复杂地面目标前视红外景象匹配算法  

FLIR scene matching algorithm for complex ground target

作　　者：刘婧[1] 孙继银[1] 朱俊林[2] 巨西诺[2] 

机构地区：[1]第二炮兵工程学院四系 [2]中国人民解放军96411部队七室

出　　处：《计算机应用研究》2010年第1期350-352,395,共4页Application Research of Computers

基　　金：国防"十一五"预研项目(513220208)

摘　　要：针对复杂地面目标前视红外图像边缘模糊、背景杂乱无章,无直接可用的同源基准图的情况,研究了一种灰度与形状特征相结合的由粗到精景象匹配算法。在粗匹配阶段,将积分图像引入归一化积相关算法中,并对其计算公式进行了近似,使运算速度大大加快;精匹配阶段,利用实时图与所制备基准图形状相似的特点,采用形状上下文进行匹配。实验结果显示,该算法匹配精度高、速度快,对于复杂地面目标前视红外图像的匹配识别具有一定的应用价值。According to the FLIR image of complex ground target have blurred edge and background clutter,and no base image available,this paper studied an algorithm that combined the gray information with shape feature to achieve precision guidance. In course of coarse matching introduced the integral image into NPROD algorithm and approximated its formula, the efficiency of algorithm had been improved obviously. In the process of fine matching, based on characteristic of the real image and base image had similar shape, the algorithm adopted the shape context descriptor. The experimental results show that the method has highly precision and quickly performance. The algorithm has highly practicality for complex ground garget

关 键 词：景象匹配 归一化积相关 形状上下文 积分图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]