三维轴对称目标的简化识别  

Simplified Recognition of 3D Axes-Symmetric Object

作　　者：易亨瑜[1] 胡晓阳[2] 张凯[2] 吕百达[1] 

机构地区：[1]四川大学激光物理与化学研究所,四川成都610064 [2]中国工程物理研究院应用电子学研究所,四川绵阳621900

出　　处：《中国激光》2005年第4期466-470,共5页Chinese Journal of Lasers

基　　金：中国工程物理研究院应用电子学研究所所长基金(SJ0111004)资助项目。

摘　　要：基于对空中轴对称目标识别的需要,根据目标投影变换和姿态测量,利用匹配滤波和坐标变换,提出一种简化的识别方法。选择合适的目标投影图像作为识别模板,并用计算全息的方法编制成匹配滤波器;在目标的三维姿态被探测后,根据其方位角和俯仰角,对滤波器进行坐标变换,再利用光学相关方式识别目标的投影图像。通过数值模拟,分析了不同姿态下二维任意目标的识别情况,以及三维轴对称目标的模板选择及其识别效果。分析表明,利用这种简化识别方法,只需将目标在45°和90°俯仰角下的投影图像制成识别模板,并进行组合,即可得到目标的最小相关系数为0.4472,远大于飞机干扰物与模板的最大相关系数0.06。In order to recognize axes-symmetric targets in the sky, based on target's projection transform and measurement of its three dimensional (3D) orientation, a simplified approach is put forward in this paper, which is realized in the way of matching filter and coordinate transform. An appropriate projected image of a target is chosen to make a discriminating template and is encoded to a matching filter by using computer-generated hologram. After 3D orientation of the target is explored, coordinate transform of the matched filter can be carried out according to its pitching angle and azimuth angle, and then this target is detected in the way of matching correlation. By this simplified detection method, it is studied how to select template for 3D axes-symmetric object, and recognized effect is simulated for 2D arbitrary target and 3D axes-symmetric object at random angle. With the combination of only two filters, which are made from target's projection of 45° and 90°, the minimum correlation coefficient between the target and template in various pitching angle is 0.4472 which is far more than maximal correlation coefficient of 0.06 between plane interference objects and template.

关 键 词：信息光学 目标识别 坐标变换 计算全息 傅里叶频谱 三维姿态 

分 类 号：O438[机械工程—光学工程]