基于曲率匹配原理的圆形雷达天线识别方法  

Circular radar-antenna recognition method based on curvature matching theory

作　　者：杜天军[1] 黄自力[1] 高升久[1] 刘桂芬[1] 

机构地区：[1]西南技术物理研究所,四川成都610041

出　　处：《红外与激光工程》2008年第4期629-633,共5页Infrared and Laser Engineering

摘　　要：识别圆形雷达天线具有重大的军事价值,但由于探测器本身固有的特性以及军事上极限使用的要求,红外图像普遍存在目标-背景对比度较差、目标边缘模糊、噪声较大等缺点,采用常规方法难以取得理想的检测效果。提出了一种低信噪比红外图像中圆形雷达天线目标的识别方法,基于曲率匹配原理,结合Hough变换,建立了椭圆或类圆形边缘结构的检测方法。利用多个分段圆弧的二维累积空间取代标准变换的高维空间,有效克服了以前方法中时间和存储空间的开销问题。通过建立曲率中心极大值点图,将目标识别问题转化为数据挖掘过程,并给出了数据挖掘规则。最后,利用Dijkstra算法求解由挖掘规则衍生的最短路径问题。理论分析与数值试验验证了方法的有效性。Circular radar-antenna recognition has an important military value. Due to inherent characteristic of infrared detectors and utmost demands in military application,low-SNR infrared images often have many disadvantages such as low target-background contrast,blurred edges,large noises and so on,which make it difficult to get the ideal detected results using traditional target recognition methods. Aimed at low-SNR infrared circular radar-antenna,a detection method of elliptical or quasi-circular edge shape for circular radar-antenna recognition was founded via Hough transform based on curvature matching theory. Multidimensional accumulation spaces of standard Hough transform were replaced by many 2-D accumulation spaces of multi-subsections,which overcame the problem of time expenses and space expenses for standard Hough transform. By establishing maximum points graph of curvature center,target recognition problem was translated into a data mining process,and the data mining rules were presented. At last,the shortest path problem which derived from data mining rules was solved via Dijkstra algorithm. Theoretical analysis and experiment results show that the proposed method is effective for circular radar-antenna recognition in low-SNR infrared images.

关 键 词：圆形雷达天线 HOUGH变换 曲率匹配 DIJKSTRA算法 目标识别 

分 类 号：TN911.73[电子电信—通信与信息系统]