Surveillance Potential of High-Resolution LEO Constellations on Urban Mobility

The growing availability of modern, accurate and precise Earth observation satellites, including emerging constellations in Very Low Earth Orbit (VLEO), has opened new opportunities for large-scale, persistent, and cost-effective traffic monitoring from space. However, such capabilities also raise significant privacy concerns, as they enable continuous monitoring of large urban areas. This thesis investigates the feasibility of tracking vehicular movements using satellite imagery using current and planned future technologies, combining high-fidelity simulation with probabilistic trajectory reconstruction techniques. Realistic vehicular mobility patterns are modeled using simulation tools like OMNeT++ coupled with SUMO, while satellite dynamics and sensing are simulated through the SpaceVeins framework, integrating existing and planned constellations such as Jilin, Skysat, and Iridium. These simulations allow analyzing how parameters such as observation frequency, spatial resolution, orbital geometry, and urban morphology influence the ability to maintain continuous tracks of vehicles for a prolonged amount of time. To reconstruct trajectories from partial and noisy measurements, the observation data are processed using probabilistic and inference-based models, such as Hidden Markov Models and the Viterbi algorithm, combining geometric proximity and motion consistency along the road network. Occlusions caused by 3D urban structures are explicitly modeled, highlighting the critical role of urban morphology in determining the visibility and continuity of satellite observations. The analysis also considers realistic sources of detection error based on recent studies in satellite- based and aerial tracking, integrating them into the simulation pipeline. Finally, the analysis evaluates the proposed tracking framework against varying constellation architectures, measurement noise levels, and temporal observation gaps. The findings demonstrate that probabilistic inference significantly enhances trajectory reconstruction, effectively compensating for sparse or inaccurate data.Ultimately, this highlights both the feasibility and the inherent challenges of deploying a satellite constellation sufficient for persistent and reliable vehicular monitoring.