Research on recognition of bedding system coverage rate using infrared thermal imaging

This study investigated a new challenging problem: how to recognize bedding system coverage rate (BSCR) in real-time without contact. To overcome this problem, a method combining infrared thermal imaging with image segmentation algorithms was proposed. The infrared thermal images of 30 subjects with common coverage forms were captured under three shooting angles (45°, 60°, 90°) and three ambient temperatures (23 °C, 26 °C, 29 °C) to evaluate the effect of these factors on recognition accuracy. The proposed method was then applied in real-time to recognize BSCR in summer. And the subjective thermal perceptions (thermal sensation votes, thermal comfort votes) and objective sleep quality were collected. Three image segmentation methods were tested. And the K-means clustering excelled in recognizing BSCR, achieving an MAE of 3.70, RMSE of 4.67, and $R^2$ of 0.901. The optimal shooting angle for infrared camera was 60°. Additionally, the summer indoor temperature (23 °C–29 °C) had no significant impact on the recognition accuracy. The average BSCR during sleep was 76.6 % under conditions of thermal comfort and good sleep quality. Notably, females exhibited the BSCR that was 8.9 % higher than that of males (p < 0.01). This study provides reference data and insights for establishing a non-contact sleep thermal comfort prediction model.