The transition to a fully electrified energy grid in the Netherlands is essential for the country's energy transition strategy. However, the existing grid infrastructure struggles to support the increased demand due to electrification, leading to congestion and potential outages anticipated before 2030. Microgrids, decentralized networks where generation, heat, storage, and consumption are coordinated, offer a promising solution by providing the required flexibility for the integration of Distributed Renewable Energy Resources (DERs) and potentially reducing energy losses associated with long-distance energy exchange.

This study explores the implementation of Deep Reinforcement Learning (DRL) in a district-level microgrid in the Netherlands. The primary objective is to minimize grid connection dependency and reduce energy exchange distances within the microgrid cost-effectively, thereby enhancing grid reliability and sustainability, answering the research question: \textit{Main Research Question}: \emph{How can Distributed Energy Resources be deployed and managed in a cost-effective manner within an electrified microgrid to achieve a balance between energy consumption and production in order to minimize the burden on the central grid given fluctuating demand?}

A comprehensive framework was developed offering a structured approach for simulating and optimizing microgrid scenarios, providing practical solutions and frameworks for the efficient management and deployment of microgrids. The study demonstrates that DERs can be effectively deployed and managed within an electrified microgrid through a combination of simulation techniques with geographical data input, a DRL model with a Deep Q-Network (DQN) that controls the system, and dynamic mappings and visualization. While the DQN demonstrated its potential in optimizing microgrid configurations, challenges were noted due to the large action spaces required for routing optimization. To overcome these limitations, future work is proposed to combine DRL with Graph Neural Networks (GNNs). This novel method shows promise in enhancing the scalability and efficiency of the optimization process, enabling distance and routing optimization without an aggregated model.

Scenarios were created to test the influence of different components, demonstrating how various factors affect overall system performance. The results of the model were as expected, and undertakings to optimize the model further led to significant improvements in the variability of the learning curve and of the rewards.

The study concludes that the developed pipeline contributes to existing works by offering a structured approach for simulating and optimizing microgrid scenarios and testing different setups. While this research approach has proven effective, future work should enhance the realism of components and address some computational limitations encountered. Refining these methodologies by incorporating GNNs, integrating real-world data, and extending the model with additional algorithms for performance comparison will further advance the field. This effort ultimately supports the vision of a reimagined, sustainable, resilient, and efficient energy grid.

Semantic 3D city models are essential for visualising, analysing, and managing the built environment. CityGML is an international standard for representing 3D spatial information with a Unified Modelling Language (UML)-based data model to address data heterogeneity and facilitate data exchange. Common formats for encoding CityGML data include Extensible Markup Language (XML), CityJSON, and relational databases, with PostgreSQL preferred for its spatial data management capabilities enhanced by PostGIS.
Although relational databases like 3D City Database (3DCityDB) support CityGML v.1.0 and 2.0, challenges remain due to complex schemas and the need for advanced Structured Query Language (SQL) knowledge to access nested features and attributes of the encoded CityGML data, especially through Geographical Information System (GIS) software like QGIS, which is widely used by Architecture, Engineering and Construction (AEC) professionals. To address these issues, the 3DCityDB-Tools plug-in for QGIS (plug-in) developed by the 3D Geoinformation group at TU Delft simplifies interactions with 3DCityDB-encoded data by providing a user-friendly QGIS interface, enabling the creation of GIS layers composed of unique feature geometries and associated with attributes. With the release of CityGML v.3.0 in 2021, 3DCityDB is being updated to version 5.0, requiring corresponding changes to the plug-in for compatibility.
This thesis investigates the changes in the CityGML spatial concepts and the differences in 3DCityDB encoding. The methods are derived based on the 3DCityDB v.5.0 structure, which consists of schema-wise scans for checking the existing feature geometries and attributes. The scan results are then stored in the metadata tables for users to select the desired feature geometries and attributes for generating GIS layers. In the implementation, feature geometries are determined by the inherited fixed spatial properties of space or boundary features. In contrast, the feature attributes are classified into four types: ”Inline-Single”, ”Inline-Multiple”, ”Nested-Single” and ”Nested-Multiple” according to the modified 3DCityDB encoding. Each type requires specific flattening (linearisation) strategies to be joined with the geometries. Finally, users can generate GIS layers by joining the queried feature geometries and attributes. Several query time performance tests are conducted to determine the method for storing query results and creating the layers.
The generated GIS layers demonstrate flexible access to the feature geometries and attributes with enhanced attribute management. The attribute flattening method facilitates the consumption of complex attributes, making them accessible for batch querying in QGIS. While direct editing of geometries and attributes in GIS layers is not yet supported, these advancements increase the usability of CityGML data. Coping with the XML complex feature schema is a persistent technical challenge in the GIS applications; the proposed approach provides promising alternatives that align with the ongoing development efforts in the QGIS community, offering a complementary pathway for handling complex geospatial data.

