This thesis explores the development and validation of a synthetic framework for oblique aerial image adjustment and object point detection, with the goal of improving photogrammetric workflows in complex urban environments. The research is motivated by the inherent challenges of oblique imagery, such as occlusion, perspective distortion, and variable visibility, which complicate traditional adjustment procedures. To address these issues, the study employs a novel approach by generating synthetic test cases that emulate real-world oblique aerial data, enabling controlled experiments and sensitivity analyses. Utilizing data from recent aerial campaigns over Rotterdam, including both nadir and oblique images, the research implements and evaluates various adjustment and feature detection algorithms, including Bundle, DISK, SIFT, and LightGlue. The synthetic framework allows systematic testing of key parameters and environmental conditions, such as occlusion and lighting variations, providing insights into the robustness and limitations of different methods. Although the results demonstrate promising potential for synthetic data to replicate key geometric and photogrammetric behaviors, challenges remain in achieving full photorealism and seamless transferability to real-world applications. The findings underscore the importance of synthetic data in advancing urban geospatial systems and support the early-stage design of aerial collection systems, with particular relevance for municipal maintenance, planning, and infrastructure management in the Netherlands. The study concludes with recommendations for future research directions, emphasizing the integration of more photorealistic synthetic imagery and improved synthetic-to-real transfer methods to enhance the accuracy and reliability of oblique aerial mapping workflows. Overall, this work contributes to the growing body of knowledge on synthetic data use in photogrammetry and opens pathways for more resilient and efficient urban mapping solutions.

LIght Detection And Ranging (LiDAR) imaging technology has advanced over the past two decades, being used for applications such as Digital Terrain Models (DTM) and Building Integration Modeling (BIM) integration. The resulting products, point clouds, serve diverse purposes, each demanding specific quality standards. Failing to meet the standards risks rendering the data ineffective or even unusable. Contractors, including Rijkswaterstaat, therefore, specify adherence to set requirements or standards. Current manual sampling for assessing the quality highlights the need for an automated tool.

This study proposes a workflow for automating quality validation of LiDAR infrastructure point clouds. The workflow assesses point cloud quality based on three primary components: coverage, relative accuracy, and absolute accuracy. The methodology includes:
• Point Cloud Density assessment: involves analyzing 2D horizontal cells and partial-3D spaces to ensure compliance with density requirements.
• Overlapping Regions Alignment: identifies and compares surfaces in overlapping areas of point clouds to determine relative accuracy.
• Benchmark alignment: extracts points corresponding to spherical targets, estimates the center coordinates, and evaluates adherence to absolute accuracy standards.

Quality assessments were conducted on static, mobile, and airborne point clouds. The static point cloud analysis revealed non-compliance with density requirements in 2D, with approximately half the points failing to meet standards. In 3D analysis, compliance was observed for 1 m2 horizontal cells, but individual 1-meter sections often fell short upon closer inspection. These findings highlight the need for tailored quality standards: detailed 3D analysis is crucial for complex environments like tunnels, while road environments can be effectively evaluated in 2D. Relative accuracy assessments for static and mobile datasets showed compliance with scanner specifications, with RMSE values meeting the specified requirements. Absolute accuracy assessments on static point clouds met requirements with minimal deviations in both XY and Z directions.

Recommendations include defining density requirements for different environments, establishing acceptance criteria, and defining allowable deviations for all requirements.

Semantic segmentation, a task vital in the creation of 3D point cloud models for buildings, is aimed at assigning meaning to individual points. However, due to the vast volume of unstructured point cloud data, precise semantic segmentation remains challenging. Significant progress has been observed in recent years with the application of deep learning techniques to point cloud segmentation, and the effectiveness of Dynamic Graph Convolutional Neural Network (DGCNN) and K-Nearest Neighbors (K-NN) in handling point cloud data has been recognized.
In this study, Dynamic Graph Convolutional Neural Network (DGCNN) was utilized for semantic segmentation on a building's point cloud scene. We adopted K-Nearest Neighbors (K-NN) as a crucial component of our methodology to optimize the segmentation process. By varying 'k' values in K-NN and exploring different block sizes, we aimed to obtain various segmentation results for comparison. When a block size of 1 meter was employed and 'k' was set to 20, an overall accuracy of 90.32%, mean accuracy of 87.64%, and IoU of 80.71% were achieved. However, the most favorable segmentation outcomes were observed when the block size remained 1 meter, and 'k' was set to 30, resulting in an overall accuracy of 93.86%, mean accuracy of 90.68%, and IoU of 84.97%.
These experiments underscore the significance of parameter selection in optimizing the performance of DGCNN for point cloud segmentation. The findings reveal that adjustments to 'k' values and block sizes can significantly influence segmentation accuracy and quality, emphasizing the importance of parameter optimization in the context of semantic segmentation for building point clouds using deep learning techniques. The utilization of K-NN played a crucial role in achieving these improvements by allowing us to adapt to the inherent variability in point cloud data.

