MLS-based recognition and parameter extraction of roadway roof bolts/cables from 3D point clouds

The installation quality and long-term stability of roadway roof bolts/cables are critical to the safety of deep coal mines, yet conventional manual inspections are time-consuming, subjective, and difficult to perform frequently. This paper presents a mobile laser scanning (MLS)-based method for automatic recognition and parameter extraction of roof bolts/cables from 3D point clouds. A SLAM-based MLS system is first used to obtain a continuous 3D model of the roadway. The roof is then globally aligned and separated from surrounding objects using a cloth simulation filter (CSF), which yields candidate roof points. Instance-level clusters are extracted by density-based clustering combined with geometric constraints, and principal component analysis (PCA) is applied to derive key geometric parameters for each bolt/cable, including exposed length, inclination, and row/column spacing. Field experiments were carried out on five consecutive segments of a return airway in a deep coal mine, with 127 manually labeled bolts/cables. The proposed method correctly identified 118 of them, achieving a precision of 96.72% and a recall of 92.91%, while also providing an automatically generated parameter database for engineering evaluation. The results indicate that the method can effectively support intelligent assessment of roadway support quality, although its performance remains dependent on MLS data quality and the visibility of exposed bolts/cables, highlighting the need for further validation in additional mines and under more extreme monitoring conditions.