面向模拟试箱的大场景点云数据处理算法研究

Research on Large-scale Point Cloud Data Processing Algorithm for Simulated Container Loading Test

传统的集装箱船货舱精度验收方法完全依赖手工和目视操作,该过程耗时长且资源消耗大。随着三维激光扫描技术的发展,目前已经能够对货舱大场景进行高精度的三维点云扫描采集。针对基于三维激光扫描的集装箱船试箱的大场景点云中难以提取并分析导轨、底锥等关键数据的问题,结合三维激光测距技术和点云处理算法在舱段场景中应用的优势,论文提出了基于三维激光扫描的集装箱船试箱的算法处理流程。在该流程中,针对导轨点云的特性,改进了RANSAC算法的平面筛选条件;针对底座点云所占比例小且难以精确提取的问题,基于Bhattacharyya距离实现了一种自动提取底座点云的方法;针对底锥点云难以确定其轮廓特征和中心点的问题,提出了一种结合遗传算法和ICP算法实现对底锥轮廓拟合的方法。最后在实际点云数据上进行了试验,验证了所提方法的有效性。

Traditional methods of the cargo hold of container ships rely entirely on manual and visual operations,which are time-consuming and resource-intensive.Advanced 3D scanning devices are applicable to acquire high-precision 3D point clouds of large-scale cargo hold scenes.To address the challenge of extracting and analyzing key data,such as cell guides and bottom cones,from the large-scale point cloud of the cargo hold,this paper proposes a framework to analyze the construction quality of the cargo hold based on the 3D point cloud.In this framework,the plane filtering condition of the RANSAC algorithm is improved based on the geometrical properties of the cell guide.To address the problem of the small proportion and extracting difficulty of the foundation point cloud,an automatic extraction method for the bottom base point cloud is proposed based on Bhattacharyya distance.To overcome the difficulty of determining the contour features and the center point of the bottom cone point cloud,a method combining genetic algorithm and ICP algorithm is proposed.Experiments are conducted on actual point cloud data to verify the effectiveness of the proposed method.