基于三维视觉的服装信息智能提取方法  

Intelligent extraction method for clothing information based on 3D vision

作　　者：于述平[1] YU Shu-ping(Clothing Department,Dalian Art College,Dalian 116600,China)

机构地区：[1]大连艺术学院服装系,辽宁大连116600

出　　处：《沈阳工业大学学报》2018年第4期420-425,共6页Journal of Shenyang University of Technology

基　　金：辽宁省教育厅基金资助项目(UPRP20140031)

摘　　要：针对二维图像处理技术存在鲁棒性弱、补充效果差的问题,提出了基于三维视觉的服装信息智能提取方法.利用三维视觉成像系统,从服装生产视频的重点帧中采集三维视觉服装图像,利用聚类技术优化三维视觉图像的色彩度特征差值切割结果,分别对像素点和切割块进行显著性计算.利用高斯函数进行像素估计,并融合背景估计结果和显著性计算结果提取图像前景,得出服装信息.结果表明,所提方法的鲁棒性强,并具有良好的三维填补效果.Aiming at the problem of weak robustness and poor supplementary effect in the 2D image processing technology,an intelligent extraction method for clothing information based on 3D vision was proposed. Using the 3D vision imaging system,the 3D vision clothing images were collected from the key frames of clothing production video. The difference cutting results of color degree characteristics in 3D visual image were optimized with the clustering technology,and the significance calculation for the pixel points and cutting blocks were respectively carried out. In addition,the pixels were estimated with the Gauss function to extract the image foreground and obtain the clothing information through fusing the background estimation results and significance calculation results. The results showthat the proposed method has strong robustness and good 3D filling effect.

关 键 词：三维视觉 服装信息 智能提取 聚类技术 显著性 高斯函数 像素 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]