特征提取的点云自适应精简

发布时间：2018-10-05 08:12

【摘要】：作为一种反映物体形貌的三维信息,点云数据的原始数据量十分庞大,直接对过多的数据进行操作会影响后续重建等工作。本文提出了一种新的点云特征提取自适应精简算法。首先对原始点云进行空间划分,构建点的k邻域,设置特征参数,进行特征分析,识别不同区域的信息和数据。然后针对平面数据预先进行边界的检测和提取,对剩余部分进行精简。最后,针对非平面区域,先提取特征,再根据曲率的不同进行不同程度的精简。办公室数据扫描实验结果表明,处理大小为百万以内点的点云模型可以在几秒之内完成,精简比能够达到90%以上,与原始数据间的误差较小:平面部分在精简前后平均偏差均在0.02mm以内,波动很小,为0.005 7mm;非平面区域精简前后的平均偏差均在0.08mm左右,差值仅为0.000 3mm,精简精度得以保证。因此,利用提出的算法处理后的数据能更好地展示物体的形貌。
[Abstract]:As a kind of 3D information which reflects the shape of objects, the original data of point cloud data is very large, and direct operation of too many data will affect the work of subsequent reconstruction. In this paper, a new adaptive reduction algorithm for point cloud feature extraction is proposed. Firstly, the original point cloud is divided into space, the k-neighborhood of the point is constructed, the characteristic parameters are set up, the feature analysis is carried out, and the information and data of different regions are identified. Then, the edge of the plane data is detected and extracted in advance, and the remaining part is reduced. Finally, for non-planar regions, features are extracted first, then reduced to different degrees according to different curvature. The results of the office data scanning experiment show that the point cloud model can be completed in a few seconds, and the reduction ratio can reach more than 90%. The error between the plane part and the original data is small: the average deviation of the plane part is within 0.02mm, the fluctuation is very small, and the average deviation of the non-plane area is about 0.08mm, the difference is only 0.000 3 mm, so the precision of the reduction can be guaranteed. Therefore, the data processed by the proposed algorithm can better display the shape of the object.
【作者单位】：河北工业大学机械工程学院;
【基金】：国家自然科学基金资助项目(No.51675160,61171048) 河北省应用基础研究计划重点基础研究资助项目(No.15961701D) 河北省高层次人才资助项目(No.GCC2014049) 河北省人才工程培养经费资助项目(No.A201500503)
【分类号】：TP391.41

【相似文献】