3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and

Geodesy

Theresa Meyer

and Ansgar Brunn

University of Applied Sciences Wuerzburg-Schweinfurt, Roentgenring 8, Wuerzburg, Germany

Keywords:

3D Point Clouds, Point Cloud Tiling, Geodatabase, GIS, 3D Applications.

Abstract:

Besides the common approach of an exclusively ﬁle based management of 3D point clouds, meanwhile it is

possible to store and process this special type of massive geodata within spatial database systems. Users beneﬁt

from the general advantages of database solutions and especially from the potentials of a combined analysis of

original 3D point clouds, 2D rasters, 3D voxel stacks and 2D and 3D vector data in order to gain valuable geo-

information. This paper describes the integration of 3D point clouds into an open source PostgreSQL/PostGIS

database using the Pointcloud extension and functions of the Point Data Abstraction Library (PDAL). The

focus is on performing three-dimensional spatial queries and the evaluation of different tiling methods for

the organization of 3D point clouds into table rows, regarding memory space, performance of spatial queries

and effects on interactions between point clouds and other GIS features within the database. A new approach

for an optimized point cloud tiling, considering the individual geometric characteristic of a 3D point cloud,

is presented. The results show that an individually selected storage structure for a point cloud is crucial for

low memory consumption and high-performance 3D queries in PostGIS applications, taking account of its

three-dimensional spatial extent and point density.

1 INTRODUCTION

3D point clouds are increasingly gaining importance

for various applications mainly in ﬁelds of geodesy,

geoinformatics, architecture and archaeology.

Even though modern measuring devices such as

laser scanners, UAS (unmanned aerial systems) and

mobile mapping systems enable a very fast genera-

tion of 3D point clouds, the handling of large amounts

of point data is a challenging task. Even though

point clouds are meanwhile indispensable input data

for many projects processed by even small engineer-

ing ofﬁces. These include e.g. as-built documen-

tations of buildings and plants (Tang et al., 2010),

change detection and deformation analysis of en-

gineering constructions (Mukupa et al., 2017) and

three-dimensional terrain modeling as a basis for

building projects (Burger et al., 2016).

The management of 3D point clouds is currently

usually ﬁle based. Common exchange formats for 3D

point cloud data are LAS (ASPRS, ), PCD (Rusu and

Cousins, 2011) and E57 (Huber, 2011). These point

https://orcid.org/0000-0002-3146-7645

https://orcid.org/0000-0002-8692-3636

cloud formats enable a lossless transport of up to bil-

lions of 3D point coordinates (X,Y,Z) and dozens of

additional point attributes such as Intensity, Return

Number, Red, Green, Blue, NIR, Classiﬁcation and

Point Source ID.

As an alternative to an exclusively ﬁle based

handling, it is meanwhile possible to manage 3D

point clouds within spatial database systems (geo-

databases). General advantages of a database solu-

tion are data consistency, security, reduced storage

space and multi user access. Furthermore another

even greater beneﬁt is that 3D point clouds can be re-

lated to all types of spatial and non-spatial data, such

as raster, vector and administrative data, within the

geodatabase.

This paper describes the integration of 3D point

clouds as GIS features next to common raster and vec-

tor data in spatial database systems on the example of

PostgreSQL/PostGIS with the Pointcloud extension

(PostgreSQL, 2019), (PostGIS, 2019), (Pointcloud,

2019). Pointcloud extends PostgreSQL/PostGIS with

two new data types and various operations to store and

analyze 3D point clouds using GIS tools.

The focus is on the evaluation of different ap-

proaches for a blockwise organization of 3D point

154

Meyer, T. and Brunn, A.

3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and Geodesy.

DOI: 10.5220/0007840901540163

In Proceedings of the 5th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2019), pages 154-163

ISBN: 978-989-758-371-1

clouds in table rows. With regard to 3D applica-

tions, a new method for a three-dimensional tiling of

3D point clouds is presented. Additionally, the paper

includes a practical guide to performing multistage

operations for querying 3D point clouds with three-

dimensional vector features.

