一种新的细胞神经网络阴影检测器的研究  

A NEW KIND OF SHADOW DETECTOR BASED ON CNN-UBN

作　　者：汪海明[1] 郭仕德[2] 余道衡[1] 

机构地区：[1]北京大学电子学系视觉与听觉信息处理国家重点实验室,北京100871 [2]北京大学遥感与地理信息系统研究所,北京100871

出　　处：《电子与信息学报》2003年第11期1499-1504,共6页Journal of Electronics & Information Technology

基　　金：国家863项目(No.2001AA136030)资助课题

摘　　要：该文深入研究了一种新的二维细胞自动机(CA),找到了几种新的算法规则可以用来实现字符的粗化处理和阴影检测,并且用这些规则设计了几种新的细胞神经网络,文中详细介绍了这些算法规则的布尔代数表达式和细胞神经网络学习算法。仿真结果证明了这种新的细胞神经网络是简单而有效的,同时也证明了可以用 CA规则来设计新的细胞神经网络,为细胞神经网络的设计找到了一种新颖有效的方法。In this paper, a new kind of two-dimensional Cellular Automata (CA) is studied. Several algorithmic rules are discovered that can be used to design new Cellular Neural Networks (CNN) which can implement shadow detector and overstriking for character recognition. Their Boolean expressions and learning algorithms are introduced detailedly. The results of computer simulation confirm that this new approach is simpler and more effective than that given in the literature. Furthermore, these results also confirm that new CNN can be designed with CA rules and find a novel method to design CNN.

关 键 词：细胞神经网络 CNN 细胞自动机 CA 阴影检测器 复值权重 非线性处理器阵列 图像处理 

分 类 号：TN911.73[电子电信—通信与信息系统] TP183[电子电信—信息与通信工程]