基于Matlab仿真的图像旋转不变识别  

Rotation-invariance Recognition Based on Matlab Simulation

作　　者：冯升同[1] 郭立群[1] 欧阳荣[1] 

机构地区：[1]北京石油化工学院数理系,北京102617

出　　处：《实验室研究与探索》2011年第1期5-8,17,共5页Research and Exploration In Laboratory

基　　金：国家自然科学基金项目(60377003);北京市自然科学基金项目(4042010)

摘　　要：从相干光学信息处理系统出发,以Vander Lugt相关器为基础,利用Matlab软件进行光学相关识别仿真。为了实现图像旋转不变识别,利用圆谐函数二级分量,采用合适而精确的数值算法,综合出圆谐滤波器,并改进成纯位相滤波器。将该滤波器制作成匹配滤波器,用于光学图像识别。仿真结果表明,该滤波器能够识别任意旋转角度的同一类图像,很好地解决了图像旋转不变识别问题。在此基础上,为了多类识别和减少匹配滤波器的个数,引入了圆谐-综合判别函数滤波器。该滤波器能够很好地完成多类目标的旋转不变识别。In this paper,image recognition is carried out based on Vander Lugt correlator using appropriate algorithm and Matlab simulation.A new approach,called circular harmonic expansion,is presented to achieve rotation-invariance recognition.Second order circular harmonic component is selected,pure-phased,synthesized as matched-filter（called as CH filter）.The input image rotated by any angle can be discriminated by the second order circular harmonic component.In order to reduce the number of filters a new filter,called CH-SDF filter,is synthesized by combining circular harmonic expansion with SDF（Synthetic Discriminant Function）.Using this filter,multi-class targetrotated by any angle can be discriminated.

关 键 词：数值分析 Vander Lugt相关器 旋转不变识别 圆谐函数 MATLAB仿真 综合判别函数 

分 类 号：O439[机械工程—光学工程]