Global Navigation Satellite System is a spatial data acquisition technique, mostly used in navigation and positioning. One of the main components of this technique is the satellite visibility, which refers to the connection between the satellite and the ground receiver. It is known that the GNSS positioning systems are not as performing in urban areas due to the dense coverage of obstacles (buildings, trees, high terrain etc.). These obstacles can obstruct and reflect the lines of sight between the satellite and the ground receiver which can affect the quality of the performance of the GNSS service. The geometry configuration of the satellites above the receiver is another important aspect that has to be taken into consideration.

This research focuses on implementing a simulation similar to that of GNSS mission planning tools, but using point cloud data as the 3D representation of the surroundings of the receiver and using only the GPS constellation of satellites. Due to the large size of a point cloud sample, two visibility algorithms have been implemented to filter the necessary 3D data. The main output of the simulation are the dilution of precision values which give further information about the satellites' positions. The main purpose of this research is to understand the dilution of precision values, which are directly related to the geometry of the satellite configuration above the receiver. Understanding the behaviour and how the receiver's environment influences the DoP values can result in leading GNSS surveying missions with better results.

This output is then compared with the data acquired from a GNSS receiver in a real scenario. While the results are not favorable for the implemented simulation, it gives a better understanding of the surroundings of the receiver's location by using point cloud data than the already existing online GNSS tools.

Water management is an integral part of Dutch history, driven by the continuous need to reduce flood risk. Because a large area of the country is located below Normaal Amsterdams Peil (NAP), there is an ongoing challenge to safely discharge all the water to the sea. Therefore, flood safety policy has become crucial to protect the Netherlands from natural hazards. An essential part of this strategy involves the Waardegedreven Onderhoudscontract Uiterwaarden (WOCU) Rijntakken project, which is responsible for managing the floodplains adjacent to the Rijntakken within the Netherlands.

The current lack of efficiency and effectiveness regarding change inspection in the large and sometimes inaccessible areas of the floodplain requires the use of remote sensing change detection to move toward a data-driven maintenance process, in particular, by using point cloud data. This is nowadays a widely used data source in a variety of fields to capture elevations and in this way extract valuable information from terrains. Despite its usage in a variety of applications, the data is often underused since the data is frequently processed directly to other data formats. This research therefore aims to reveal the potential of explorative point clouds in floodplain maintenance.

Light Detection and Ranging (LiDAR)- and multispectral data were acquired at two moments, one before and one after the summer, with a time interval of 45 days. Subsequently, these acquired datasets evolved into an explorative point cloud by adding attributes, including vegetation health, also known as Normalized Difference Vegetation Index (NDVI), and the distance between these two point clouds, the cloud-to-cloud distance. This explorative point cloud with the integrated additional information was visualised to several disciplines involved in the WOCU project. This was done in Three Dimensional (3D) by using Virtual Reality (VR). This collaborative approach revealed the potential use cases of the Red, Green, Blue (RGB), cloud-to-cloud distance, and NDVI point clouds highlighting the potential of explorative point clouds.

Potential use cases that were found are; highly detailed area modeling, vegetation overgrowth monitoring, bank erosion detection, flora status assessment, monitoring of vegetation types, digital inspection of remote sites, participation medium, and identification of atrophied ground patches. Attributes added to point clouds enhanced insights. Especially the RGB point cloud sparked excitement due to its realistic appearance. The Cloud-to-Cloud Distance (C2CD) attribute showed potential, especially for erosion detection. However, due to the short timeframe between measurements, it could not be detected. The NDVI attribute was perceived as less interesting.

The use of explorative point clouds, generated from raw LiDAR point cloud data, offers potential uses and insights for floodplain maintenance. The interdisciplinary value of explorative point clouds was clearly visible. This thesis emphasizes that underused raw LiDAR data, by making it explorative, can act as a valuable resource.