The 3D point cloud representation of building holds significant importance in developing a comprehensive smart city model. However, due to the substantial volume of point cloud data and its inherently unstructured nature, accomplishing semantic segmentation of the point cloud poses a formidable challenge. Recent years have witnessed substantial advancements in employing deep learning techniques for point cloud segmentation. Both PointNet and PointNet++ have demonstrated their prowess in managing point cloud data. In the context of this study, we harness the potential of DGCNN for segmenting building point clouds. DGCNN plays a pivotal role in partitioning the input point cloud, thereby influencing the sensor domain's network range. Consequently, this research delves into the role of parameter k in K-NN (K-Nearest Neighbors) and its implications. The findings illustrate that augmenting the 'k' value contributes to enhancing the overall precision of DGCNN. Nonetheless, complete precision in segmenting every module cannot be assured. The research objectives in the study is building façade, and the targets are windows, walls, balconies and doors. In this study, the key parameters are the values of 'k' and the 'block size.' Experiments were conducted by varying the values of these two parameters to obtain different segmentation results of building facade point clouds. Notably, setting k to 20, block size to 1m achieves an overall accuracy rate of 89.75%, a mean accuracy rate of 85.03%, an IoU of 76.24%, while elevating k to 30 and block size to 1m results in a heightened accuracy rate of 95.74%, a mean accuracy rate of 89.37%, an IoU of 81.52%.

Everything around us is rapidly changing. Whole new blocks of buildings are built, huge infrastructural projects are constructed and so on. Hence, there is a need of a reliable and up-to-date inventory of the area and the objects of interest for mapping and monitoring assets and their changes. An answer of this upcoming need is an automated inventory of infrastructure using Remote Sensing and artificial intelligence (AI) techniques. Rail sector shares the same need for fast and reliable inspection on its infrastructure. Monitoring frequently the condition of the railway infrastructure can improve the maintenance efficiency and the avoidance of hazards. The traditional monitoring techniques are costly, time consuming and in some cases dangerous, due to their reliance on the physical presence of the inspector. Hence, new state-of-the-art techniques that are able to frequently and without putting in risk human lives, inspect the condition of the railway and its infrastructure. This master thesis aims at developing an efficient workflow for combining 2D imagery and 3D light detection and ranging (LiDAR) point clouds for the automated detection and localization of the railroad infrastructural objects into 3D world coordinate system, for monitoring the railway infrastructure. Using deep learning (DL) methods in imagery we detected and mapped, approximately the 60% of the railroad equipment of our interest (i.e. light signals and equipment boxes). These detected equipment were analyzed with stereoscopic techniques to retrieve their position in 3D world coordinate system. That led to the automated creation of a geographical information system (GIS) map having the positional and class information of railway equipment. Once the detected objects were mapped, then the point cloud data were automatically cropped into voxels including the same objects. Hence, using various sophisticated machine learning (ML) techniques, the points referring to the objects were classified. Furthermore, combining the positional information provided via 2D analysis with 3D point clouds, the vertical position was refined and the height of the mapped objects was estimated. Lastly, the positional information estimated from the 2D analysis enhanced the unsupervised ML classification in point clouds. The product of this classification, has the potential to be used as training data to train supervised point cloud classifiers.

In SAR interferometry, the data acquired by satellite are large. The points in the dataset are usually expressed in spatial-temporal dimension in which many epochs may be included. Extra information including contextual values and data from other sources sometimes are important to be taken into account when analyzing the deformation based on the time series. However, this kind of extra information is not supplied in the dataset. Here we show a way to augment the InSAR dataset by updating the new acquisitions, adding contextual values and combining data with other sources. We create an enriched dataset which provides one more knowledge on the points in the dataset, such that more accurate analysis and decision can be made. Also, a way to query a certain bunch of points with the same feature category is descripted, which allows one to find the points of interest based on their features. Our result provides a way to utilize multiple data sources and to combine them into an InSAR dataset in order to make the dataset be more efficiently used. The way depicted in this paper might be a hint of how to do a more accurate and comprehensive analysis based on a dataset. Furthermore, an automatic or semiautomatic way to modify the dataset with some tools are helpful.