2 3D POINT CLOUDS IN GIS

Integration of original 3D point clouds is useful for

many GIS applications such as generation of 3D

building models as an important part of 3D city mod-

els (Haala and Brenner, 1997), detection of vegetation

in urban areas (Hoeﬂe and Hollaus, 2010), for terrain

and object modeling in ﬂood simulations (Merwade

et al., 2008) and for the extraction of roof and facade

surfaces for solar potential analyzes (Voegtle et al.,

2005), (Jochem et al., 2011).

When dealing with point clouds two fundamen-

tal cases have to be distinguished. If there is only or

mostly one Z value for each combination of X and

Y, the point cloud can be considered as a 2.5D point

cloud. Such point clouds are usually generated from

airborne data acquisition e.g. airborne laser scanning

or UAV photogrammetry. They can easily be con-

verted to a raster format in order to process and an-

alyze them with common GIS tools. On the other

hand, unstructured point clouds with a pronounced

3D character can be considered as ”true” 3D point

clouds. They are typically generated from terrestrial

or mobile laser scanning, even combining indoor and

outdoor scenarios. A raster interpolation would cause

an irreversible loss of information for such types of

3D point clouds. As a result, the need for suitable 3D

storage structures and 3D operations arises for ”true”

3D point clouds.

The Feature Geometry Model is a general stan-

dard for 3D GIS. It supports three-dimensional coor-

dinates and object classes for 3D surfaces, triangular

irregular networks (TINs), 3D bodies and 3D topol-

ogy. Part 3 of the SQL Multimedia and Application

Packages (SQL/MM Spatial) extend the Simple Fea-

ture Model with useful 3D operations.

While there are well-established and standardized

models and structures for 2D raster and vector data

in geodatabase systems, the integration of original

3D point clouds is not yet speciﬁed or standardized.

Currently, the Open Geospatial Consortium (OGC)

Point Cloud working group is dedicated to deﬁning

requirements for interoperability and implementing

standards in the application of 3D point clouds (OGC,

2019).

3 PostgreSQL/PostGIS WITH

POINTLCOUD EXTENSION

Common geometric data types such as POINT Z and

MULTIPOINT Z are not appropriate for the huge

number of points within a typical 3D point cloud,

e.g. from a terrestrial 3D laser scan. The Pointcloud

extension enables the persistent storage of 3D point

clouds and at the same time efﬁciently accessing all

additional point attributes. Pointcloud extends Post-

GIS with the new data types PcPoint and PcPatch.

Several hundreds or thousands of spatially adjacent

points are organized into lossless compressed byte ar-

rays, called patches (blocks). Each patch is stored in

an individual table row and corresponds to a PcPatch

object.

Pointcloud deals with the challenge of varying

point attributes by using an XML schema document,

which describes the contents of any particular point.

Every point might contain up to dozens of additional

attributes/dimensions, and each of it can be of any

data type, e. g. with scaling and/or offsets applied.

3.1 Extended Point Cloud Tiling with

PDAL

Point clouds are tiled into patches while writing to the

Postgres database in order to manage and query them

efﬁciently in tables.

The point cloud tiling is implemented by default

during the database import as part of the execu-

tion of command line instructions and functions of

the Point Data Abstraction Library (PDAL) (PDAL,

2019). PDAL provides two basic functions for point

cloud tiling:

• Irregular tiling with ﬁlters.chipper()

All patches are spatially contiguous and non-

overlapping. Each patch has the capacity point

count speciﬁed. The default capacity is 5000

points per patch. Patches are ﬁlled with points

considering their proximity in X/Y plane. Z co-

ordinates are not considered while tiling with the

Chipper function. For each patch, the algorithm

controls the maximum extend only in X and Y di-

rection and rearranges in case of imbalance.

• Regular tiling with ﬁlters.splitter()

Points are organized into regular square patches

when applying the Splitter function. The user

determines a grid size (in meter). The tile origin

is either chosen randomly or is also determined

by specifying an X and Y coordinate. This regu-

lar tiling method creates patches with the same ex-

3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and Geodesy

155

tent in the X/Y plane, but with varying numbers

of points.