Geometric algorithms are usually designed based on the assumption of using arbitraryprecision representations. However, in modern computer systems floating-point arithmetic and finite-precision approximation are often used as implementing exact computations would require large computational resources and would be rather slow in the applications. Snap rounding (SR) is a widely used technique to transform representations of infinite precision into fixed-precision formats. It slightly modifies the original geometry of the input in order
to obtain a better organized geometric object arrangement and avoid the issues caused by applying floating-point arithmetic in geometric algorithms. The existing implementations of SR primarily focus on processing sets of line segments, while in practice polygons are of great significance as well yet has not received adequate attention so far. In this thesis, a new method is proposed to perform SR on 2-dimensional polygons. Firstly the input polygons are embedded into a triangulation. Subsequently, the triangulation is tagged with the ID information of polygons and the polygon boundaries are extracted and
stored in a container. The boundaries and the triangulation are then dynamically processed according to the identification of close polygonal vertices and close vertex and boundaries (under a predefined tolerance). After having snapped all the close elements the rounded polygons will be reconstructed from the processed boundaries. The testing results show the proposed method is capable of eliminating small gaps between elements (i.e. vertices and
boundaries) and is adaptable to various inputs. The topological and geometrical properties of polygons are preserved as much as possible. Finally, an overview of the limitations is provided, along with potential directions for future research.

The present report is the end result of the project that was carried out as part of the Geomatics Synthesis Project in cooperation with AllMaps, an open-source platform dedicated to the viewing and georeferencing of historic maps. The main objective of the project was to automatically georeference historic map series curated and digitised by the Dutch National Archive. This was based on the  corner coordinates of the map sheets. The first issue that had to be tackled was the reprojection of the original coordinates which were in Bonne projection
to WGS84 coordinates. To determine the corners of the map content within the sheets two methods were implemented. The first one detects the lines based on HoughLines Probabilistic Transformation and the second one detects lines based on the distribution of black pixels in the rows and columns of the images. In addition to map sheets with corner coordinates, there are two other sets of images which were georeferenced utilising a convolution neural network that performs feature matching. The feature matching was performed by running the two sets of images against the georeferenced sheets with known corner coordinates. To minimise the search space for this process a geocoder was used to determine the approximate location of the image. The implemented methods appear to hold the potential for georeferencing old map series. It is worth noting that the developed algorithms, while effective in many cases, may encounter challenges when dealing with irregularities on map sheets caused by the passage
of time, such as damage. Consequently, there is a great opportunity to further enhance the algorithms to ensure they can consistently and accurately georeference images, even when faced with such irregularities. This ongoing development will lead to improved georeferencing accuracy and user confidence.

Urban areas are home to over half of the global population and consume significant natural resources, necessitating effective city management for urban resilience and sustainability. Resilient cities are better prepared to face challenges, from natural disasters to economic downturns, and adapt to climate change, resource scarcity, and population growth. Urban digital twins, which are virtual replicas of cities integrating data from various sources, play a pivotal role in enhancing and supporting urban resilience. These digital twins empower urban planners, designers, and the public to make informed decisions, identify vulnerabilities, and respond to acute shocks in real-time. Leveraging technologies like the Internet of Things (IoT) and 3D city models, urban digital twins offer a dynamic representation of the city’s functions, allowing for better visualization, monitoring, and decision-making. However, challenges related to interoperability and data acquisition need to be addressed to maximize the potential of urban digital twins.

This thesis investigates how environmental sensor data can be collected, processed, and integrated into 3D city models to visualize the dynamic elements of a city comprehensively. It explores the use of international standards, such as the OGC SensorThings API standard to acquire, store, and manipulate real-time and historical crowd-source, sensor data in a consistent and interoperable manner. The concepts of sensor data pre-processing and interpolating are also explored through OGC standards, in the context of providing detailed spatio-temporal information for the urban environment, while the results are visualized in a 3D web application.

