Identifying the orientation of discontinuity planes is crucial for determining rock mass characteristics. Traces on rock outcrops typically serve as the primary data source for structural characterization. Obtaining precise trace data is vital, as they are often the only visible and measurable parameters in most outcrops. While manual recognition and detection of traces on high-resolution textured 3D models are relatively straightforward, the process can be enhanced using machine learning to detect and measure the shape of these 3D traces. This paper employs an AI-aided trace detection process where trace data are structured through 2D thinning, 3D re-projection, clustering, segmentation, and segment linking. The resulting linked segments are exported as 3D polylines, each corresponding to a trace. Our trace network analysis algorithm incorporates the concepts of curved and straight traces, with curved traces representing their discontinuity planes and intersecting straight traces identifying co-planar traces. We employ stereonet analysis and fuzzy c-means clustering techniques to identify principal joint sets. The study concludes by calculating normal set spacing from 3D joint sets, determining perpendicular distances between discontinuity planes within sets along a virtual scanline based on full-persistency or non-persistency assumptions. The methodology’s effectiveness is theoretical and practical, as demonstrated through its application to a real case study, marking a significant advancement in characterizing rock mass structural properties.