Regardless of the chosen tiling method, the user has

to be aware that both, applying ﬁlters.chipper() and

ﬁlters.splitter(), is not a 3D tiling but only a 2D tiling

of a 3D point cloud. The individual extend along the

Z axis is not considered for a point cloud. As a result

of a 2D tiling, points that may be far apart because of

different Z values in 3D space may still be assigned to

the same patch array. This leads to very unbalanced

patches especially for point clouds with a very pro-

nounced 3D character (cf. ﬁg. 1).

An improved tiling method can be applied in or-

der to achieve a 3D tiling. That approach takes ad-

vantage of an additional PDAL function named ﬁl-

ters.range(). The Range function enables the ﬁlter-

ing of a point cloud by any attribute/dimension. The

maximum Z extension for each patch can be lim-

ited when applying ﬁlters.range() on the Z dimen-

sion. This procedure corresponds to a contouring

e.g. with 1m interval. The Range function can be

used to create a ’Z’ extended three-dimensional point

cloud tiling when combining with ﬁlters.chipper() or

ﬁlters.splitter() (cf. ﬁg. 2).

3.2 Integration of PostGIS

All functions of the Pointcloud extension refer to the

two central Pointcloud objects PcPoint and PcPatch.

Overall, PcPoints play only a minor roll. Although

it is possible to create Pointcloud tables of PcPoints

with a valid schema, these points are primarily needed

only for special queries e.g. as an intermediate step in

order to convert a small result set of point cloud points

into PostGIS points of the type POINT Z. Full-scale

3D point clouds are exclusively managed in PcPatch

tables, and Pointcloud provides a number of special

functions for this purpose.

There are functions for creating and dis-

solving, sorting and compressing/uncompressing

patches (e.g. PC MakePatch(), PC Union(),

PC Sort() and PC Compress()) and functions

for returning content information of patches

(e.g. PC Summary(), PC NumPoints() or

PC PatchMax() and PC PatchMin()). Furthermore

the Pointcloud functions PC FilterGreaterThan(),

PC FilterLessThan(), PC FilterBetween() and

PC FilterEquals() enable a ﬁltering of points based

on attribute values such as intensity, classiﬁcation

code, Z value or time-stamp.

Interactions between Pointcloud and PostGIS ob-

jects become possible only with enabling the addi-

tional extension Pointcloud PostGIS for the database.

Patches can be checked to see whether they over-

lap the geometry of a PostGIS object using the

function PC Intersects() and new patches can be

generated which represent the result set of an in-

tersection between a PcPatch and e.g. a poly-

gon using PC Intersection(). Both PC Intersects()

and PC Intersection() only accept 2D objects and

therefore do not enable spatial queries in 3D. The

PC EnvelopeGeometry() function returns the two-

dimensional boundary of a patch as a PostGIS poly-

gon and PC BoundingDiagonalGeomery() enables

the creation of a 3D index by returning the 3D bound-

ing box diagonals of a patch as LINESTRING Z.

3.3 Multistage 3D Query

The Pointcloud PostGIS extension offers the func-

tions PC Intersects() and PC Intersection() for inter-

section queries and operations involving Pointcloud

and PostGIS features. Both functions do not accept

three-dimensional query objects of the type POLY-

HEDRALSURFACE Z but only two-dimensional

PostGIS features e.g. polygons. Thus, point cloud

points can initially only be queried projected onto

the X/Y-plane. But there is a practical multistage

workaround for spatially querying a 3D point cloud

with a 2.5D volumetric object in 3D space:

1. The Pointcloud function PC BoundingDiagonal-

Geometry() returns the diagonal of the 3D

bounding boxes of all patches as 3D Lines

(LINESTRING Z).

2. In order to enable high-performance queries in

3D, a 3D index is set up for the diagonals resp.

their 3D bounding boxes.

3. By using the index operator &&&, the 3D bound-

ing boxes of all patch diagonals are selected,

which overlap the 3D bounding box of any 3D ob-

ject of the type POLYHEDRALSURFACE Z.