Today, in the urban planning, energy modelling and other fields, semantic 3D city models are used in various applications like visualization, data exploration, analysis and more. As a result, standard data practices needed to be set in order to facilitate the storage and exchange of these city models. For this purpose, the Open Geospatial Consortium ( OGC ) adopted CityGML as an international standard for effective use of 3D city models. Generally, the models are encoded in Extensible Markup Language ( XML ) files, however, other file encodings can also be used like JavaScript Object Notation ( JSON ) files with CityJSON. Moreover, CityGML can also be adapted for a database encoding like the 3D City Database ( 3DCityDB ), on which this thesis is based upon. The benefit of using a database encoding is that databases are built to handle and organize large amount of data, which 3D city models usually consist of. The 3DCityDB is an open source project developed for PostgreSQL and Oracle databases. It is supported by other software in the 3DCityDB suite that facilitate its use in different applications. The 3DCityDB tries to simplify the complexity of CityGML, however, its approach remains difficult for users to access data directly without technical knowledge of databases, Structured Query Language ( SQL ), CityGML and/or 3DCityDB structure. Derived frm thios limitation, the primary objective of this research is to develop an approach that could simplify user interaction with the 3DCityDB from within a Q Geographical Information System ( QGIS ) environment. To achieve this, ”3DCityDB-Loader”, a QGIS plugin, is developed to handle complex server operations in the background, whilst providing a user-friendly workspace environment. The complete functionality of the plugin is segmented into client and server-side parts. This thesis focuses on the client-side development but both parts were jointly developed in a common iterative process of requirement identification, development, testing and assessment. The most important requirements for the plugin is to have layers that can interact with 3DCityDB data, be able to work with multiple users with different privileges, allow for multiple scenarios (database schemas), allow to edit attributes, handle different Levels Of Detail ( LOD ) and geometry representations and finally be able to operate from a Graphical User Interface ( GUI ) in QGIS . Regarding the client-side part of the plugin, it can manage database connections, manage the server-side installation, manage and create layers for multiple scenarios from a GUI , include CityGML generic attributes, enumerations and codelists and automatically set their relations, automatically structure a hierarchical QGIS Table Of Contents ( TOC ) and finally automatically apply standard colors on different features. At the time of writing this document, the plugin is at version 0.4. The limitations are mostly related to functionalities that are not yet supported, with future development being tracked from the project’s GitHub repository. All in all, ”3DCityDB-Loader” facilitates the use of 3DCityDB for users of different fields and expertise with the common denominator being the well-accustomed QGIS environment.

Nowadays, the military stakeholders take advantage of the various geospatial infrastructure and technology while exploiting the wide use and distribution of geographic data, to model military scenes and/or conduct geospatial analysis for military operational scenarios. While a host of technology is offered, in potential joint military operations or joint civil-military operations, several difficulties elapse in geospatial communication, due to different Coordinate Reference Systems (CRS), non-conforming resources, different data formats and different areas of responsibility perception between the stakeholders. The multidisciplinary geospatial communication, data integration and a common modelling geospatial framework is integral for the accurate and fruitful modelling of a military scene assisting in joint operations. Discrete Global Grid Systems (DGGS), while not being a new concept, recently emerged in the geospatial community with various distributed implementations, as a framework to integrate, analyze and manage geospatial data. This thesis attempts to apply a DGGS, using an existing distributed implementation, to integrate different format, two dimensional and elevation geospatial datasets, coming from military and civilian stakeholders, to identify the potential of this approach for military scene modelling. The different approaches of quantization of the datasets are explored for their efficiency, quality and the ability to assist in performing geospatial military analysis, while according APIs are developed for the DGGS conversion. Given the model of the military scene the research tackles the geovisualisation alternatives and a case study is also conducted for a military geospatial operation (ranging) using the DGGS approach. The results show the promising potential of the DGGS approach to model military scenes, providing a uniform area based framework, encapsulating the different qualitative and quantitative military information, offering strong connectivity and hierarchy relations and increasing the qualitative spatial perception and multidisciplinary geospatial communication. The datasets’ DGGS conversion showed advanced integration, segmentation, aggregation and visualization capabilities, also exploiting the 3rd dimension data. Uniform storage through the DGGS integration can be realized in a database, facilitating the distribution of military data. A DGGS can also support common military geospatial operations (i.e. ranging). However, the quality of the results is highly dependent of the original data accuracy/spatial uncertainty and the demanded precision requirements of the scene’s coverage. In parallel, due to the fact that this technology is still emerging and dynamically optimized, it is not mature yet to handle high precision requirements and be treated as a frame with high positioning accuracy for high scale military scene coverages.

There has been an ever increasing focus on the development of smart cities, both by big corporate in the field of Information Communication and Technology as well as respective governments responsible for these cities worldwide. This in combination with the influx of easily accessible data owing to cheaper and efficient sensors has risen to the concept of digital twinning of urban spaces and cities.

