Due to safety or preservation reasons, certain objects or areas need to be inspected regularly. Currently, the inspection of objects is mostly done in-person, which is labour-intensive and not very effective. With the use of drones, areas and objects can be inspected from new angles at a much faster rate. To effectively monitor these objects with drones, the position and location needs to be extracted from drone data in real time.
In this thesis, a case study is done on the localisation of fisher boats in restricted areas. Several components are integrated to create a prototype. The pretrained YOLOv3 detection model is trained on acquired nadir boat images, which makes it able to predict the bounding boxes of boats on images captured with drones. A positioning algorithm is constructed, which calculates the geographical coordinates from the pixel coordinates for images taken both in a nadir and an oblique angle. A real time connection is constructed between the drone and the prototype. This is done by creating a connection with Google Drive with the drone controller and the prototype. The positioned polygon bounding boxes are localised using a real time dashboard, which visualises the bounding boxes in a map with other relevant layers.
The results indicate that the performance of the components and prototype as a whole are satisfactory for this use case. To deploy this prototype in other object localisation use cases, it is recommended to train the pretrained model further, use a drone with more accurate equipment and run the prototype on the drone controller.

Understanding the characteristics of photovoltaic system (PV) adopters can help policymakers realise energy transition more effectively. In this study, we developed a model that predicts PV adoption per building using geometric and socioeconomic variables. Seven geometric variables were created by processing building registration data, airborne laser scanning data and 3D building models based on airborne laser scanning data. Additionally, eight socioeconomic variables were generated from building registration data and socioeconomic postal code statistics. The random forest machine learning model, which was trained and evaluated on 646 000 buildings in the province of Overijssel, The Netherlands, displays good overall performance with an AUC of 0.77. Moreover, the model demonstrates that buildings have an increased probability of PV adoption if they (i) have a suitable area above 30 m2, (ii) have a rooftop higher than 6 m, (iii) have a non-flat roof, (iv) were built after 1970, (v) only have one address registered and (vi) are used for residence. Similar experiments involving a different type of machine learning model (i.e. a neural network) and province (i.e. North Holland) yield similar results. Future improvements of the model could focus on increasing performance in residential areas and studying the effect of PV stimulation by including a temporal component.

This graduation project is part of the Shared Heritage Lab; a collaboration studio between the master tracks of Heritage&Architecture, Architectural Engineering, Landscape Architecture and Urbanism of the Faculty of Architecture and the Built Environment from the TU Delft. All students joining this graduation lab focus on the question ‘How to evolve important heritage structures and areas of Bandung, in order to realise inclusive, thriving, and healthy environments for working, living and leisure?’. Within the scope of this question the issues found in the Pasar Baru Area in Bandung are addressed in the research and the design of this graduation project. The area is located in the middle of the city centre, has always been one of the major trade hubs of Bandung and is the old Chinese and Arabic neighbourhood. During the Dutch colonialization of Indonesia, the area developed in an interesting spatial way, as the city blocks, are characterized by an ‘inner’ and an ‘outer world’. The outer world consist of the periphery of the city blocks, where the shophouses, owned by Chinese merchants, define the spatial and urban character of the area. Behind those shophouses, in the middle of the city blocks which is called the ‘urban pocket’, the local Sundanese community resided in their kampung houses. This urban development resulted in an interesting spatial, social and economical balance in this part of the city. Nowadays, however, this balance is under threat, because of commercial investments, densification and pollution. In order to stop the process of decay in the Pasar Baru area and to create a pleasant environment for working, living and leisure, the design aims to restore the former balance in the area and to create a comfortable place for all the inhabitants, traders and customers. In this design the values of the heritage, the community and the environment are taken into account. This results in an urban strategy focussed on the restoration of the balance between the ‘inner’ and the ‘outer world’ of the city block, whilst still allowing for further densification to happen in this urban context. These rules guarantee a comfortable environment to live, work and relax, for a diverse group of people and also stimulate the use of more traditional building techniques in the architecture, in order to create a new local atmosphere in the area, and to take the issues of climate change into account. This strategy is further elaborated into an architectural design for the urban pocket, which consists of an expandable kampung house based on easy-to-understand building methods using only timber and bamboo. This house allows for cheap and comfortable housing and guarantees a pleasant residential environment for all the residents of the urban pocket. In order to guide the process of densification and commercialisation in the area, the urban strategy is also focussed on the design of the architecture in the periphery of the city block. To provide space for small businesses and to reintroduce housing into the area, four existing shophouses are retrofitted into a mini-market for street vendors and small shops and on top of those a timber high-rise structure is designed to house small studio apartments. The flexibility and the climate adaptivity of the building is created by the design of several movable panels in the interior and in the façade. Felt and glass panels in the interior allow the residents to arrange their own privacy and comfort, whilst a system of louvre and woven bamboo panels in the façade allow for enough natural ventilation to create a comfortable indoor climate. The bridge between the urban pocket and the periphery is created by low-rise apartments for families, where a similar building system is applied. The use of plants and grasses in the roofs, the facades and in the public realm of the urban pocket, create a comfortable environment for all residents and finalise the strategy, where all the urban pockets together are the lungs of the Pasar Baru area, both metaphorically as physically.

