available.

Point clouds in statistics

In statistics and exploratory data analysis, point clouds are used to graphically depict bivariate relationships (see scatter plot , correlation ). They allow you to get a simple visual impression of the direction and closeness of the relationship and to detect outliers in the data set. Areas of a point cloud that are denser than others are called clusters .

See also

Web links

Individual evidence

  1. J. Otepka, p Ghuffar, C. Waldhauser, R. Hochreiter, N. Pfeifer: georeferenced Point Clouds: A Survey of Features and Point Cloud Management. In: ISPRS Int. J. Geo-Inf. 2, 2013, pp. 1038-1065. doi: 10.3390 / ijgi2041038
  2. F. Leberl, A. Irschara, T. Pock, P. Meixner, M. Gruber, S. Scholz, A. Wiechert: Point clouds: Lidar versus 3D vision. In: Photogram. Closely. Remote Sens. 76, 2010, pp. 1123-1134.
  3. ^ C. Teutsch: Model-based Analysis and Evaluation of Point Sets from Optical 3D Laser Scanners. (= Magdeburg writings for visualization. Volume 1). Shaker Verlag , 2007, ISBN 978-3-8322-6775-9 .
  4. R. Richter, J. Döllner: Potentials of massive 3D point cloud data streams. (Geoinformatics 2012 - "Mobility and Environment"). Shaker Verlag, 2012, pp. 215-222.
  5. ^ Jochem Andreas, Bernhard Höfle, Martin Rutzinger: Extraction of Vertical Walls from Mobile Laser Scanning Data for Solar Potential Assessment. In: Remote Sensing. 3, 4, December 2011, pp. 650-667.
  6. R. Richter, J. Döllner: Integrated real-time visualization of massive 3D point clouds and georeferenced texture data. In: Photogrammetry, remote sensing, geographic information. (PFG), vol. 2011, no. 3, 2011, pp. 145–154.
  7. P. Goswami, F. Erol, R. Mukhi, R. Pajarola, E. Gobbetti: An Efficient Multi-resolution Framework for High Quality Interactive Rendering of Massive Point Clouds using Multi-way kd-Trees. In: The Visual Computer. 29, 1, 2013, pp. 69-83.
  8. ^ J. Elseberg, D. Borrmann, A. Nüchter: One billion points in the cloud - an octree for efficient processing of 3D laser scans. In: ISPRS Journal of Photogrammetry and Remote Sensing. February 2013.
  9. P. van Oosteroma, O. Martinez-Rubib, M. Ivanovab, M. Horhammerc, D. Geringerc, S. Ravadac, T. Tijssena, M. Kodded, R. Gonçalves: Massive point cloud data management: Design, implementation and execution of a point cloud benchmark. In: Computers & Graphics. 49, 2015, pp. 92-125.
  10. M. Wimmer, C. Scheiblauer: Instant points: Fast rendering of unprocessed point clouds. In: Proceedings Symposium on Point-Based Graphics 2006. 2006, pp. 129-136.
  11. ^ R. Richter, J. Döllner: Out-of-Core Real-Time Visualization of Massive 3D Point Clouds. 7th International Conference on Virtual Reality, Computer Graphics, Visualization and Interaction in Africa, 2010, pp. 121–128.
  12. ^ R. Richter, M. Behrens, J. Döllner: Object class segmentation of massive 3D point clouds of urban areas using point cloud topology. In: International Journal of Remote Sensing. vol. 34, no. 23, 2013, pp. 8394-8410.
  13. R. Pintus, E. Gobbetti, M. Agus: Real-time rendering of massive unstructured Raw Point Clouds using screen-space operator. In: The 12th International Symposium on Virtual Reality, Archeology and Cultural Heritage. October 2011, pp. 105-112.
  14. Paolo Cignoni: Meshing Point Clouds. September 7, 2009, accessed February 19, 2017 .
  15. A. Palha, A. Murtiyoso, J.-C. Michelin, E. Alby, P. Grussenmeyer: Open Source First Person View 3D Point Cloud Visualizer for Large Data Sets. In: I. Ivan, A. Singleton, J. Horák, T. Inspektor (Eds.): The Rise of Big Spatial Data. Springer International, Cham 2017, pp. 27–39.
  16. Point Cloud Library (PCL). Retrieved March 2, 2017 (English).
  17. M. Rutzinger, B. Höfle, M. Vetter, N. Pfeifer: Digital Terrain Models from Airborne Laser Scanning for the Automatic Extraction of Natural and Anthropogenic Linear Structures. In: Mike J. Smith, Paolo Paron, James S. Griffiths: Geomorphological Mapping: a professional handbook of techniques and applications. Elsevier, Oxford 2011, pp. 475-488.
