双目面结构光传感器点云分层修正方法

打开文本图片集

中图分类号：TB9;TN247 文献标志码：A文章编号：1674-5124（2025）08-0131-06

Abstract： Binocular planar structured light sensors are composed of a left camera， a right camera，and a projector.Any two of the three components can be combined to complete point cloud measurement， thus obtaining three point clouds. Compared to binocular intersection for measurement， this approach can improve the completenessof point cloud measurement. However， under high and low temperature conditions，the three point clouds may have layering problems，affecting their usage.Existing methods require individual sensor calibration in a temperature chamber， which has low production eficiency， or require additional projection of stripes for on-site correction， occupying the measurement cycle.This paper proposes a method that can effectively solve the three-scan point cloud layering problem of binocular planar structured light sensors， without the need for calibration in a temperature chamber or additional projection of stripes.The method achieves point cloud registration by back-projecting the three point clouds into a two-dimensional space to find corresponding points，and then uses these corresponding points to complete the transformation between the three point clouds. Each measurement uses the binocular point cloud as the reference， completing the correction of the layering problem in the three point clouds. Experimental results show that in the case of 5.5mm layering， it can be reduced to 0.3mm a 95% reduction in the point cloud layering problem. Compared with the exiting method， the propose method is faster and the efficiency is increased by 30% ： Keywords： binocular structured light; triple scan; point cloud layering

0 引言

三维成像技术作为感知真实三维世界的重要信息获取手段，为重构物体真实几何形貌及后续的三维建模、检测、识别等方面提供了数据基础I。（剩余8171字）