Over the last decades, laser scanners are becoming more and more established for the acquisition of geo-information. Depending on the sensor platform where the laser scanners are mounted, there are MLS, ALS and TLS techniques for both indoor and outdoor environments. The high-quality 3D point clouds produced from laser scanners is an important source of 3D spatial information and it is increasingly used in a wide field of applications from engineering to medical modeling, gaming and agriculture. Although laser scanners provide dense and accurate point clouds, a scene coverage and the creation of a complete 3D representation requires multiple scans of the same area. The procedure integrating multiple scans does not always result to a perfect match. Errors that exist in the datasets or errors in the transformation of datasets create difficulties during the matching procedure. Matching data-sets with the best alignment is a topic that has been researched in many fields and a variety of methods have been analyzed for registering point clouds. A co-registration can be compared to a mathematical model and hence it is important to present not only the functional model and the set of functional relationships between the variables but also the stochastic model that describes the variability among the values and gives insights about the level of satisfying predefined demands. As a result, it becomes necessary to establish an evaluation of the output of a procedure and each result to be accompanied by its quality description. The aim of this graduation thesis is about presenting the stochastic model of a co-registration with a quality description of the result of co-registration between point clouds. The approach will be applicable to results acquired from an image-based registration and the goal is to implement a method in order to quantify the quality of the co-registered result of two different point clouds. A probabilistic analysis of the results of an image-based method results to a quantification of the quality of the output. Response images acquired from an image-based co-registration are tested and the quality indicators of precision and reliability are determined. Moreover, the shift parameter is also defined, fact that enhances the applicability of this research to co-registrations referring to big areas where multiple local registrations have to be combined in a globally consistent manner and each local result contributes to the global registration according to its quality indicators. Particularly, response images that have clearly defined peak areas and pass the tested criteria, are considered more reliable and contribute to the global registration with a greater weight. On the contrary, blurry response images, fail to satisfy the tests, are considered less reliable results and contribute with a smaller weight. The analysis of the images performed by testing the distribution of pixel values in both directions of the images. The implemented approach provides information about the precision and reliability of the images that have normally distributed values. Furthermore, by fitting a Gaussian line to the discrete pixel values, a sub-pixel accuracy of the result is achieved as values are generated in between the discrete pixels. The developed method quantifies the quality of an image-based registration but further improvements and investigation of the recommendations can also attach additional value.

The contemporary trend shows a shift from rule-based algorithms to deep learning. In the last few years, this field has been developing rapidly and its popularity has increased to a large extent. This happened for a good reason, since deep learning was able to solve some of the hardest problems in fields like image recognition, natural language processing and speech recognition. A large proportion of its success has to be credited to the explosion of big data in the past two decades. Structured data sets are essential to deep learning systems. The rising amount of automatic identification system data is a key example in the current big data boom. The automatic identification system produces spatiotemporal movement data of vessels. It was designed as a collision avoidance system, but researchers have been looking into ways to leverage its data for other tasks. Analyzing behavior in the movement data of vessels can help policymakers and monitoring operators with decision making processes. Improving these processes lead to safer and more resilient marine environments. Unfortunately, the possibility of applying deep learning on vessel movement data is an underexposed topic. This project attempts to explore the research gap in this topic. The objective is therefore to give an overview of the possibilities, complications and opportunities given the current state of the art. Ultimately, this project may serve as a rough guide for those who wish to explore the crossings where deep learning and vessel movement data meet.

