In this work, we present a method of determining human subjects via a low-resolution thermal imaging sensor. Since the image quality of the low-resolution thermal imaging sensor could be suffering from heat signatures and recognizable patterns of human subjects are unable to be determined due to resolution issues, it is recommended to employ a probabilistic method. This paper presents how human subjects can be expressed in terms of pixel size, standard deviation, label movement, vector tracking, label lifetime and a rewarding system based on those. Various pre and post-image processing methods will be covered including background collection, Gaussian filtering, segmentation, local/global adaptive threshold and background learning.