摘要
本文以文献[1]定义的特征不变量为基础,对原文的算法进行了一定的改进.通过对其提取算法的分析,指出其特征不变量存在的缺陷,定义了一种新的特征量,消除了特征不变量的不唯一问题.实验结果和分析表明,本文算法能够住O(n)时间内提取一种特征不变量,并且所提特征不变量具有平移、旋转及比例不变特性,可以很好地用于图像目标识别.
Based on invariant defined in reference [1], the original extraction method is improved in this paper. After the analysis of the algorithm, its limitation is pointed out and a new kind of character invariant is defined. In the original method, if the original character that is defined as the data sequence has a unique maximum or minimum, a maximizing operation and a transposition operation can describe the invariant. By this method, two different characters may result in same invariant. Thinking reversely, in this paper three properties are defined as the rules that an invariant should conform to and they assure a certain relationship between the original character and the invariant. Both experimental results and analysis show that the new character invariant can be computed within O(n) time period by the algorithm in this paper and it has the invariance with respect to translation, rotation and scale. So it can be easily applied to image recognition.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第2期237-241,共5页
Pattern Recognition and Artificial Intelligence
基金
国家博士点基金(2000018509)
