改进的统一Hu矩在图像目标识别中的应用  

Improved United Hu Moments and Their Application to Image Recognition

作　　者：张琪[1] 屈严[1] 董志芳[1] 

机构地区：[1]东南大学电子科学与工程学院,南京210096

出　　处：《电子器件》2012年第6期679-682,共4页Chinese Journal of Electron Devices

摘　　要：以图像的平移、旋转及尺度变换为研究对象,结合Hu提出的基于区域的不变矩和Chen提出的基于边界的不变矩,进一步改进了统一Hu矩,并证明了改进后的不变矩在离散状态下的不变性。其中针对图像旋转的特点引入权向量的改进发法,并通过实验证明了这种方法极大提高了图像在旋转变换后的识别命中率,使其具有一定的应用前景。By researching on translation, rotation and scaling transformation of images, combined with Hu＇s invariant moments based on regions and Chen＇s invariants moments based on boundary, a new kind of improved united Hu moments is proposed, and its scale factor in discrete is also proved. According to features of image rotation, weight vector is introduced, which is proved to increase the hit rate in image recognition significantly and will make a promising prospect in the future.

关 键 词：目标识别 不变矩 Hu不变矩权向量 相似度匹配 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]