Broken glass from a bus stop, litter on the ground, a broken lamppost, and more things like this can be encountered in a municipality. To improve the process of mapping the issues and fixing them Fixi is developed by Decos. Fixi is an application in which citizens can report issues in public space. A citizen can report the issue he or she encounters in Fixi and the handlers of the municipality will use this inventory to improve the quality of the municipality. The issues are mapped using GNSS or pin-pointing it out on the map, but these options are not always available. This thesis research explores object identification based on the description of neighbouring elements. By providing a method to identify the object by a description, an additional option for reporting issues is made available in Fixi. The neighbouring elements which are described by the user will function as reference points for the method. By combining the information, a suggestion of one or more objects can be presented to the user which should include the described object. The flow from description to output consists of five different elements: data input, preprocessing of the data, user input, processing, and output. Three versions of input and processing are developed and tested in this thesis research. (1) With spatial relationships and distances indicated by the user, (2) with spatial relationships, and (3) with distances indicated by the user. The type of elements that can be described by the user are based on the theory of Lynch [1960] and can be categorised in five different types: path, edge, district, node, and landmark. To test the method, the selectivity of the elements is tested and the output of the process is evaluated. This last test is done based on six criteria: (1) number of questions, (2) presence of object, (3) amount of suggested objects, (4) average distance to elements, (5) covered area of suggested objects, and (6) completion time of description. The tests have been executed on thirty scenarios spread over three municipalities: Westervoort, Amsterdam, and Joure. All scenarios presented the described object as one of the suggested objects, which means the object has been identified. In general scored version 3 with the distances indicated by the user the best results in the tested cases. Improvements are possible in elements like number of questions, average distance, and completion time.

Since GPS tends to fail for indoor positioning purposes, alternative methods like indoor positioning systems (IPS) based on Bluetooth low energy (BLE) are developing rapidly. Generally, IPS are deployed in environments covered with obstacles such as furniture, (partition-) walls, people and electronics influencing the signal propagation. The major factor influencing the system performance and to acquire optimal positioning results is the geometry of the beacons. The geometry of the beacons is limited to the available infrastructure that can be deployed (number of beacons, basestations and tags), which is dependent on the budget and deployment effort of the customer.
This leads to the following challenge: Given a limited number of beacons, where should they be placed in a specified indoor environment, such that the geometry contributes to optimal positioning results?

This challenge is approached by using theoretical design computations. The design computations require the definition of a chosen 3D space, the number of beacons, possible user tag locations and a performance threshold (e.g. required precision). For any given set of beacon and receiver locations, the precision, internal- and external reliability can be determined on forehand. The results of a given geometry have been validated by deploying an IPS of BlooLoc and comparing the observed precision with the modeled precision for a chosen set of user tag locations. The theoretical model showed a precision pattern with several equivalent precision patterns of the measured data, however, some significant differences could not be explained physically and the model has to be adapted further. Besides determining the precision based on a set of beacon and receiver locations, the model is able to select the optimal geometric configuration based on a performance threshold (e.g. required precision). Depending on the performance threshold, the optimal configurations can either be a single solution or consist of multiple solutions that satisfy the performance threshold of the customer. The performance threshold varies depending on the use case and the user requirements. Therefore, the amount of possible combinations in terms of 3D space, amount of available beacons, possible beacon locations, user tag locations and performance thresholds are limitless and the model can thus be used for all kind of applications.

All in all, the initial model (design computations) can be used by IPS customers for all kind of applications. The model is able to select the optimal geometric configuration in terms of precision based on a performance threshold specified by the user. Furthermore, the model requires user specified input parameters and the amount of possible combinations are therefore unlimited. Although the initial model has to be adapted further to account for the differences in modeled and measured data as a consequence of environmental factors, the initial model is a good initiative for the rising indoor positioning market and its customers. Therefore, the model can be adapted further such that it can explain significant differences between the modeled and the measured data by including factors that influence the system performance in real life, such as materialistic properties, signal attenuation, interference, multipath, NLOS etc.