基于信息融合的运动图像姿态轮廓提取研究  

Research on Extraction of Pose Contour of Moving Image Based on Information Fusion

作　　者：宋建华[1] 黄达武[1] SONG Jian-hua;HUANG Da-wu(Jiangxi Normal University,Nanchang Jiangxi 330022,China)

机构地区：[1]江西师范大学,江西南昌330022

出　　处：《计算机仿真》2020年第9期395-398,473,共5页Computer Simulation

基　　金：江西省社科规划(15TY03)。

摘　　要：现有的运动图像姿态轮廓提取结果清晰度低,且细节丢失严重。为此,利用信息融合技术设计了一种运动图像姿态轮廓提取方法。首先利用信息融合过程对具有运动信息的图像进行统计分析,判断运动区域位置,经过形态学处理削弱噪声后获得初始运动姿态轮廓;然后利用HIS颜色模型对图像进行分割,得到源图像色彩分割图。由于运动过程中存在场景统计误差,会对运动分割结果产生影响,因此需在姿态轮廓图像的基础上引入色彩力,找回目标内部的所有颜色区域,并预测姿态轮廓边缘,得到目标的宽和高;最后利用序列影像轮廓线提取当前运动图像的轮廓线,从而得到运动图像的姿态轮廓。实验结果表明:经上述方法提取的运动图像姿态轮廓清晰,细节丢失量较少,证明所提方法具有较强的实用性能,能够有效寻找出所需的运动目标轮廓。In this article,a method to extract the pose contour of motion image was designed by information fusion technology.Firstly,the information fusion was adopted to perform the statistical analysis on the image with motion in⁃formation,so as to determine the motion area,and then the initial posture contour of motion after the noise reduction was obtained.Secondly,HIS color model was used to segment the image and obtain the color segmentation map of source image.However,the scene statistical error in motion might influence the motion segmentation results,it was necessary to introduce color force on the basis of the pose contour image,and thus to retrieve all the color areas from the target and predict the edge of contour.And then,the width and height of target was obtained.Finally,the outline of the current motion image was extracted by the contour of image sequence,and the pose contour of motion image was obtained.Simulation results show that the motion image extracted by the proposed method has clear pose contour and less detail loss.This method can effectively find the required contour of motion target,with strong practicality.

关 键 词：信息融合 运动图像 姿态轮廓提取 色彩力 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]