基于改进光流场算法在复杂背景变化图像处理中的应用  

Target Registration in Complex Images Based on Improved Optical Flow Field

作　　者：许磊[1] 戚睿 袁杰 刘研[4] 温乃峰 XU Lei;QI Rui;YUAN Jie;LIU Yan;WEN Naifeng(School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,Heilongjiang,China;Hainan College of Vocation and Technique,Haikou 570100,Hainan,China;Shanghai Aerospace Electronic Technology Institute,Shanghai 201109,China;School of Mathematics,Harbin Institute of Technology,Harbin 150001,Heilongjiang,China;School of Electromechanical Engineering,Dalian Minzu University,Dalian 116600,Liaoning,China)

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,黑龙江哈尔滨150001 [2]海南职业技术学院,海南海口570100 [3]上海航天电子技术研究所,上海201109 [4]哈尔滨工业大学数学学院,黑龙江哈尔滨150001 [5]大连民族大学机电工程学院,辽宁大连116600

出　　处：《上海航天》2019年第4期110-114,共5页Aerospace Shanghai

基　　金：上海航天科技创新基金(SAST2017-104);智能感知与先进控制国家民委重点实验室开放课题

摘　　要：提出了一种基于Demons算法的全自动非刚性图像配准算法,用于复杂结构的图像配准。将一种基于牛顿第一定律的新驱动力集成到原始光流方程中,以重新计算Demons力。在图像空间的任何点处,Demons模型不仅能产生来自图像局部梯度信息的力,以驱使浮动图像中的像素点向参考图像中扩散,而且还可以根据Moore邻域系统中全局灰度变化的趋势产生一种新的“惯性力”,从而促使浮动图像中的像素点更有效地扩散到相应的参考图像中,尤其是当参考图像的梯度场不稳定时。采用具有复杂背景变化的航拍图像来评估所提方法的性能。实验结果表明:该方法能够快速有效配准背景变化复杂的图像,且对图像中的噪声具有稳健性。A fully automatic non-rigid image registration algorithm based on Demons algorithm is proposed for complex structure image registration.In the proposed method,a new force based on Newton's first law motion is integrated into the optical flow equation to estimate the Demons force.At any points in the image space,the Demons model not only generates forces from local gradient information to drive the pixels in the floating image to diffuse into the reference image,but also produces an "inertia force"derived from the global grayscale overall trend in the Moore neighborhood system,which helps the pixels to more properly spread into the corresponding reference image,especially when the gradient field of the reference image is unstable.In the experiment,aerial images with complex structures were used to evaluate the performance of the proposed method.The experimental results show that the method can quickly and efficiently register complex images and is robust to noise in images.

关 键 词：目标跟踪 图像配准 复杂结构 高噪声 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]