Occupant-centric cabin thermal sensation assessment system based on low-cost thermal imaging

• Published in: Building and environment 2024-08, Vol.261, p.111692, Article 111692
• Main Authors: Hou, Zhenyu, Lyu, Junmeng, Wu, Dongyuan, Chen, Jiangping, Shi, Junye, Lian, Zhiwei
• Format: Article
• Language: English
• Subjects: Assessment system; Facial recognition; Low-cost; Thermal imaging; Thermal sensation; Vehicle cabin
• ISSN: 0360-1323; 1873-684X
• DOI: 10.1016/j.buildenv.2024.111692

Assessing cabin occupants' thermal sensation in real-time enables automatic control of car air conditioning, improving driving safety and energy efficiency. Traditional methods, limited by their need for numerous physiological parameters, have restricted practicality. For this purpose, this study employs a low-cost thermal imaging sensor as the hardware core to establish a cost-effective and non-contact real-time assessment system for the thermal comfort of occupants within a cabin. The system comprises a thermal sensation assessment model developed based on subject experiments, along with facial recognition and segmentation algorithms optimized for thermal images. The thermal sensation assessment model developed, which employs cheek temperature, solar radiation intensity, and cabin air temperature as its features, exhibited an R2 of 0.617 on the test set. Furthermore, the facial recognition algorithm established for thermographic imaging achieved a mean accuracy of 96.5% and a recall rate of 99.0%. The system underwent validation in real-world vehicle environments, proving its ability to accurately detect and measure the cheek temperatures of occupants in the cabin and execute thermal sensation assessments. With a mean absolute error of 0.5 thermal sensation units in its output, its accuracy in practical applications was affirmed. This research provides an effective solution for automatically adjusting cabin air conditioning. •Developed a non-contact thermal sensation assessment system for cabin occupants.•Output of system can drive automatic controls for cabin air conditioner in the future.•Developed a simple thermal sensation assessment model for cabin, with an R2 of 0.617•Developed facial region segmentation algorithm for low-resolution thermal imaging.•Established system validated with a mean absolute error of 0.49, showing high accuracy.