视觉初级皮层区超柱结构的自组织适应模型

在视皮层的初级皮层区中，有许多非常规整的柱形功能结构，它们形成的局域网络具有抽取视图像中最基本的微小特征的能力．视皮层的等级组构为许多心理现象提供了生理解释，这不仅对模式识别、计算机视觉有重要的价值，而且对人工智能系统知识的获取和知识表示都具有非常重要的意义．通过构造一个金字塔状的神经网络层次模型，对模拟视网膜的输入点阵信息进行逐级加工．在计算实现上使用了无监督的自组织语义映射结构，以使模型具有一定的自适应能力．强调的是认知心理功能的计算依托和生理基础．

A SOM network model for feature extraction by hyper columns architecture of primary visual cortex

The primary visual cortex consists of six layers of neural cells. Apart from discovering that there are three types of cells in the visual cortex, the cells, specifically the simple and complex cells, are arranged in a very orderly way. They are organized in narrow columns that cut across the six layers of the cortex, each column constituting cells with the same line orientation. In the language of pattern recognition, they works as feature extraction units. The architecture and mechanism of natural biological system, which processing visual image, are taken as a model for the development of artificial electrical system. The self\|organizing mapping algorithm are used to train the network for the acquisition of sensitivity to special linear feature. All these research are significant for knowledge representation, pattern recognition and computer vision, which is based on cognitive neuropsychology.