用于成像引信图像识别的局部特征提取技术  

The Study of Feature Extraction Used to Recognize Incomplete Image for Imaging Fuze

作　　者：王炎[1] 郑链[1] 王克勇[1] 

机构地区：[1]北京理工大学机电工程学院,北京100081

出　　处：《红外技术》2001年第5期20-23,28,共5页Infrared Technology

摘　　要：根据成像引信的扫描成像机理 ,在详细分析了引信目标图像特性的基础上 ,论述了四种适用于引信目标图像识别的局部特征 :角点、轮廓曲线段形状特征、线矩和弯曲能量 ,并提出了一种子块特征的提取算法。最后 ,分析了在应用局部特征进行图像识别时遇到的问题并指出了解决方法。本文的研究对于其他不完整图像识别问题 ,如工件识别、机器人视觉、场景分析、医学图像的解释、飞行器的识别等 。According the imaging principle of imaging fuze, this paper analyzes the characteristics of fuze target image in detail. Then, four features used to recognize incomplete target image, including corners, the feature of contour curve shape, line moment and bending energy, are discussed, and a new feature so called block feature is presented. Finally, the problem of recognizing image based on these features is described and the method to it is introduced. The study of this paper is also very useful for other image recognition fields, such as workpiece classification, machine vision and scene analysis.

关 键 词：局部特征提取 图像识别 神经网络 成像引信 

分 类 号：TJ43[兵器科学与技术—火炮、自动武器与弹药工程] TP391.41[自动化与计算机技术—计算机应用技术]