By combining the RACMO2.3 RCM output with GRACE non-tidal ocean and atmospheric pressure anomaly as well as geo-center motion corrections from 2003 to 2017, a physics-based model to estimate the vertical displacement of the Greenland surface bedrock is created. This model is able to convert the mass and pressure anomalies into vertical displacement in the spherical harmonic domain with the use of load deformation coefficients based on the PREM earth model. The computed vertical displacement has a weak correlation with the observed displacement as measured by the GNET GNSS station network. Both the computed and observed vertical movement show seasonal signals, however the computed signal has both a lower amplitude and different phase to the observed displacement. This could possibly be explained by the selection of the GNET stations used in the analysis or possibly the implementation of the non-tidal ocean and atmospheric pressure anomaly. Questions regarding the optimization of the processing within the model are discussed but remain open. Recommendations for further research into the complexity of the system and shortcomings of the model are discussed in detail.

The Sentinel-5 Precursor satellite has a payload of the TROPOspheric Monitoring Instrument, TROPOMI.The satellite was launched in 2017 by ESA with the intended goal of measuring trace gases in the atmosphere.One of the products of TROPOMI is the Tropospheric NO2 column. This product is based on thespectral measurements to obtain the column abundance of NO2 in the troposphere. This product alsorelies on a-priori data and one of  these a-priori datasets is the albedo dataset.The currently used dataset has a resolution of 0.5°x 0.5°, which corresponds to approximately 55 kmx 34 km at mid-latitudes. The TROPOMI pixel size is significantly smaller, 3.5 km x 7 km. Due to this large difference in resolution the discussion arises if this used dataset is sufficient for accurate results. This researchmakes a comparison between the current a-priori dataset and possible replacements.
This paper makes this comparison by calculating Air Mass Factors (AMFs) using the OMI LERalbedo climatology as a reference and the two alternative high resolution surface reflectance datasets,Sentinel-2 and Landsat-8. These surface reflectance datasets were regridded and averaged on the corresponding TROPOMI grid. The focus area of this paper is the Greater Rotterdam region in the Nether-lands.Before these AMF calculations were done, a comparison between Sentinel-2 and Landsat-8 surfacereflectance datasets is made. This is done both on their own high resolution and regridded onto theTROPOMI grid. Above water surfaces and land covered by vegetation a bias of approximately 0.01 was present between the two high resolution surface reflectance datasets. These differences are relatively small. The differences calculated for the datasets regridded to the TROPOMI grid were also relatively small, with a bias of 0.01 above the water and vegetation surfaces.
Two cases were studied during this research: the 21st of April 2018 and the 6th/7th of May 2018. Theresults show that significant improvements can be made by using a higher resolution surface reflectancedataset. A median bias of -10.4% (-15.6%) was calculated for the 21st of April for Sentinel-2 (Landsat-8)compared to the AMFs based on the OMI albedo dataset. For May this was -3.9% (-9.3%). Furthermorethis study showed extreme AMF-biases of 68.0% overestimation and 39.8% underestimation by the OMIalbedo dataset compared to Sentinel-2, where the overestimation was observed over the greenhouses inthe Westland region and the underestimation in the rural region to the East of the domain in April.For May the underestimation was mostly observed to the West (North Sea), indicating that over regionswith a low surface reflectance the atmospheric correction greatly influences the AMF. The comparisonbetween Landsat-8 and OMI showed similar results in the AMF differences.These findings are supported further by a recent Sentinel-5P validation study, which comparedground based observations to the TROPOMI observations. This project found an NO2 underestimationof approximately 20% for many different stations. This research suggest that, at least partly, this difference can be explained by the coarse resolution of the a-priori albedo dataset used. 