Urban digital twins require the city to be represented in 3-dimensional digitally, these datasets are oftentimes too large to be provided on as-is basis for most platforms. To tackle this problem of large 3D datasets, the OGC 3D tiling specifications were formed and accepted in 2019.

This thesis attempts to develop a neighbourhood designing platform which would enable users to design neighbourhoods in 3D while being able to experience the design in an immersive manner by utilizing virtual reality’s capability to observe 3D assets in scales which are not possible using a physical model or a 2 dimensional neighbourhood plan, employing the OGC 3D tiling technique to understand the extent to which a combination of existing 3D datasets can be made with a user-interactive neighbourhood designing platform in a VR environment.

The results of the study reveal that using our methodology, it is possible to combine existing 3D datasets with user made neighbourhood design with their own 3D assets of any level of detail and furthermore, it is possible to utilize instanced tiling format to disseminate the neighbourhood design as a standard 3D tiled design.

With regard to climate change and air pollution within cities, interest in sustainable modes of transportation for regular use has taken a rise. Utilitarian cycling is being seen as a frontrunner for replacing everyday motorized travels within and between cities, supported by the rapid emergence of the electric bicycle. Governments are trying to use the increasing opportunities involving bicycle transportation to reduce car traffic and the related air polution, by stimulating the use of bicycles. In this light, interest is drawn to cyclist travel behavior to uncover preferences of cyclists. Existing literature shows a significant impact of weather conditions on cyclist travel behavior in terms of tranportation mode choices. Especially adverse weather condtions leave their mark on the use of bicycles as a means of transportation, as it is recognized by many studies as a main deterrent for cycling. On the other hand, the relation between weather conditions and cyclist route choice is an underexplored topic in existing literature. Consequently, it has remained unclear to what extent cyclists attempt to mitigate the influence of weather condtions through choice of route, and based on which determinants. Insights in ways to mitigate unchangeable external circumstances like weather conditions could be another step forward in stimulating utilitarian use of bicycles in the search for transportation modes that can replace motorized trips. This thesis made an attempt to partially address the research gap in existing literature, by departing from findings in the field of pedestrian mobility. In these studies, pedestrians have been found to adapt their choice of route to the degree of shelter that is offered by the built environment as a measure to change level of weather exposure. These findings were projected on cyclist route choice, to evaluate to what extent cyclist alter their choice of route based on weather conditions and the degree of shelter that can be found within a built environment. An elaborate methodology was proposed in which observed routes throughout the study area of Tilburg (the Netherlands), comprising trips made with conventional and electric bicycles, were compared with shortest and fastest alternatives. The weather conditions under which a route was conducted were modelled through a set of individual meteorological factors, spatially related to the location of an observed route. To operationalize the degree of shelter provided by the built environment, a new method was developed using aspects from existing theories on street climate design and spatial openness in order to provide a detailed description of the potential shelter along a route. Three different shelter factors were established, describing the degree of mean building shelter, maximum building shelter, and vegetational shelter in the form of tree density along a route. Through estimation of a set of linear regression models, independent and combined effects of the meteorological and shelter factors on cyclist route choice were modelled. Initial moderate influences of windspeed, temperature, and cycling under twilight conditions on the choice of route were found, while cyclists generally chose routes with a lower degree of building and vegetational shelter compared to alternative shortest and fastest routes. Interactions between the effects of meteorological and shelter factors showed very limited additional effects, suggesting that utilitarian cyclists in the study area did not value the degree of built environment shelter along a route sufficiently as a mitigator of weather conditions to diverge from the shortest or fastest route. These findings imply that built environment shelter does not have to be accounted for in policy design to stimulate utilitarian cycling.

Indoor localisation is a highly relevant topic. It can be used for many applications,
such as indoor navigation. Current indoor localisation approaches all have certain
downsides. In this report, the results of a completely new indoor localisation approach are described. The aim of this approach is to perform indoor localisation on room level based on a fingerprint solution using point clouds of the ceilings. The ceiling is used, because the ceiling does generally not change much and therefore it is easier to keep an up-to-date database. This research considers both the use of Dense Image Matching (DIM) input from pictures or videos made with a mobile phone and Light Detection And Ranging (LiDAR) input.

