Camera Sensor Qualification for Multi-View Vehicle Detection

Lassen V, Lübke M, Russwurm E, Franchi N (2026)


Publication Type: Journal article

Publication year: 2026

Journal

DOI: 10.1109/LSENS.2026.3709849

Abstract

This work presents a systematic evaluation of the impact of camera sensor quality on vision-based vehicle detection and multi-camera fusion in a controlled parking-lot model. It investigates whether multiple low-quality cameras can achieve performance comparable to high-quality systems. Sensor properties, including spatial resolution, noise behavior, and dynamic range, are analyzed using a reproducible evaluation procedure. A bird's-eye-view detection and fusion pipeline enables direct attribution of performance variations to hardware characteristics. Experiments under controlled illumination (sunny, cloudy, nighttime) show that detection performance does not increase monotonically with resolution, achieving best results at intermediate resolutions (mIoU up to 0.9156), while both lower (0.8054) and higher resolutions reduce performance. High-quality sensors provide higher localization accuracy and robustness, particularly under low-light conditions. Fusion achieves IoU values up to 0.83, with sensor quality directly affecting accuracy and variance. Multi-view fusion can partially compensate for lower sensor quality, though with increased noise sensitivity. The proposed methodology enables objective comparison and selection of camera systems for robust parking monitoring.

Authors with CRIS profile

How to cite

APA:

Lassen, V., Lübke, M., Russwurm, E., & Franchi, N. (2026). Camera Sensor Qualification for Multi-View Vehicle Detection. IEEE Sensors Letters. https://doi.org/10.1109/LSENS.2026.3709849

MLA:

Lassen, Vincent, et al. "Camera Sensor Qualification for Multi-View Vehicle Detection." IEEE Sensors Letters (2026).

BibTeX: Download