Abstract: Aiming at the problems of traditional original geological logging modeling, such as map pollution, low vectorization efficiency and low model accuracy, this paper proposes a vectorization 3D modeling method of original geological logging based on pattern recognition. This method mainly includes reading and cutting the original geological logging drawings, and Rosenfeld algorithm is used to reduce noise and refine graphics, and adopt Douglas-Pucker algorithm to extract and dilute the geological sketch boundary, and three-dimensional geological logging boundary, and the Coons surface is used to accurately build the 3D geological body model. This method has been successfully applied to the 3D geological modeling of Gejiu Mining Area. The numerical simulation results show that compared with the traditional original geological logging 3D modeling, the modeling results using this method can more truly reflect the change of ore body shape, and the modeling speed is reduced from 20 min to 30 s, which greatly improves the efficiency of geological logging modeling, and further reduces the risk of underground production exploration and mining operations.