4. The 3D bounding boxes of the patch diagonals

are converted to solids using the PostGIS func-

tion Box3D() in order to enable an exact test for

intersection using the ST 3DIntersects() operation

(SFCGAL backend only) (cf. ﬁg. 4).

5. The selected solids are returned to the original Pc-

Patch objects with an inner join via a key attribute

(ID).

6. The preselected point cloud patches are tailored

to the limitations of the query object in X and

Y using the Pointcloud function PC Intersection()

(cf. ﬁg. 5).

7. In dependence of the tiling method, the patches

still contain 3D points that are above or below the

query object. For queries with 2.5D objects, the

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

156

Figure 1: 2D tiling of a 3D point cloud with 3.3 million points with ﬁlters.chipper() and 100 points capacity.

Pointcloud function PC FilterBetween() is rec-

ommended. It can be used to exactly limit the

minimum and maximum Z extents of the ﬁnal re-

sult patches (cf. ﬁg. 6).

The result of such a multistage 3D query is a selec-

tion of patches which only contain 3D points that are

within the query object or on its boundary.

A possible scenario is e.g. interactions between a

point cloud and a 3D building model. The query ob-

ject could be a model of an indoor wall and task is

the selection of all 3D points from a terrestrial laser

scan that probably represent the object. Such a point

subset can be further processed and analyzed e.g. as

part of an as-built documentation. Subsets of a point

cloud can be exported with PDAL from the database

into any point cloud format, further processed with

external functions and direct database connection or

converted as PostGIS points of type POINT Z. The

latter also enables point queries with any 3D objects

in PostGIS. But in practice works only for dispropor-

tionately small subsets of 3D points that barely corre-

spond to the essence of a 3D point cloud.

3.4 Comparison of Tiling Methods

The decision for a speciﬁc tiling method and con-

ﬁguration is entirely up to the user. In the PDAL

documentation, the only recommendation is that the

space requirement of a single patch should not ex-

ceed the default 8 KB page size of a PostgreSQL

database, and for most LIDAR data, this should prac-

tically mean a patch size of between 400 and 600

points (PDAL, 2019). However, the default capac-

ity of ﬁlters.chipper() is at 5000 points and therefore

exceeds this recommendation by a factor of 10.

For the user, the question arises which tiling

method (regular, irregular, 2D, 3D) to choose for

which type of point cloud and concrete project def-

inition. The impact of ﬁlters.chipper() and ﬁl-

ters.splitter() in various conﬁgurations on storage

space requirements within the database were exam-

ined. It is recommended to choose a large capacity for

Chipper tiling resp. a large grid size for Splitter tiling

in order to achieve a maximum and lossless compres-

sion of the data, independent of effects on interactions

in database operation. The more 3D points within a

patch, the fewer patches are created in total and the

more effective the standard compression method ”di-

mensional”. This compression mode stores patches as

collections of dimensional data arrays, with an “ap-

propriate” compression for each attribute/ dimension

applied (e.g. run-length encoding for dimensions with

low variability) (Boundless, 2015).

Furthermore, a comparison between 2D and 3D

tiling shows that a 3D tiling has clearly positive ef-

fects on storage space requirements of Pointcloud ta-

bles (cf. ﬁg. 7 and 8). It can be assumed that this posi-

tive effect is also due to the compression of the points

within the patches, since points within a 3D tile are

strongly adjacent. Closely adjacent points will tend

to have more homogeneous dimensions (e.g. coorid-

inates, color, intensity, classiﬁcation codes, etc.), and

as a result, can be compressed more effectively. Com-

parisons of 2D and 3D tiling, with and without com-

3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and Geodesy

157

Figure 2: 3D tiling of a 3D point cloud with 3.3 million points with ﬁlters.chipper(), 100 points capacity and 1m contour

interval.

pression have conﬁrmed this assumption.

The tiling method also has an impact on spatial

queries concerning query duration and number of se-

lected points in single stages of a multistage opera-

tion. Generally, there is a difference between queries

on patch level and queries on point level. The ﬁrst

ﬁlter stage is a preselection of patches based on an in-

dex operation and approximated geometries (stage 1).

