Infrared thermography is a non-invasive measurement method that can accurately describe immediate temperature changes of an object. In the case of continuous in vivo hand measurements, extracting correct thermal data requires a first step of image segmentation to identify regions of interest. This step can be difficult due to parasitic hand movements. It is therefore necessary to regularly readjust the segmented areas throughout the recording. This process is time-consuming and presents a particular obstacle to studying a large number of areas of the hand and long duration sequences. In this work, we propose an automated segmentation methodology that can automatically detect these regions on the hand. This method differs from previous literature because it uses a secondary visual camera and a combination of computer vision and machine learning feature identification. The obtained segmentation models were compared to models segmented by two human operators via Dice and Intersection-over-Union coefficients. The results obtained are very positive: we were able to decompose the images acquired via IRT with our developed algorithms, regardless of the temperature variation, and this with processing times of less than a second. Thus, this technology can be used to study the long-term thermal kinetics of the human hand by automatic feature detection, even in situations where the hand temperature experiences a significant variation.