Over the years, the pace at which data is generated keeps on increasing. As a consequence, the data itself no longer holds the highest value, but rather the information and context the data captures are. This principle also holds in the 3D environment modelling scene, as accurately depicting an environment holds more value than the number of models there are of it.

One of the major problems in 3D environments, especially when the environment represents a building, is the presence of glass. A lot of the data captured to model these 3D environments is captured using LiDAR laser scanning. This is where glass becomes a problem as glass is almost completely transparent to laser beams at the typical wavelengths used when using LiDAR laser scanning. As a consequence, glass can lead to problems with navigational routes as it is invisible in the environment but still blocks the path. It can also create false spaces in the captured environment as it can also partially act as a mirror reflecting the laser beam and showing these reflections in space as if they were captured in a straight line.

Alternative manners for capturing and identifying glass in environments captured with laser have been created over the years, but they often need a dedicated set-up, expensive equipment or a lot of data. These solutions are not always feasible for users of point cloud data.

Therefore, in this thesis a focus is put on how can a low entry solution be created for this problem, which leads to the main research question: How can the location of glass be deduced using only information acquired from 3D point clouds and a reference position?

To answer this question, this thesis focuses on the deduction of the locations of glass windows in the provided input. To find these a projection from 3D data to 2D is performed. In 2D image space, contours are then detected that match the criteria of window frames. These contours are then used to segregate parts of the 3D point cloud that should contain the window detected in the projection. After clustering these parts and the best matching cluster is deduced to be a window.

In this thesis, it is shown that using the proposed methodology it is possible to deduce the location of glass in a LiDAR point cloud using only an additional reference position, but there are some flaws with the simplified input of the method.

Unlike outdoor environments, there is no wide-spread solution to positioning inside a building. Indoorsolutions rely on pre-installation of infrastructure, such as Bluetooth beacons or ultra-wide bandtechnology. Recently, there has been growing interest in the use of Augmented Reality (AR) for indoorpositioning. AR devices use an algorithm known as Simultaneously Localisation and Mapping (SLAM)to scan an environment and find a position inside. This make it possible to estimate a position on-theflywithout any pre-installations. The Microsoft Hololens (MH) is a head-mounted display AR devicethat is able to perform SLAM. The use of SLAM for indoor positioning can be beneficial in the case ofEmergency Response (ER). The place of impact in ER is unknown and there is no time to install anyinfrastructure beforehand.Two problems arise with the use of SLAM for indoor positioning. (1) A position is a pin point in spacethat is defined by a reference frame. In the case of SLAM, the frame is the scanned object. In mostsituations, a map or floor plan is more appropriate as reference frame, in order to give a full context.(2) The SLAM algorithm suffers from drift, a growing error over time. This research tries to solve theseproblems using shape registration. The SLAM output can be aligned to a reference floor plan, by use ofa spatial matching technique. This alignment is a continuous process to account for the drift errors ofthe SLAM algorithm.Three spatial matching techniques are compared: Iterative Closest Points (ICP) iteratively tries to minimizethe distance between two shapes using least squares; Instantaneous Kinematics (IK) is a varianton ICP that makes use of a velocity vector; and Hough Transform makes use of the Hough domainproperties, where rotation is invariant to translation and scale. These algorithms are compared on theiraccuracy, computation time and robustness.It is concluded that the Hough Transform algorithm gives the most accurate results and is fastest. In thisresearch it was found that an average accuracy of < 1m can be maintained over 80% of the experiments,with maximum errors up to 5 meter. In 20% of the experiments, the error can extremely diverge upto >100 meter. It is suggested that the quality of the scan and the existence of artefacts (e.g.furniture,people) are the cause of these errors. That makes the method unsuitable for indoor positioning in thecase of Emergency Response, that needs extremely reliable systems. Research is needed to create a morerobust method, that uses better outlier detection methods. However, the results are promising and doopen the door for indoor navigation using Augmented Reality.