Stage 2 represents the intersection between patches

and the original shape of a query object instead of its

3d bounding box. Finally, the last stage is a selection

on point level using Pointcloud functions (stage 3).

The chosen tiling method has no impact on the

fact that the stage of preselection always returns the

biggest number of patches resp. points. The num-

ber of candidate patches and points is continuously

reduced during the query procedure until the ﬁnal

subset of points is reached. Regardless of the tiling

method, the number of resulting single points is ex-

actly the same, but there are differences in the preci-

sion of the result after ﬁlter stages 1 and 2 depending

on the individual tiling of a point cloud.

A query procedure eventually returns the same set

of points for each tile method, but there are signiﬁcant

differences in the intermediate results for each ﬁlter

stage. These differences can affect the performance

of practical applications. The smaller the number of

points in a patch, the ﬁner the tiling, and the more pre-

cise the query results, but the higher the total number

of patches and thus the overhead of operations in the

database.

The individual geometric characteristic of a 3D

point cloud (especially the ratio of Z extend to X and

Y) is decisive for the different effects of varying 2D

and 3D tiling methods. Besides, the shape of an in-

volved query object also has an impact on the results

of a multistage query.

The diagrams in ﬁgure 9 show the effects of dif-

ferent tiling methods on the result sets of point re-

quests of stages 1 and 2 for an example point cloud.

All points are requested that are inside a cuboid in 3D

space (cf. ﬁg. 4 - 6). The example point cloud extends

over several ﬂoors inside a building and therefore has

a very strong 3D character. The results of 2D and 3D

tiling methods, each with patches of different sizes,

show that 3D tiling is generally better suited than the

standard 2D variant for the most accurate results pos-

sible from patch level queries in ﬁlter stages 1 and 2.

4 ALTERNATIVE SOLUTIONS

FOR THE INTEGRATION OF

3D POINT CLOUDS IN GIS

The processing of original 3D point clouds is not a

classic GIS task, but is largely implemented in spe-

cial software for point clouds, laser scanning, digital

photogrammetry and/or 3D modeling. For example,

CloudCompare (CloudCompare, 2019) is a very ver-

satile open source point cloud software but it does not

provide a programming or database interface.

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

158

Figure 3: Result of PC Intersection().

There exist some open source libraries for pro-

cessing and analyzing 3D point clouds independent

of any proprietary software solutions. Such libraries

can be used to extend the functionality of open point

cloud management systems. In the context of Post-

GIS/Pointclound, the PDAL library is one of the

most important. PDAL functions realize the practical

database import of point clouds by processing instruc-

tions (so-called PDAL pipelines), which are written

in JSON syntax and executed via the command line.

So-called ﬁlter functions can additionally be applied

within the execution of a PDAL pipeline, e.g. neigh-

borhood calculations such as normal vectors, eigen-

values or voxel centroids. Other functions can be used

to colorize a point cloud from an aerial photograph of

an area, to create meshes or to apply arbitrary rotation

and translation transformations. There is also a bridge

between PDAL and the Point Cloud Library (PCL),

whose focus is even more on the development of

new algorithms for processing 3D point clouds (PCL,

2019). Furthermore, PDAL supports the Python pro-

gramming language in the way that Python code can

be embedded into PDAL pipelines. The other way

around, there is a PDAL extension for Python, which

makes it possible to execute PDAL functions within

Python applications.

Another library for 3D data and point clouds is

Open3D (Zhou et al., 2018). It provides general pro-

cessing algorithms for 3D point clouds and function-

ality for point cloud rendering. Open3D is also exe-

cutable via a python extension and therefore suitable

for interactions with point clouds within the PostGIS

database.

The LAStools (only partially free) (Rapidlasso,

2019) and LibLAS library (LibLAS, 2019) are spe-

cially designed for handling LiDAR point clouds in

the LAS format.

Program libraries can expand point cloud data

with additional attributes, and thus increase the infor-

mation content within a database.

