摘要
为实现不同视角点云的快速精确拼接,利用多邻域曲率变化信息,通过多阈值自适应地进行关键点选取,较好地排除噪声的干扰,保证了关键点质量。对以关键点为中心不同半径邻域曲面进行拟合,利用曲面拟合系数描述关键点的特征,用较小的计算量最大程度提取出了点云表面的几何形状信息,识别精度高。选取特征描述符间距离最小的对应点对作为初始对应关系进行初次配准,在变换后的对应关系中设定距离阈值,大于阈值的对应关系则予以去除,方法计算简单、剔除错误匹配点对效果良好。运用聚类分选方法对对应关系进行优化,保留局部特征向量相似程度最大的对应关系,使得对应关系分布合理。最后利用优化后的对应关系二次配准得到最终结果。实验结果表明,本文方法运行速度快且有较好的抗干扰能力,适合实时三维测量应用。
In order to integrate point clouds generated from different points of view quickly and accurate- ly,key points are selected adaptively in this paper by using thresholds according to multiple neighbor- hoods curvature changing information. We describe the key points＇ characteristics using the coefficients which are got when fitting key points~ neighborhood surfaces with different radii. This method has a high degree of recognition and can extract the point cloud surface geometry information with a smaller amount of calculation. The initial registration depends on the correspondence selected from the minimum distance between corresponding key points＇ feature descriptors. After the corresponding points have been trans- formed using initial registration result, the distance threshold is set to remove wrong corresponding rela- tion whose corresponding points＇ distance is larger than the threshold. The strategy is simple in calcula- tion and has good ability to eliminate wrong matching points. Clustering sorting method is used to opti- mize the correspondence,which keeps the local corresponding relation that has the maximum similarity between characteristic vectors. It makes correspondence distributed evenly. At last, the final result is ob- tained through secondary registration using the optimized corresponding relation. The experimental re- suits show that the proposed registration method is fast and has good anti-interference ability,and it is suitable for real time 3D measurement applications.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2015年第9期1724-1731,共8页
Journal of Optoelectronics·Laser
基金
黑龙江省自然科学基金(F201123)
中央高校基本科研业务费专项基金(HEUCFX41304)资助项目
关键词
多曲面系数特征
关键点提取
点云拼接
coefficient characteristics of multi-surface~ key point extraction~ point cloud integrating
