对数极坐标变换识别算法在成像制导中的应用  

The Application of Target Recognition Algorithm Based on Log-polar Transformation for Imaging Guidance

作　　者：王立[1] 李言俊[1] 张科[1] 

机构地区：[1]西北工业大学航天工程学院导航制导教研室,西安710072

出　　处：《宇航学报》2005年第3期330-333,共4页Journal of Astronautics

基　　金：高等学校博士学科点专项科研基金资助(20020699014);航天科技创新基金资助(N4CH008)

摘　　要：图像匹配是成像制导系统中常用的目标识别算法,但实时目标图与基准目标图存在的尺度差异以及旋转不可避免,这也导致失配率提高。针对这个问题,引入对数极坐标变换运算,根据对数极坐标图像的尺度、旋转不变性提出了新的目标匹配识别算法,一方面保证了在尺度变化及旋转下的高匹配率;另外新算法结合了变换图目标区域的轴向投影统计策略,保证了匹配运算量的降低;文中仿真实例验证了算法的优势与有效性。Image matching is often used in target recognition for imaging guidance, but the scale change and rotation of target make this ineffective. Aiming at this question, we introduce a novel recognition algorithm based on log-polar transformation, which provides an invariant representation of the object's rotation and scaling. Collecting the distance axis projection, this algorithm not only work well when the target has the scale change or rotation, but also has the low operation cost. Indeed the validity of this method is verified in the paper.

关 键 词：对数极坐标变换 目标识别 轴向投影 

分 类 号：TN911.7[电子电信—通信与信息系统]