Railway transportation is crucial to reducing greenhouse gas emissions and meeting increasing global demand for passenger and freight transport. To maintain a reliable, high-capacity railway infrastructure, regular and accurate maintenance is necessary. The recent development of a Rail-based Track
Surveying System that can be mounted on any operational train, presents a comprehensive solution to acquire track geometry measurements while limiting track unavailability. The system is able to simultaneously acquire measurements of both the absolute- and relative track geometry using a combination of a GNSS / INS integrated navigation system and a laser ranging sensor. The measurement unit is able to obtain consistently accurate absolute track geometry measurements (σ2D = 8 [mm] (Wang et al.2019)) in areas with good GNSS signal reception. However, in areas with limited GNSS coverage, the accuracy of these measurements can fall below the requirements for rapid track inventory acquisition (3 [cm], p = 0.95 (Specht et al. 2016)). This thesis, developed and conducted in lose collaboration with Fugro, aims to improve the accuracy of the trajectory solution for rail-based track surveying systems in such limited GNSS environments.

At the time this research was performed, there was no ground truth or reference trajectory available to assess the accuracy of a trajectory solution. Therefore, the research first establishes quality metrics to provide a measure of the accuracy and certainty of the integrated GNSS/INS estimated trajectory solution. By nature of the data acquisition process of the RTSS, multiple runs, or trajectory estimates, are available of an arbitrary track segment. The observed cross-track trajectory spread (precision) provides a qualitative metric to assess the level of accuracy of multiple passes over the same track segment, given that the
individual measurements are unbiased. To quantify the cross-track trajectory
spread, the standard deviation and smallest circle radius are selected as statistical and absolute metrics.

Next, the research identifies the existence of unreliable or inaccurate GNSS positioning updates in the integrated trajectory solution by linking the cross-track trajectory spread to the Quality Control metrics of individual runs. The level of accuracy of the GNSS / IMU integrated trajectory solution was found to be decreased at track segments where one (or multiple) runs showed sustained periods of high (> 5) PDOP values. A decreased accuracy in the order of 10 [cm] was observed for the high PDOP GNSS positioning updates. Furthermore, these inaccurate position updates were weighted too heavily in the Loosely Coupled integration scheme, reducing the accuracy of the integrated GNSS/IMU
trajectory solution. To improve the accuracy and certainty of the integrated trajectory solution, the GNSS positioning updates in segments with sustained high PDOP values can be inactivated. Inactivating these high PDOP GNSS positioning updates, reduced the observed cross-track trajectory spread by
as much as c.50%.

The thesis provides valuable results for the improvement of the accuracy and certainty of GNSS/IMU trajectory determination in rail-based applications. Furthermore, the framework and data-analysis algorithm developed and presented in this thesis research could be used to quantify the effect of different
quality parameters and processing thresholds on the accuracy of the trajectory solution. The main conclusion and recommendation of this report - processing the trajectory against a tight PDOP constraint - leads to a trajectory estimate for a rail-based surveying application with improved accuracy and certainty. Moreover, recommendations for research implementation, further trajectory accuracy gains and continuation of this research are also presented in this report.

There are multiple types of objects distributed along the rail trackside to support the operation of railroads. Damage to those objects will result in delay and cancellation of trains. Therefore, the maintenance of the objects is important to ensure the safety of railway operations, and the key to maintenance is to know the accurate position of the objects. RILA - a railway mobile mapping system developed by Fugro - provides continuous monitoring of the rail environment without interrupting the daily operation of rail tracks. The RILA dataset contains different types of data including video frames, point clouds, and measurement trajectories from the Global Navigation Satellite System and Inertial Measurement Unit (GNSS-IMU). This enables a remote and automatic object positioning process based on object recognition in video frames and positioning using photogrammetric techniques.

Most of the rail trackside objects are parts of the railway electrification system and the railway signaling system. The available ground truth data contains more information about signaling objects, so this thesis mainly focuses on the positioning of signaling objects. Three typical components: signal lights, railway cabinets, and railway side markers are selected to make case studies and develop the object detection and localization pipeline for them.

There are two main issues that existed in former research: One is object recognition only gives a 2D pixel position for an object, but we want to know its 3D position in the world; the other is the contextual information between frames is ignored in object positioning. Based on that, this thesis developed a new workflow: firstly, this thesis proposes the use of an existing 3D object detection model, Single-Stage Monocular 3D Object Detection via Keypoint Estimation (SMOKE), to recognize objects of interest from the RILA video frames. This model can not only predict an object's 2D pixel position on the image, but also its 3D position with respect to the camera. This enables the localization of an object in each of the frames it is detected in, which provides redundancy to define its final position. Secondly, this thesis developed a pipeline to clean the detection output of SMOKE so that most false detections can be removed. In addition, this thesis developed a method to use the contextual information in image sequences to position objects: first, extract the sequences by classifying all the predictions in the detection output through the Euclidean distances, then for each sequence, analyze the varying trend of predicted positions to select reliable positions. The mean of the selected positions is the final improved position of that object. By making tests on a part of the railway near Ely, UK, with 2983 frames, the workflow detected and positioned all of the 4 signal lights, all of the 10 markers, and 13 of the 14 cabinets in the testing dataset. Most signal lights and cabinets can be positioned within 1 meter to the ground truth, and the mean positioning offset for the railway side markers can be controlled around 2 meters. Complex scenarios could be solved, when up to three cabinets were located close together.