3-Dimensional Spatial Analysis of Parking Lot Wall Scratch Using Mobile Point Cloud Data

Abstract. This study investigates the application of point cloud data for identifying and analyzing scratch patterns on walls within underground parking lots. As parking demands increase, narrow passages, intricate turns, and suboptimal layouts in parking facilities heighten minor collision risks, leading to substantial financial and operational costs. Conventional assessment methods, relying on on-site surveys and video surveillance, often fail to capture accurate spatial details and minor wall damages. This research employs high-precision point cloud data, complemented by image data, to precisely model and analyze parking lot layouts and scratch-prone areas. A novel approach integrating YOLOv10 object detection and PTv3 point cloud processing algorithms is developed to detect and localize scratches, while spatial analysis evaluates design factors affecting scratch distribution. Using handheld SLAM scanning devices, point cloud data was efficiently collected from five representative underground parking lots. The analysis of these datasets, which captured 327 wall scratches, reveals that structural layout and lighting conditions significantly influence scratch occurrence patterns, highlighting the potential of point cloud data in improving safety-oriented parking facility design.