There are also proprietary software solutions for

the integration of original 3D point clouds in GIS ap-

plications. ESRI’s LAS-datasets enable an efﬁcient

management of 3D point clouds within the ArcGIS

software-suite (ESRI, 2019). Such data sets store ref-

erences to LAS point clouds in any ﬁle directory. This

provides central access to point clouds that are dis-

tributed in individual ﬁles. Combined with tools for

point cloud generation from aerial image data (the so-

called Ortho Mapping Tools), ArcGIS, as a single sys-

tem, already covers large parts of the overall workﬂow

for generating, processing and GIS based analysis of

e.g. UAV products. Rapidlasso’s LAStools software

suite includes a wide variety of point cloud command

line tools, which also refer to individual LAS ﬁles

(Rapidlasso, 2019). These tools can also extend the

functionality of ArcGIS, QGIS, and Erdas Imagine as

a plugin.

A similar software product for point cloud pro-

cessing is OPALS (Orientation and Processing of

Airborne Laser Scanning Data), developed from the

Technical University Vienna (OPALS, 2019). OPALS

is specially designed for the handling of LiDAR point

clouds even though its functionality is not restricted

to this type of point data. With its various modules

and functions, the software largely covers the entire

processing chain for georeferencing, quality control,

classiﬁcation, ﬁltering and terrain modeling. Even

massive point clouds can be processed with the in-

tegrated database manager. For a direct connection

to GIS software, OPALS can be used via Python

and there exists a special OPALS extension for QGIS

(QPALS, 2019).

Finally, the commercial geodatabase Oracle Spa-

tial also supports the storage and processing of n-

dimensional point clouds with the SDO PC PKG

package. As with Pointcloud for PostGIS, point

clouds are tiled into blocks of adjacent points with

a given point capacity. In contrast to the PostGIS

database, the PC Clip() function enables the selec-

3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and Geodesy

159

Figure 4: Result of ST 3DIntersects().

Figure 5: Result of PC Intersection().

Figure 6: Result of PC FilterBetween().

Figure 7: Comparison of the storage space of a point cloud

tiled with ﬁlters.chipper() in 2D and 3D.

tion of subsets from a point cloud using both 2D

(polygons) and 3D (parallelepiped) geometries (Or-

acle, 2019).

5 CONCLUSIONS

The integration of massive 3D point clouds as features

in geodatabases is quite possible and there are many

Figure 8: Comparison of the storage space of a point cloud

tiled with ﬁlters.splitter() in 2D and 3D.

examples for the beneﬁt of point clouds in GIS appli-

cations. Valuable information can be exchanged be-

tween point clouds and common GIS features based

on their proximity in 2D and 3D. Even general engi-

neering projects, which are not necessarily associated

with classical GIS applications, can be optimized by

managing point clouds in spatial databases. Work-

ﬂows beneﬁt from a combined storage and analysis of

large 3D point clouds with other project data (whether

spatial oder non-spatial/administrative) as well as the

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

160

Figure 9: Effects of different tiling methods in 2D and 3D on the result sets of point requests of stages 1 and 2 for an example

point cloud with approx. 3 million points.

beneﬁts of centralized data management and multi-

user operation.

PostGIS with Pointcloud extension enables a

blockwise storage of 3D point clouds in tables. Un-

fortunately, PostGIS is generally restricted to two-

dimensional operations with point clouds. In addi-

tion, the question of suitable point cloud tiling is still

an open topic with missing solutions and standards.

Although there are some default values, these are

not necessarily the best choices for any type of point

cloud. The Pointcloud extension was originally devel-

oped for LiDAR point clouds from airborne laserscan-

ning, which have rather a 2.5D than 3D character. A

rough 2D tiling is sufﬁcient for such point clouds and

small-scale projects but it is clearly not for 3D point

clouds with a large Z extend and high point density in

large-scale scenes.

In this paper, a new extended and optimized point

cloud tiling has been developed and realized taking

into account the point cloud extend along the Z axis

by using and combining functions of the open source

PDAL library. The results show, that such a 3D tiling

has positive effects on the storage space requirements

of point cloud tables in PostGIS. Furthermore, this

optimized kind of point cloud organization provides

much more precise point sets in early ﬁlter stages on

patch level for 3D applications.

