How Low Can You Go?: Performance Trade-offs in Low-Resolution Thermal Sensors for Occupancy Detection: A Systematic Evaluation
Mikko Rinta-Homi, Naser Hossein Motlagh, Agustin Zuniga, Huber Flores, Petteri Nurmi
摘要
We contribute by systematically analysing the performance trade-offs, costs (privacy loss and deployment cost) and limits of low-resolution thermal array sensors for occupancy detection. First, to assess performance limits, we manipulate the frame rate and resolution of images to establish the lowest possible values where reliable occupancy information can be captured. We also assess the effect of different viewing angles on the performance. We analyse performance using two datasets, an open-source dataset of thermal array sensor measurements (TIDOS) and a proprietary dataset that is used to validate the generality of the findings and to study the effect of different viewing angles. Our results show that even cameras with a 4 × 2 resolution - significantly lower than what has been used in previous research - can support reliable detection, as long as the frame rate is at least 4 frames per second. The lowest tested resolution, 2 × 2, can also offer reliable detection rates but requires higher frame rates (at least 16 frames per second) and careful adjustment of the camera viewing angle. We also show that the performance is sensitive to the viewing angle of the sensor, suggesting that the camera's field-of-view needs to be carefully adjusted to maximize the performance of low-resolution cameras. Second, in terms of costs, using a camera with only 4 × 2 resolution reveals very few insights about the occupants' identity or behaviour, and thus helps to preserve their privacy. Besides privacy, lowering the resolution and frame rate decreases manufacturing and operating costs and helps to make the solution easier to adopt. Based on our results, we derive guidelines on how to choose sensor resolution in real-world deployments by carrying out a small-scale trade-off analysis that considers two representative buildings as potential deployment areas and compares the cost, privacy and accuracy trade-offs of different resolutions.
问问这篇 Paper
问问你的智能体。
Lune 读过与它相关的顶会 Paper,每个回答都会注明依据哪几篇。
相关 Paper
- Modeling the Trade-off of Privacy Preservation and Activity Recognition on Low-Resolution ImagesYuntao Wang, Zirui Cheng, Xin Yi, Yan Kong 等CHI 2023 · 被引用 10 次
- TAPOR: 3D Hand Pose Reconstruction with Fully Passive Thermal Sensing for Around-Device InteractionsXie Zhang, Chengxiao Li, Chenshu WuUbiComp 2025 · 被引用 3 次
- HeatDeCam: Detecting Hidden Spy Cameras via Thermal EmissionsZhiyuan Yu, Zhuohang Li, Yuanhaur Chang, Skylar Fong 等CCS 2022 · 被引用 16 次
- THOR: Thermal-Guided Hand-Object Reasoning via Adaptive Vision SamplingSoroush Shahi, Farzad Shahabi, Rama Naboulsi, Glenn Fernandes 等UbiComp 2026
- PriView- Exploring Visualisations to Support Users' Privacy AwarenessSarah Prange, Ahmed Shams, Robin Piening, Yomna Abdelrahman 等CHI 2021 · 被引用 39 次
