摘要
基于Mean Shift原理,利用高斯滤波方法交替地对数据模型进行缩放,不断剔除噪声,但存在光滑系数不容易确定,滤波时易造成处理数据偏离实际模型或产生变形的问题。鉴于此,提出了一种简单的改进方法,通过交换缩放操作的顺序,使得由平滑系数不合理造成的模型变形尽可能小;并通过实验证明了改进方法的有效性。
On basis of the Mean Shift, λ/μ filtering method is used to remove the noise through scaling the input point cloud or model.Though it can obtain fine results in some applications, yet, the main defect lies in how to determine the value of scaling coefficients to control the deformation after filtering. We proposed a simple improved method by alternately performing zooming in and zooming out operations which could offset the influence of parameter in this paper. And then, we proved the validity of this method by some experiments.
出处
《地理空间信息》
2018年第5期113-114,122,共3页
Geospatial Information
基金
国家自然科学基金资助项目(41371426)
地理国情监测国家测绘地理信息局重点实验室资助项目(2014NGCM09)
关键词
高斯滤波
点云模型
去噪
Gaussian filter
point cloud model
denoising