Thermal image drift is observed in prevalent industrial-level cameras because their optomechanical design is not optimised to reduce this phenomenon. In this paper, the effect of temperature on industrial-level cameras is investigated, focusing on the thermal image drift resulting from ambient temperature changes and warming-up process. Standard methods for reducing thermal image drift are reviewed, concentrating on the lack of repeatability aspect of this drift. Repeatable thermal image drift is crucial for applying a compensation model as random thermal deformations in sensors cannot be compensated. Moreover, the possible cause of this issue is explored, and novel optomechanical camera modifications are proposed that maintain the thermal degrees of freedom for the deforming sensor, limiting the lack of repeatability aspect of thermal image drift to a low level. The improvement is verified by conducting experiments using a specialised test stand equipped with an invar frame and thermal chamber. Considering the results from the application of the polynomial compensation model, the standard deviation of the central shifts of image drift is reduced by ×3.99, and the absolute range of image drift is reduced by ×2.53.