Automatic non-rigid registration of multi-strip point clouds from mobile laser scanning systems

作者：Yan, Li; Tan, Junxiang^*; Liu, Hua; Xie, Hong; Chen, Changjun^*

来源：International Journal of Remote Sensing, 2018, 39(6): 1713-1728.

DOI：10.1080/01431161.2017.1410248

• TRAFFIC SIGN DETECTION
• TERRESTRIAL LIDAR
• URBAN OBJECTS
• EXTRACTION
• CLASSIFICATION
• SEGMENTATION
• RECOGNITION
• PERFORMANCE
• INVENTORY

Mobile laser scanning (MLS) systems equipped with precise Global Navigation Satellite System (GNSS) and inertial measurement unit (IMU) positioning devices are being used at an increasing rate for production of the high-accurate driving maps because of its safety and high performance in collection of 3D spatial data. In practice, GNSS signals may be blocked out by trees or buildings etc., and the errors of IMU are accumulated over time, leading to misalignments ranging from decimetre level to sub-metre level between point clouds from back and forth scans, or among multiple excursions. In this article, we propose a new time-variant model and an automatic solution to align the multi-strip MLS point clouds. Our methods are divided into three key steps: preprocessing to get representative points, two-step Iterative Closest Point registration to obtain correspondences, and time-variant errors estimation and correction of point clouds. We verified the solution using test data scanned in city road and highway environment. The experimental results demonstrate that the precision of the point clouds is significantly improved and the root mean square errors are about 4-5 cm.

• 出版日期2018
• 单位武汉大学