FRACTAL DIMENSION SPECTRUM AS AN INDICATOR FOR TRAINING NEURAL NETWORKS

被引：0

作者：

Crisan, Daniela Alexandra ^{[1
]}

Dobrescu, R. ^{[2
]}

机构：

[1] Romanian Amer Univ, Bucharest, Romania

[2] Univ Politehn Bucuresti, Dept Control Engn & Ind Informat, Bucharest, Romania

来源：

UNIVERSITY POLITEHNICA OF BUCHAREST SCIENTIFIC BULLETIN SERIES C-ELECTRICAL ENGINEERING AND COMPUTER SCIENCE | 2007年 / 69卷 / 01期

关键词：

character recognition; fractal dimension; fractal spectrum; boxcounting algorithm; neural network;

D O I：

暂无

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

In this paper an original system for recognizing scanned characters is proposed. The system consists in a neural network trained on fractal features of characters. The advantages of this new method consist in its ability to recognize characters represented in gray levels and its invariance to scaling and rotating forms. For fractal feature we used the fractal dimension spectrum.

引用

页码：23 / 32

页数：10