The authors have approached some system design problems to achieve an interval of spatial resolutions for different characteristics of the observed scene, in which the observation probability can vary because of the thresholds of observation conditions. The paper proposes a simple analytical model for the estimation of the observation range (on the base of an imposed value for the observation probability), starting from a Boltzmann approximation which extends the observation prediction model based on the Johnson's classic criterion. Supplementary, several real image scenes, at different known observation ranges, were acquired on PC and also some patterns at different thermal contrasts, in laboratory conditions. The results of experiments have been extended into a function similar to the initial model proposed and which depends on the thermal camera used. The observation range was modelled, not only depending of target dimension and system resolution, but by the observation probability and the difficulty degree of observation, too. The deterioration of the thermal contrast has been simulated with image processing software, by a contrast controlled degradation of image, to estimate the observation probability in different environmental conditions. By a graphic-analytical optimization one can select some spatial resolution values which assure desirable acquisition probabilities of the targets at different thermal contrast values of the scene. The observation probability was analyzed for detection, recognition and identification.