Occluded Multi-Human Target Segmentation in Dynamic Indoor Scenes Based on Thermopile Array Sensor

Low-resolution thermopile infrared array sensor (TPAS) has gained increasing attention for real-time perception and monitoring of indoor human targets. Most recent researches focus on single individual just because of its low resolution of TPAS. In addition, there are many dynamically changing heat sources besides human targets in indoor environments. The dynamic background removal and occluded human instances' distinction encounter significant challenges in low resolution based on TPAS. To address the interference of dynamic heat sources in multiperson indoor scenarios, this article proposes an algorithm called adaptive Gaussian background removal algorithm applying a priori map for multi-human (AGBR-PMM). In addition, to enhance the subsequent human behavior recognition performance, it is necessary to segment occluded human targets after removing indoor heat source interference and extracting all human targets. PANet is used to segment different human regions, particularly the areas with occluded individuals. A combination of these two methods in low-resolution scenes achieves a good segmentation of occluded human instances while eliminating interferences from dynamic indoor scenes. It outperforms AGBR-PMM and PANet alone in terms of accuracy, precision, and F1-score of 97%, 92%, and 94%, respectively. This lays the foundation for subsequent human behavior recognition using low-resolution TPAS.