The clustering of discontinuity orientation is crucial for revealing the distribution and characteristics of various types of discontinuities. Conventional clustering methods based on discontinuity pole density maps often rely on geological experience and lack objectivity. Therefore, the mean shift clustering algorithm was introduced to study the clustering of discontinuity orientation. Initially, discontinuity orientations with different degrees of dispersion were manually generated. Subsequently, these orientation data were converted into coordinates in 3-D space, and the sinusoidal value γ of the unit normal vector was employed as the similarity measure. The mean shift algorithm was then used to perform clustering analysis on the measured data set. Compared with the conventional pole density map method and the K-means clustering algorithm, the validity test index and clustering error recognition rate were close to those of the K-means clustering algorithm. Finally, taking the Chongqing Sangong rock slope as an example, the rationality and effectiveness of the new method were verified by field data. The results show that the performance of the proposed method surpasses that of the conventional discontinuity pole density map and K-means clustering algorithm. The clustering results are objective and reasonable, and the clustering effect for near-horizontal discontinuities is also satisfactory.