随机时分神经计算理论用于视觉曲线自激增强  

Random Time Division Neural Computational Theory for Self-excitation Enhancement of Visual Curves

作　　者：郭雷[1] 刘天明[1] 

机构地区：[1]西北工业大学,西安710071

出　　处：《电子学报》2000年第1期35-38,共4页Acta Electronica Sinica

基　　金：国家自然科学基金!(NO.69625507);国家重点基础研究发展规划;863计划等资助课题

摘　　要：利用空间分布满足曲线规则（位置和定向连续性）的局部刺激之间的相互激励，称为曲线自激，这是发现视觉边界曲线和抑制局部噪声的有效手段．过去的工作均采用并行结构区分的计算方式，曲线自激并没有达到满意的效果．本文提出随机时分动态自激的计算方案，对不同的曲线实施随机时分的搜索、时间一致性滤波、和自激积累等机制．本文给出了实现的原理方案．The mutual excitation between the local stimuli satisfying the rules of curve distribution (position and orientation continuity) called curve self-excitation is a useful method to discover and enhance curves and to inhibit noise. The Present approaches used parallel connection structure division which did not acquire effect. This paper presents the idea of random time division and dynamic self-excitation, for different curves performing random time-division Searches, time coincidence filtering, and Self excitation accumulation . The Principle is given.

关 键 词：神经计算 随机时分 曲线自激 视觉 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]