陈伯云,陆安江,赵麒,黄际玮,彭熙舜.激光点云中散乱电子元器件分割方法[J].包装工程,2022,43(13):216-224.
CHEN Bo-yun,LU An-jiang,ZHAO Qi,HUANG Ji-wei,PENG Xi-shun.Segmentation Method of Scattered Electronic Components in Laser Point Cloud[J].Packaging Engineering,2022,43(13):216-224. |
| 激光点云中散乱电子元器件分割方法 |
| Segmentation Method of Scattered Electronic Components in Laser Point Cloud |
| |
| DOI：10.19554/j.cnki.1001-3563.2022.13.028 |
| 中文关键词: 点云分割 数量检测 边缘提取 散乱电子元器件 区域生长 |
| 英文关键词: point cloud segmentation quantity detection edge extraction scattered electronic components region growth |
| 基金项目:国家自然科学基金(62065002，61865002) |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 针对散乱电子元器件计数过程中电子元器件分割困难的问题，提出一种基于点云簇平均法线夹角、平均点云密度边缘提取和区域生长阈值自适应的散乱电子元器件分割方法。方法 通过体素化处理、RANSAC算法和统计离群滤波算法对原始点云数据进行预处理，去除大量无关点云;使用欧式聚类算法对预处理结果粗分割得到电子元器件点云簇，以点云簇为阈值设置单元，避免阈值设置不合理的情况;通常边缘点较非边缘点法线夹角更大、邻域点更少，提出通过点云簇平均法线夹角和平均点云密度自适应约束来去除点云簇中边缘点的方法;对去边缘点后的点云簇细分割，根据细分割后点云簇的平均法线夹角进行区域生长阈值的自适应选择，通过改进的区域生长算法将每个电子元器件从点云簇中分割出来。结果 实验结果证明，文中方法分割正确率达97%以上，每10个目标分割耗时约345 ms。结论 提出的方法具有良好的准确性和实用性，分割效果优于传统分割算法，能够准确地将每个电子元器件从复杂场景中分割出来。 |
| 英文摘要: |
| The work aims to propose a segmentation method of scattered electronic components based on edge extraction and adaptive region growth threshold of average normal angle and average point cloud density of point cloud cluster, so as to overcome the difficulty of dividing electronic components in counting of scattered electronic components. The original point cloud data were preprocessed by voxel processing, RANSAC algorithm and statistical outlier filtering algorithm to remove a large number of irrelevant point clouds. The European clustering algorithm was used to roughly segment the preprocessing results to obtain the point cloud cluster of electronic components. The point cloud cluster was used as the threshold setting unit to avoid the unreasonable threshold setting. In general, edge points had larger normal angle and fewer neighborhood points than non-edge points. A method was proposed to remove the edge points in point cloud cluster by adaptive constraints of average normal angle and average point cloud density. Finally, the point cloud cluster with the edge points removed was finely segmented, and the region growth threshold was adaptively selected according to the average normal angle of the point cloud cluster after fine segmentation. Thus, each electronic component was segmented from the point cloud cluster through the improved region growth algorithm. Experimental results proved that the segmentation accuracy of the method was over 97%, and it took about 345 ms to segment 10 targets. The proposed method has good accuracy and practicability, and has better segmentation effect than the traditional segmentation algorithm, which can accurately segment each electronic component from the complex scene. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
渝公网安备 50010702501716号