Neonates require precise temperature management in incubators due to their immature thermoregulatory functions. Traditional methods which use skin-attached probes, are challenging because of the neonates’ delicate skin. Therefore, efforts are underway to develop non-contact temperature measurement using thermography. However, thermography can be obstructed during medical procedures by factors such as the medical staff’s hands. To address this issue, we employ Gagge’s two-node model, a human thermal model, to simulate body temperature changes. Our method integrates real-time sensor data, including thermography and incubator conditions, to estimate skin and core temperatures. We utilize backpropagation for rapid parameter optimization within the model. Importantly, the proposed method is not a black-box approach; it ensures explainability, providing a clear understanding of how temperature estimates are derived. Evaluation results based on data from five cases demonstrated that the accuracy is comparable to that of the probes currently used in NICUs, with the method achieving core temperature estimation with an average absolute error of 0.075 \(^\circ \text {C}\) , even 10 min after thermography data becomes unavailable.