基于径向基函数网络的图像三维恢复 技术在雕刻加工中的应用

谢明红

doi:null

您当前的位置：

首页 >

文章列表页 >

基于径向基函数网络的图像三维恢复技术在雕刻加工中的应用

更新时间：2020-08-12

- 基于径向基函数网络的图像三维恢复技术在雕刻加工中的应用
- Application of 3D image recovering technology based on radial basic function network to engraving
- 光学精密工程 2007年15卷第1期页码：117-123
- 作者机构：
  
  华侨大学机电及自动化学院, 福建泉州 362021
- 作者简介：
- 基金信息：
- DOI：
  中图分类号： TP391.4
- 收稿日期：2006-04-16，
  
  修回日期：2006-11-12，
  
  网络出版日期：2007-01-30，
  
  纸质出版日期：2007-01-30
- 稿件说明：
移动端阅览
谢明红. 基于径向基函数网络的图像三维恢复技术在雕刻加工中的应用[J]. 光学精密工程, 2007,15(1):117-123.

XIE Ming-hong. Application of 3D image recovering technology based on radial basic function network to engraving[J]. Optics and precision engineering, 2007, 15(1): 117-123.
谢明红. 基于径向基函数网络的图像三维恢复技术在雕刻加工中的应用[J]. 光学精密工程, 2007,15(1):117-123. DOI：

XIE Ming-hong. Application of 3D image recovering technology based on radial basic function network to engraving[J]. Optics and precision engineering, 2007, 15(1): 117-123. DOI：

摘要

基于图像的三维恢复技术的目的是采用单幅图像中物体表面的明暗变化来恢复其表面各点的相对高度

从而对三维物体进行曲面恢复。提出了一种基于径向基函数网络模型进行图像三维恢复的新方法

建立了径向基函数的网络模型

并构造出一个曲面公式

其中输出对两个输入变量的偏导数满足反射函数方程

反射函数方程中的灰度值已知

利用反射函数与灰度值之间的误差作为约束条件调节网络权因子、径向基函数中心和宽度

使得构造的曲面中每一点都满足反射函数方程

从而拟合整个曲面。该方法无需光滑约束和积分约束条件

求出的解是一个连续解。通过不同方法对合成图像圆球和花瓶的实验表明

该算法恢复的曲面的高度点最大误差精度提高1倍到4倍

平均误差精度提高5倍到20倍

恢复的曲面具有良好的连续性和光滑性

它不仅可以恢复图像中各网络点的高度

而且可自动内插网络点之间任意点

适合于任意反射模型图像的曲面反求

恢复出的曲面便于进行后置处理生成刀具路径

并进行雕刻加工。

Abstract

The purpose of the 3D image recovery technique is that the height of a surface is recovered by shape from shading on single image to generate the curve face of 3D object. This article puts forwards a new method of 3D image recovering based on the Radial Basis Function(RBF) network model. The curve face formula has been constructed

in which the partial derivates of output to two inputs satisfy to the equation of reflection function and the gray value is known. According to the inhibit condition between reflection function and the gray value

adjusting power factor formula and the center and width of the radial basis function to satisfy the equation of reflection function for every point of the curve face constructed

the whole curve face is recovered. This method does not need smooth control and integral control

and can get a continuous solution. The experimental results of different methods on synthesizing for example ball and vase show that the precision of max error by this method is enhanced 1~4 times

and the precision of average error by this method is enhanced 5~12 times

and the recovered curve face has consecution and smoothness. It can also recover the height of each network point in the picture

automatically interpolate any network point. It is shown that this method is suitable for the curved face of the picture for any reflection model

easy to post handle and make tool path from curve face recovered and to be engraved.

关键词

Keywords

references

. IKEUCHI K, HORN B K P. Numerical shape from shading and occluding boundaries[J]. Artif. Intel., 1981,17(8):141-184.

. HORN B K P. Shape from Shading: A method for obtaining the shape of smooth opaque object from one view . MAC TR-79 and AI-TR-232,Artificial Intelligence Laboratory,MIT,November 1970.

. DUPUIS P, OLIENSIS J. Direct method for reconstructing shape from shading . IEEE Proc.Comput. Vision & Pattern Recogn., 1992:453-458.

. BICHSEL M, PENTLAND A P. A simple algorithm from shape from shading . IEEE Proc.Comput. Vision & Pattern Recogn., 1992, 459-465.

. LEE C H, ROSENFELD A. Improve methods of estimating shape from shading using the light source coordinate system[J]. Artif. Intel.,1985,26(2):125-143.

. TSAI P S, SHAH M. Shape from shading using linear approximation[J].Imag. & Vision Comput. J., 1994, 12(8):487-498.

. HOU C, PETTET M W, VILDAVSKI V Y,et al.. Neural correlates of shape-from-shading[J].Vision Res., 2006,46(6-7): 1080-1090.

. NEFS H T, KOENDERINK J J, KAPPERS A M L. Shape-from-shading for matte and glossy objects[J]. Acta Psychol., 2006,121(3):297-316.

. 李睿凡,郭学彬,瞿安连. 基于奇值分解的三维光学切片显微图像恢复算法研究[J]. 光学精密工程,2001,9(3):212-215. LI R F, GUO X B, QU A L. Image restoration algorithms based on singular value decomposition of 3-D optical sectioning microscopy[J]. Opt. Precision Eng., 2001,9(3):212-215.(in Chinese)

. 俞鸿波,赵荣椿,廖熠,等. 基于平均差分计算的任意光照环境下三维表面重建算法[J]. 西北工业大学学报2005,23(3):303-306. YU H B,ZHAO R C,LIAO Y,et al..A shape reconstruction method for any light direction based on weighted difference scheme[J]. J. Northwestern Polytechn. Univ., 2005,23(3):303-306.(in Chinese)

. 尹爱军,秦树人,周传德. 基于支持矢量机的从明暗恢复形状方法[J]. 机械工程学报,2006,42(7):186-190. YIN A J,QIN S R,ZHOU C D. Method on shape from shading based on support vector machine[J].Chinese J. Mech. Eng., 2006,42(7):186-190. (in Chinese)

. 李鸣鸣,龚振邦,欧阳航空,等. 实验数据RBF神经网络模型中噪声的处理方法[J]. 光学精密工程,2003,13(增):227-231. LI M M, GONG ZH B,OU-YANG H K, et al.. Strategies to the noise contained in experimental data in RBF neural network model [J]. Opt. Precision Eng., 2003,13(Supp.):227-231.(in Chinese)

浏览量

648

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

LightDiffu-DCE：基于光照强度扩散的低光照图像增强

多级图特征融合引导相机位姿回归

基于机载的红外动态目标视频实时超分辨率重建

基于知识蒸馏的Transformer视觉跟踪器

3D-CNN与Transformer混合结构的高光谱图像空谱联合分类