Research on a real-time dimensions detection method for roller tank lugs in non-orthogonal attitude within vertical shaft hoisting systems

Abstract Roller tank lugs serve as critical guiding devices in vertical shaft hoisting systems, playing a vital role in ensuring stable operation. Currently, existing dimensional measurement methods suffer from poor real-time performance and significant errors under tilt conditions. To address this, this paper proposes a real-time detection technique for roller tank lugs based on YOLOv8. This method acquires images and LiDAR point cloud data of the roller tank lugs without disrupting normal mine operations. Through joint calibration of LiDAR and camera systems, it integrates image and point cloud data to calculate the roller tank lugs’ dimensions. Building upon this foundation, this paper further proposes an inclination-adaptive measurement algorithm for roller tank lugs. Based on the YOLOv8 object detection method, the algorithm first performs corner detection on a calibration plate parallel to the roller tank lugs. It then indirectly calculates the tilt angle using a homography matrix decomposition. Following target identification, edge detection is performed to extract the diameter cross-section. Finally, the actual dimensions of the roller tank lugs are output via the measurement algorithm. Repeated measurements validated that the maximum error of this method consistently remained below 3 mm, with a standard deviation of 0.75 mm in measurement results, demonstrating high accuracy and stability.