This paper presents an alternative to measuring structural features in rock masses. The procedure is based on the reconstruction of the slope surface from a stereoscopic pair of images, taken with a calibrated camera. Complementary, hardware elements for referencing in the field and a software for generating 3D images, measurements and assessments are used. Once the 3D image is generated, it is possible to measure the orientation, spacing and geometry of discontinuities, as well as to process the data to generate the pole stereo networks of the measured planes. Additionally, these poles can be classified by supervised clustering techniques, to define the most representative families of discontinuities in the rock mass and to generate separation statistics for each family. The methodology allows to collect information of the entire free face of the slope quickly, reducing the exposure of the field team to falling blocks risk. Also, orientation and average inclination measurements of planes are made, reducing the error related to local variations in the normal vectors to the planes. Finally, systematic collection of information allows generating a more reliable probabilistic description of discontinuities, compared to traditional methods.