摘要
图象匹配作为一个优化问题，过去一直为许多人所研究，但多数使用传统的方法，这些方法随着结点数量的增多，计算量也随着增大。Ｈｏｐｆｉｅｌｄ网络以其独特结构和计算能力成功地解决了这个问题。但是Ｈｏｐｆｉｅｌｄ网络能量函数极易陷入局部最小点。本文提出的改进算法能收敛到全局最小点，从而解决了这个问题。实验结果表明改进算法适合于任何匹配问题。
Image matching as an optimum problem,was researched by many people using some traditional methods.But it needs more time to calculate with its nodes increasing.The Hopfield network model proposed by Hopfield solves this problem successfully by its unique calculating structure and capability,but the energy function for network is easy to convergence on a local minimum.The modified algorithm proposed in this paper can convergence on a global minimum.The experimental results show this modified algorithm is very available for any matching problems and the results are satisfied.
出处
《华东船舶工业学院学报》
1996年第2期45-49,共5页
Journal of East China Shipbuilding Institute(Natural Science Edition)
基金
船舶行业预研基金项目
关键词
神经网络
模式识别
图象匹配
HOPFIELD网络
neural networks
pattern recognition
integration / Hopfield model
energy function
