论文部分内容阅读
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。
Recognition of circular radar antenna has great military value, but due to the inherent characteristics of the detector itself and the military requirements of the extreme use of the common infrared images of the target - the background contrast is poor, the target edge fuzzy, noise and other shortcomings, using conventional Method is difficult to achieve the desired test results. A method to identify circular radar antenna in low SNR image is proposed. Based on curvature matching principle and Hough transform, a detection method of elliptical or quasi-circular edge structure is proposed. Using the two-dimensional cumulative space of multiple segmented arcs instead of the high-dimensional space of the standard transform, the problem of the overhead of time and storage space in the previous method is effectively overcome. By establishing the point map of curvature center, the problem of target recognition is transformed into the data mining process, and the data mining rules are given. Finally, the Dijkstra algorithm is used to solve the shortest path problem derived from mining rules. Theoretical analysis and numerical experiments verify the effectiveness of the method.