Image based evaluation of textured 3DSEM models

Photogrammetric methods enable the construction of 3D SEM models from 2D images. Most software for this purpose is designed for photographic images. The software tries to minimize modelling error but some uncertainty usually remains in the model. In such approaches no ground truth measurement for microscopic objects exists for comparison with the finished model. In the proposed method, a textured model surface is compared to the SEM images to map the error locations on the model. The method is illustrated using two datasets.