<p>The identification of discontinuities is essential for rock stability analysis. However, most studies focus on small-scale slopes, limiting their applicability to large-scale slopes due to computational efficiency and accuracy constraints. In addition, the influence of exposure patterns of discontinuity on automatic recognition remains underexplored. This study introduces a discontinuity recognition method, optimizing the traditional computational logic of sequential edge identification to enhance computational efficiency. Unmanned aerial photogrammetry was used to acquire point cloud data from six slopes with varying scales and rock structures, and the proposed method was applied to identify discontinuities. Results show that the algorithm processes over ten million point cloud points within 2 h, exhibiting satisfactory performance in computational efficiency and recognition accuracy for planar discontinuity. A detailed quantitative error analysis further reveals the impact of discontinuity exposure patterns on accuracy when applying planar discontinuity recognition method to large-scale slopes. In blocky slopes, outward-dipping discontinuities are more exposed and recognized more accurately, while inward-dipping and oblique ones depend on structural characteristics. When the structural bodies of a slope have similar three-dimensional sizes, recognition accuracy improves; otherwise, it declines. In layered slopes, when the dip direction of the bedding is parallel to the slope, discontinuities with orientations similar to bedding planes are identified more accurately than those with other orientations. Conversely, when the dip direction of the bedding is perpendicular to the slope, discontinuities with other orientations are identified more accurately. This study highlights the effect of exposure patterns in automatic discontinuity recognition and provides new insights for engineering applications.</p>