Most non-contact methods and systems for 3D spatial coordinate measurement are based on optical sensors and signal processing. Object surfaces are discretely detected to obtain their coordinates and generate a point cloud that can be processed to reconstruct the object's shape and dimensions. Optical sensor-based technologies are the most widely used because they are non-destructive and do not need to touch objects. One of their best properties is their fast measurement acquisition rate. However, one of their disadvantages is the presence of atypical values in their measurements due to their sensitivity to optical ambient noise. Filtering atypical values from point clouds allow the generation and reconstruction of mesh and geometrical models of objects with appearance and dimensions similar to reality. This work is focused on the point cloud post-processing of 3D spatial coordinates measurements obtained from a technical vision system to eliminate inaccuracies in the reconstruction models.