一种改进的不变矩特征与物体识别

摘要

引入了一种改进的不变矩特征表示方法.不变矩特征通过7个不变矩量很好地表征了图像的局部结构信息;通过压缩不变矩的值,得到了更优秀和实用的特征表示.改进的不变矩表示方法进一步包含图像的全局结构信息.实验结果表明,改进的不变矩比不变矩有着更好的识别效果,比现有的物体识别方法有着更大的优势,其易于执行,省去了参数调节,且效率高.%This paper introduces an improved moment invariant, a new representation for recognizing object. Moment invariant captures local structures of an image through the seven moment invariants. Compressing the value of moment invariant results in better representation. Improved moment invariant further incorporates global structures in the image. The experiments demonstrate that improved moment invariant outperforms the moment invariant in several object recognitions. Besides, the improved moment invariant is easy to implement. It has nearly no parameter to tune, and evaluates extremely fast.