基于SR300体感器人体扫描点云的去噪方法

来源期刊:中南大学学报(自然科学版)2018年第9期

论文作者:梁晋 张铭凯 刘烈金 梁瑜 王晓光

文章页码:2225 - 2232

关键词:SR300;人体扫描;点云去噪;深度图;双边滤波

Key words:SR300; body scanning; point cloud denoising; depth image; bilateral filtering

摘    要:为克服传统三维人体扫描系统体积大、线路复杂、成本高的缺点,采用SR300体感器获取三维人体点云,针对体感扫描点云噪声大的问题,提出由外而内,分步去除噪声的系统方法。此方法无需点云间的拓扑关系,可以直接对三维散乱点云进行处理。首先,设计体感扫描实验平台获取人体点云并采用圆柱体分割法进行点云预处理;接着,提出并实现2种体外离群点去噪的算法,即统计分析法和球半径法,将这2种算法结合使用去除不同类型的离群点;最后,调整人体深度图的双边滤波去噪算法的权重因子到合理的范围内,分别以单幅点云和全身点云为对象,去噪时间分别仅为传统双边滤波的7.52%和3.69%。借助VS2010开发环境,并调用PCL库进行大量算法实验。研究结果表明:本文方法能够有效去除内外噪点且能保持边缘特征,获得较好的三维人体点云结果。

Abstract: In order to avoid the disadvantages of traditional 3D body scanning system such as large volume, complex circuit and high cost, SR300 was used to get 3D body point cloud. To deal with the problem of large noise in the scanning data, a step-by-step denoising method featuring outside-in was proposed. This method could deal with scattered 3D point cloud directly without the topological relation. Firstly, the somatosensory scanning platform was designed to get the human body point cloud and the point cloud preprocessing was carried out by using the cylinder segmentation method. Then, two algorithms named statistical analysis method and ball radius method were proposed and implemented to remove outliers of different types. Finally, bilateral filtering algorithm for human depth image was tested to make sure the weight factors were suitable. Taking the single point cloud and the whole body point cloud as examples, the denoising time was only 7.52% and 3.69% of the traditional bilateral filtering, respectively. A lot of algorithm experiments were taken by using the PCL library under the VS2010 development environment. The results show that the proposed method can effectively remove the internal and external noise with the edge feature maintained, and can get better 3D body cloud data.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号