An example for a practical application of a well-

suited tiling of a 3D point cloud is given in ﬁgure 10.

Classiﬁed 3D point cloud tiles are supposed to repre-

sent trees within a forest GIS project. For the purpose

of forest management, these 3D tiles/patches can be

used e.g. for a simulation of the incidence of light

through the treetops on the forest ﬂoor.

In context of topics such as ”Indoor GIS” and the

integration of Building Information Modeling (BIM)

and GIS, the topic ”3D” in GIS is of particular rel-

evance. Current research is devoted to approaches

for indoor routing in building GIS (Wilkening et al.,

2019) and the semi-automatic extraction of indoor

data models based on point clouds (Tessema et al.,

2019). 3D point clouds are captured for different pur-

poses. They have very individual properties and char-

acteristics. These circumstances need to be taken into

account when choosing a suitable point cloud tiling.

There are already extensive comparisons of

3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and Geodesy

161

Figure 10: 3D tiled point cloud for forest management ap-

plications within a ”true” 3D GIS project.

database solutions in which the management of dif-

ferent sized LiDAR point clouds in different systems

(PostGIS, Oracle Spatial and MonetDB) have been

tested and compared (Van Oosterom et al., 2015).

Further investigations of tiling and querying different

types of point clouds within spatial databases is part

of future work. In any case, the focus will be on point

clouds with a pronounced 3D character in large-scale

applications which beneﬁt from ”3D” capabilities e.g.

3D point clouds within a ”Campus GIS” (Wilkening

et al., 2018).

REFERENCES

ASPRS. LAS Speciﬁcation Version 1.4 – R13. The Amer-

ican Society for Photogrammetry and Remote Sens-

ing. https://www.asprs.org/wp-content/uploads/2010/

12/LAS 1 4 r13.pdf.

Boundless (2015). Open Geo Suite 4.6 - home-

page. https://connect.boundlessgeo.com/docs/suite/4.

6/index.html (Accessed 28 Feb 2019).

Burger, S., Elﬂein, A., and Voelter, U. (2016). Photo-based-

scanning als Erweiterung des ingenieurgeodaetischen

Leistungsspektrums. In Schriftenreihe des DVW, vol-

ume 82, pages 175–189.

CloudCompare (2019). CloudCompare - project home-

page. https://www.danielgm.net/cc/ (Accessed 28 Feb

2019).

ESRI (2019). ESRI - ArcGIS documentation. http://

desktop.arcgis.com (Accessed 28 Feb 2019).

Haala, N. and Brenner, C. (1997). Interpretation of Ur-

ban Surface Models Using 2D Building Information.

In Computer Vision and Image Understanding, vol-

ume 72, pages 204–214.

Hoeﬂe, B. and Hollaus, M. (2010). Urban Vegetation De-

tection using high Density Full-Waveform Airborne

LIDAR Data – Combination of Object-based Image

and Point Cloud Analysis. In International Archives

of Photogrammetry, Remote Sensing and Spatial In-

formation Sciences, volume 38, pages 281–286.

Huber, D. (2011). The ASTM E57 File Format for 3D Imag-

ing Data Exchange. In Proceedings of the SPIE Vol.

7864A, Electronics Imaging Science and Technology

Conference (IS&T), 3D Imaging Metrology.

Jochem, A., Hoeﬂe, B., and Rutzinger, M. (2011). Ex-

traction of Vertical Walls from Mobile Laser Scanning

Data for Solar Potential Assessment. In Remote Sens-

ing, volume 3, pages 650–667.

LibLAS (2019). LibLAS - project homepage. https://liblas.

org/ (Accessed 28 Feb 2019).

Merwade, V., Cook, A., and Coonrod, J. (2008). GIS tech-

niques for creating river terrain models for hydrody-

namic modeling and ﬂood inundation mapping. In

Environmental Modelling and Software, volume 23,

pages 1300–1311.

Mukupa, W., Roberts, G. W., Hancock, C. M., and Al-

Manasir, K. (2017). A review of the use of terres-

trial laser scanning application for change detection

and deformation monitoring of structures. Survey Re-

view, 49(353):99–116.