With the rapid growth in point cloud acquisition technologies the recent years we have the ability to measure large quantities of 3D points of significantly detailed and geometrically composite scenes such as urban environments. This advantage can be exploited and used for direct analysis on point clouds. A direct point cloud analysis has several advantages over for example 3D surface reconstruction, such as the end result having more details and the computation being less expensive. In order to make a point cloud representation a suitable alternative for other types of 3D city models, they need to be semantically enriched, resulting in a rich point cloud. One element of this enrichment is the detection of objects, such as windows. Extracting these from facades is specifically what this research revolves around, which can be done by taking advantage of the fact that they show up as holes, since lasers of the point cloud scanner do not properly reflect on them. Two different general approaches are taken to detect windows in a by mobile laser scanner obtained point cloud of Noordereiland, Rotterdam, The Netherlands.

During the last three decades, the Earth’s climate is changing rapidly, with higher average temperatures every year that leads not only to the melting of the ice sheet in the arctic and on most of the glaciers all over the world but also to extreme weather phenomena. The rise of the temperature can affect the Sea Surface Height (SSH) in more than one way, and since 70% of the Earth’s surface is covered by the oceans, if the oceans are being affected then the whole Earth also is. The monitoring of the SSH can help the scientist predict the changes that will take place in the future. The SSH is a dynamic phenomenon that constantly changes not only within different decades but also from year to year, month to month even within the same day. These changes are the result of various phenomena and are called anomalies. When the SSH is monitored different phenomena are represented in different time scales and it is important to be taken into consideration if there is the need for a proper understanding of the SSH phenomenon. Many spatial data vendors are providing a large number of data-sets related to the monitoring of SSH and its anomalies and as a result, there is the need to find the most effective way to extract information from the data. Over the years has been established that one of the most effective ways to extract information from data is through the various visualization techniques and since the data of SSH is mainly spatial the main visualization technique is cartography. The advancements of the technology over the last couple of decades have led to a reality that the ”online” application is the norm and consequently the web mapping and web geographical information system (GIS).
The goal of this thesis is to propose an architecture for a web GIS application that will be able to visualize dynamic data while adding elements of interactivity to improve the chances of the Sea Surface Height Anomaly (SSHA). IN order to achieve the goal of this thesis three main research questions need to be answered: What type of animation should be used (in order to visualize the passing of time), what interactivity elements should be added (e.g. zooming/panning in space and time) and what system’s architecture is optimal for such application (server-side/client-side etc.). This document is providing guidelines on how to create such an application and is resulting in the production of a prototype. The first part of this thesis is the review of the main ideas that are introduced in this project and how they were implemented by other researchers. Then, a comparison between the different implementation techniques (for every research question) is taking place to determine the main characteristics of the application. The final part is related to the implementation of the chosen techniques that lead to the development of the prototype application. The resulted prototype even though it is not perfect, due to technical limitations that were a consequence of implementing some of the most recent concepts in web development, is functional and paves the way for the development of new improved dynamic/interactive web GIS applications (https://giorgosdimo.github.io/MSc-Thesis/).

To obtain 3D information of the Earth’s surface, airborne LiDAR technology
is used to quickly capture high-precision measurements of the terrain.
Unfortunately, laser scanning techniques are prone to producing outliers
and noise (i.e. wrong measurements). Therefore, a pre-process of the point
cloud is required to detect and remove spurious measurements. While outlier
detection in datasets has been extensively researched, in 3D point cloud
data it is still an ongoing problem. Especially, clustered outliers are hard to
detect with previous local-neighborhood based algorithms.
This research explores the possibilities of using a voxel-based approach to
automatically remove outliers from aerial point clouds. A workflow is designed
in which a series of voxel-based operations are integrated, with the
aim to detect all types of outliers and minimize false positives. Voxels can
be processed more efficiently than 3D points for two reasons: (1) A voxelgrid
can be analyzed using efficient image processing techniques; (2) Voxels
group inner points before feature extraction using neighborhood operators.
Outliers are detected in two steps. First, the source point cloud is voxelized.
Secondly, outliers are detected by computing connected components and labeling
voxels not connected to the largest region as outliers. Simultaneously,
analysis of the point’s local density, shape (planar) and intensity minimize
classification of false positives.
The presented algorithm generally detects outliers with a higher accuracy
than previous local neighborhood-based methods. A comparison with an
existing approach shows that more outliers are detected. Above all, clustered
outliers are removed. However, some issues can still be improved.
First, more research is necessary to classify outliers based on non-arbitrary
decisions. This could potentially be improved by introducing supervised
learning algorithms. Secondly, more attention is required to process massive
point clouds that do not fit in internal memory. This study proposes a
possible streaming solution.