This research focuses on GNSS users interested in a faster position fix in a situation where the receiver holds limited data. This occurs when turning on a mobile phone or sport watch to use a location-based service, driving out of a parking garage with a navigational device or activating a search & rescue beacon. In these cases, users care more about directly setting a certain action into motion, than waiting until a sub-meter level of accuracy is acquired. For this purpose, we aim to decrease the time to first fix (TTFF) of a receiver at the expense of position accuracy by focusing on its most dominant component: the time to read the first fix data, i.e. the TTFFD. In turn, the TTFFD is decreased by reducing the size of the clock correction and ephemeris data (CED) and integrating this reduced set into the existing Galileo I/NAV message while maintaining backward compatibility. Three reduction strategies are presented to reduce the size of the CED. First, the expression of the CED parameters relative to the matching almanac parameters, which requires downloading the most recent almanac. Second, the absorption of the clock and ephemeris time reference into related parameters by leveraging redundancies in the user algorithm. Third, the truncation of the CED parameters, where the optimal parameter bit allocation per CED size is determined by means of a Monte Carlo simulation. Subsequently, the optimal reduced set is integrated in the existing Galileo I/NAV message, since it is not only size but also the location and repetition rate of the reduced CED in the navigation message that determine the final decrease in TTFFD. A simulation is developed where the three reduction strategies are combined and applied to one year of Galileo navigation messages and almanacs starting from October 2017. The current I/NAV message has a size of 428 bits and an average TTFFD of 25.4 seconds. The relative expression of the CED parameters, the absorption of the clock time reference and ephemeris time reference yield a 76, 15 and 10 bit reduction respectively, without a significant decrease in position accuracy expressed as the signal-in-space ranging error (SISRE). The reduction in size that can be realized by truncation results from a trade-off with the maximum acceptable level of SISRE. When combining all three reduction strategies a 296 bit, i.e. 69%, size reduction can be realized at a SISRE level of 10.04 meters, yielding a 132 bit CED. 66% of the size reduction is in this case achieved by parameter truncation. To maintain backward compatibility, the spare and reserved bits of the I/NAV message are used to integrate the reduced set. This finally results in an average TTFFD of 6 seconds, which is a decrease of almost 20 seconds. With the reduced CED the user gains full independence of out-of-band channels for the computation of a fast position fix. In addition, it decreases the impact of bit errors on the TTFF and increases the chance of receiving an error free set of first fix data. Moreover, the reduced CED is not only of interest to fast fix applications that desire a decrease in waiting time. Positioning based on the reduced set also decreases the time the receiver has to be "on" to read the navigation data and consequently the receiver's power usage. This leads to a second type of applications that benefits from the reduced CED: low-power applications that do not require a high accuracy, such as freight management, wildlife tracking and stolen goods trackers. To conclude, the constructed reduced CED provides multiple advantages for a wide range of GNSS users and should therefore be considered as a purpose for the spare and reserved bits of the Galileo I/NAV message.

This paper analysis the effects of soil moisture and vegetation water content (VWC) on total backscattering (σ_tot^0) simulated by water cloud model (WCM) throughout a growth cycle of maize at different frequencies, polarization modes, and incidence angles. Firstly, the bare soil backscatter (σ_soil^0) was simulated by Integral Equation Method (IEM) surface scattering model [1] or Dubois empirical backscattering approach [2]. Then, to analysis the effect of vegetation cover, a standard and a double layer WCM based on parameter sets of three published studies [3]-[5] are applied in this study area to simulate the two components of σ_tot^0, direct backscatter from vegetated surface (σ_veg^0) and attenuated soil backscatter (〖γ^2 σ〗_soil^0). The input parameters of IEM, Dubois and WCM are supported by a series of ground measurements performed in Florida during the entire growth season, which includes soil moisture, surface roughness and vegetation biomass measurement. According to the analysis at different frequency and incident angle, the increase of bulk VWC can lead to either an increase or decrease in σ_(tot ). The different impact is determined by either σ_veg^0 or 〖γ^2 σ〗_soil^0 is the main contributor to σ_tot^0. At higher frequencies and larger incident angles, where the dominant part is from σ_veg^0, σ_tot^0 will increase with increasing bulk VWC. While, when 〖γ^2 σ〗_soil^0 becomes the main contributor to σ_tot^0, increasing bulk VWC leads to a denser canopy and thus more incoming microwave is attenuated. Therefore, increasing bulk VWC results in a decrease in σ_tot^0. Besides, according to the results obtained at C-band, different incident angle, HH-polarized microwaves are more sensitive to changes in bulk VWC, especially at larger incident angle. VV-polarization is less affected by vegetation cover, σ_tot^0is sensitive to soil moisture even at peak biomass and large incidence angles, which is attributed to scattering along the soil-vegetation pathway.