Navigation from a room inside a building to another room inside a building which is across the street consists of three parts: first, the indoor part at the building where you start your journey. Secondly, the outdoor part and thirdly, another indoor part inside the building of destination. As regards to the outdoor environment, a navigation aid is well used and implemented in many types of applications. However, it is not common to use an aid to navigate inside a building. While such an indoor navigation aid is not necessary in small buildings, it is a necessity in more complex buildings like hospitals, airports, conference venues and large shopping malls. The indoor navigation aid can help visitors finding their way inside these, for them, unknown locations. The systems behind a navigation aid consist of several elements like an indoor positioning system, an indoor navigable map, specific destinations (points of interest) and an appropriate guidance throughout a building. This research focusses on the creation of indoor navigable maps that can be displayed and used to plan possible routes throughout the entire building. The indoor environment is far more complex than the outdoor environment. First, most people lose their orientation inside a building after they change their direction several times. Second, since there are no pre-existing routes inside a building, there are many different possible ways to arrive at a destination. Third, there is a large variety of associated spaces which all have their own unique interior design. Therefore, automating the process of making an indoor map is more challenging and time consuming than generating an outdoor map.

Most research in the field of automatic generation of indoor maps is focused on the already available 2D floorplans and only few of them use the more complex 3D representations. Using a 2D floorplans for the purpose of indoor navigation has its limitations because of various factors. Firstly, the 2D maps are already a simplification of the complex 3D environment which can lead to difficulties in representation. Secondly, the connectivity between different floor plans can be difficult as each floor plan is a separate entity. Thirdly, the maps that are available do not always contain furniture. Fourthly, in addition to the third one, most existing methods only focus on reconstructing the indoor space as an empty hull which results in a navigation aid which has does not emphasize the attention to obstacle detection. At last, most floor plans are out of date because some buildings are not built according to their blue prints. Their interiors might change after several years through the modification of walls and doors and furniture may be repositioned to the users preferences. Therefore, information about the indoor environment must be updated in most cases. This research concentrates on the automatic generation of indoor navigable spaces for pedestrians based on laser scanning with a Mobile Laser Scanner (MLS) device. These devices scan the environment continuously along a trajectory which makes them more time efficient than terrestrial laser scanners.

To aid pedestrians in their indoor navigation, features needed for path computation such as floors, stairs, walls and furniture elements, need to be identified. These must be extracted from the point cloud which is generated by the MLS. How to identify these elements, like walls and doors, is investigated a lot. These researches are built on a set of constraints, like a Manhattan World or a flat surface constraint. These constraints are not problematic for a regular office building but they will provide difficulties in more complex buildings. This means that it is important to focus on a method with less or without constraints.

Scanning the indoor environment of an building often happens during business hours which automatically leads to the inclusion of dynamic objects like pedestrians or small vehicles in the final point cloud. These dynamic elements do not represent any type of building elements (like furniture) and thus need to be identified and removed.

Beside the point cloud, the MLS also stores the trajectory of the MLS device. This trajectory contains three types of valuable information. The points directly below the trajectory indicate areas where pedestrians can walk, since the MLS device was operated by a pedestrian. The height difference between neighboring trajectory points can be used to detect stairs, slopes and flat surfaces and the trajectory also provides information about the connection of different surfaces and represents the complexity of the building.

In this research, a method for the identification of walkable surfaces based on the analysis of a point cloud and the corresponding trajectory of the MLS is developed. First, the point cloud is voxelized. Second, the trajectory is analysed to detect the three different types of navigation surfaces: stairs, slopes and horizontal surfaces. This classified trajectory is projected vertically on the voxel model to acquire seed voxels. These seed voxels are then used to create areas by using region growing. These areas can be modified by identifying dynamic objects, entryways and furniture elements so that each area represents a specific navigable voxel space inside a building.

Experiments shows that by applying this method, it is possible to create a continuous navigable space in buildings including several floors, stairs and elevations. Data can be captured during opening hours because the method detects and removes dynamic objects in the final result. The proposed method can be used for any type of room without any constraints because the complexity of the building is already present in the trajectory of the MLS.