Creation of a CAD model from an hard model is something necessary for design modification, part replication or rapid prototyping and surface inspection. This is reverse engineering. Scientific literature presents many different approaches, even if, actually all the systems, mechanical (contact devices) and optical (non contact devices), work with constant acquisition pitches. This became a great deal in relation with the different object morphologies combinations that the same surface could show. Working in fact with a constant pitch on a free-form surface it is possible to struggle with an insufficient points cloud density, when the acquisition pitch would be a compromise between the complex and elementary features that describe the object to acquire, or with an excessive points cloud density, when the acquisition pitch represents the highest scanner resolution. Referring to this situation, this paper wants to propose, starting from a first raw acquisition, an automatic methodology, directly implemented on the acquisition device, for the selective individuation of surface zones which present sensible curvature. In this approach the curvature of the measured surfaces is analyzed by defining a threshold over which it is necessary to perform a deeper scansion of the surface. In the present paper a methodology for the definition of the threshold value based on the measurement system uncertainty is described. In the current description the method is applied to an algorithm for curvature analysis, but it could be extended to any other approaches. Furthermore, it will be demonstrated that this new methodology is simple to apply and can be easily automated directly in the control scanner software. In the end of the paper a practical example is described in order to give an experimental validation of the method.