  18. ^ AD Meijer, JL Heitman, JG White, RE Austin: Measuring erosion in long-term tillage plots using ground-based lidar. In: Soil & Tillage Research. 126, 2016, pp. 1–10.
  19. N. Micheletti, M. Tonini, SN Lane: Geomorphological activity at a rock glacier front detected with a 3D density-based clustering algorithm. In: Geomorphology. 278, 2017, pp. 287-297.
  20. ^ S. Ghosh: Hydrological changes and their impact on fluvial environment of the lower damodar basin over a period of fifty years of damming The Mighty Damodar River in Eastern India. In: Procedia - Social and Behavioral Sciences. 19, 2011, pp. 511-519.
  21. ^ D. White, KA Corcoran: Geospatial Applications in Archeology. In: International Encyclopedia of the Social & Behavioral Sciences. 2, 2015, pp. 97-103.
  22. JAB López, GA Jiménez, MS Romero, EA García, SF Martín, AL Medina, JAE Guerrero: 3D modeling in archeology: The application of Structure from Motion methods to the study of the megalithic necropolis of Panoria (Granada, Spain). In: Journal of Archaeological Science. 10, 2016, pp. 495-506.
  23. B. Höfle: Radiometric Correction of Terrestrial LiDAR Point Cloud Data for Individual Maize Plant Detection. 2014, pp. 94-98.
  24. a b G. Vosselman, HG Maas: Airborne and Terrestrial Laser Scanning. Whittles Publishing. Twente 2010.
  25. K. König, B. Höfla, M. Hämmerle, T. Jarmer, B. Sigmann, H. Lilienthal: Comparative classification analysis of post-harvest growth detection from terrestrial LiDAR point clouds in precision agriculture. In: ISPRS Journal of Photogrammetry and Remote Sensing. 104, 2015, pp. 112–125.
  26. ^ FE Fassnacht, H. Latifi, A. Ghosh, PK Joshi, B. Koch: Assessing the potential of hyperspectral imagery to map bark beetle-induced tree mortality. In: Remote Sensing of Environment. 140, 2014, pp. 533-548.
  27. ^ QY Zhou, U. Neumann: Complete residential urban area reconstruction from dense aerial LiDAR point clouds. In: Graphical Models. 75, 2013, pp. 118-125.
  28. B. Höfle, M. Hollaus, J. Hagenauer: Urban vegetation detection using radiometrically calibrated small-footprint full-waveform airborne LiDAR data. In: ISPRS Journal of Photogrammetry and Remote Sensing. 67, 2012, pp. 134-147.
  29. A. Jochem, B. Höfle, M. Rutzinger, N. Pfeifer: Automatic Roof Plane Detection and Analysis in Airborne Lidar Point Clouds for Solar Potential Assessment. In: Sensors. 7, 2009, pp. 5241-5262.
  30. R. Richter, J. Kyprianidis, J. Döllner: Out-of-Core GPU-based Change Detection in Massive 3D Point Clouds. In: Transactions in GIS. 17, 2013, pp. 724-741.
  31. ^ J. Travelletti, JP Malet, C. Delacourt: Image-based correlation of Laser Scanning point cloud time series for landslide monitoring. In: International Journal of Applied Earth Observation and Geoinformation. 32, 2014, pp. 1-18.
  32. ^ RA Kromer, DJ Hutchinson, MJ Lato, D. Gauthier, T. Edwards: Identifying rock slope failure precursors using LiDAR for transportation corridor hazard management. In: Engineering Geology. 195, 2015, pp. 93-103.
  33. M. Hollaus, C. Aubrecht, B. Höfle, K. Steinocher, W. Wagner: Roughness Mapping on Various Vertical Scales Based on Full-Waveform Airborne Laser Scanning Data. In: Remote Sensing. 3, 2011, pp. 503-523.
  34. R. Malinowski, B. Höfle, K. Koenig, G. Groom, W. Schwanghart, G. Heckrath: Local-scale flood mapping on vegetated floodplains from radiometrically calibrated airborne LiDAR data. In: ISPRS Journal of Photogrammetry and Remote Sensing. 119, September 2016, pp. 267-279.
  35. Ursula Buck: Digital photogrammetry and laser scanning in forensics. (PDF) Hochschule für Technik Stuttgart, 2006, accessed on February 24, 2017 .
  36. Ursula Buck: Laser scanning in criminology . In: Journal of Geodasia, Geoinformation and Land Management . tape 135 , no. 3 , 2010, p. 190-198 .