OGC (2019). Point Cloud Domain Working Group - project

homepage. http://www.opengeospatial.org/projects/

groups/pointclouddwg (Accessed 28 Feb 2019).

OPALS (2019). OPALS - project homepage. https://geo.

tuwien.ac.at/opals/html/index.html (Accessed 28 Feb

2019).

Oracle (2019). Oracle - documentation. https://www.oracle.

com/technetwork/database-options/spatialandgraph/

documentation/spatial-doc-idx-161760.html (Ac-

cessed 28 Feb 2019).

PCL (2019). Point Cloud Library - project homepage. http:

//pointclouds.org/ (Accessed 28 Feb 2019).

PDAL (2019). Point Data Abstraction Library - project

homepage. https://pdal.io/ (Accessed 28 Feb 2019).

Pointcloud (2019). Pointcloud - project homepage. https:

//github.com/pgpointcloud/pointcloud (Accessed 28

Feb 2019).

PostGIS (2019). PostGIS - project homepage. https:

//postgis.net/ (Accessed 28 Feb 2019).

PostgreSQL (2019). PostgreSQL - project homepage. https:

//www.postgresql.org/ (Accessed 28 Feb 2019).

GISTAM 2019 - 5th International Conference on Geographical Information Systems Theory, Applications and Management

162

QPALS (2019). QPALS - project homepage. https://geo.

tuwien.ac.at/opals/html/usr qpals.html (Accessed 28

Feb 2019).

Rapidlasso (2019). Rapidlasso - homepage. https://

rapidlasso.com/lastools/ (Accessed 28 Feb 2019).

Rusu, R. and Cousins, S. (2011). 3D is here: Point Cloud

Library (PCL). In IEEE International Conference on

Robotics and Automation (ICRA, Shanghai),9-13 May

2011, DOI: 10.1109/ICRA.2011.5980567.

Tang, P., Huber, D., Akinci, B., Kipman, R., and Lytle, A.

(2010). Automatic reconstruction of as-built building

information models from laser-scanned point clouds:

A review of related techniques. In Automation in Con-

struction, volume 19, pages 829–843.

Tessema, L., Jaeger, R., and Stilla, U. (2019). Extraction

of IndoorGML Model from an Occupancy Grid

Map Constructed Using 2D LiDAR. In German

Society for Photogrammetry, Remote Sensing and

Geoinformation, 39st Conference. Thomas P. Kersten.

https://www.dgpf.de/src/tagung/jt2019/proceedings/

proceedings/papers/21 3LT2019 Tessema et al.pdf.

Van Oosterom, P., Martinez-Rubi, O., Ivanova, M.,

Horhammer, M., Geringer, D., Ravada, S., Tussen, T.,

Kodde, M., and Goncalves, R. (2015). Massive point

cloud data management: Design, implementation and

execution of a point cloud benchmark. volume 49,

pages 92–125. Computers and Graphics.

Voegtle, T., Steinle, E., and Tovari, D. (2005). Airborne

Laserscanning Data for Determination of suitable ar-

eas for Photovoltaics. In ISPRS Workshop Laser scan-

ning 2005, volume 23, pages 12–14.

Wilkening, J., Kapaj, A., and Cron, J. (2019). Creating

a 3D Campus Routing Information System with

ArcGIS Indoors. In German Society for Photogram-

metry, Remote Sensing and Geoinformation, 39st

Conference. Thomas P. Kersten. https://www.dgpf.de/

src/tagung/jt2019/proceedings/proceedings/papers/

11 3LT2019 Wilkening et al.pdf.

Wilkening, J., Schaeffner, R., and Staub, T. (2018). Interac-

tive 3D route planner for the Campus Roentgenring in

Wuerzburg. volume 4, pages 35–41. AGIT Journal.

Zhou, Q., Park, J., and Koltun, V. (2018). Open3D: A Mod-

ern Library for 3D Data Processing. CoRR.

3D Point Clouds in PostgreSQL/PostGIS for Applications in GIS and Geodesy

163