基于联想存储级联WTA模型的旋转不变识别

提出一种多目标旋转不变分类识别的新方法，其基本思想是多对一导联想存储模型，为便于光学实现、充分利用系统的时间、空间带宽积，采用了全单极形成，并对权重进行了二值化截取。一个分布阈值用来克服单极联想存储模型０，１状态分布几率不等时存储空量急剧下降的缺点，同时通过巧妙地选取目标所对应的异联想码将异联想模型和ＷＴＡ模型有机地结合起来，提高了整个系统的存储容量和容错性。本文给出了计算机仿真模拟及一个光电混合

Recognition with Rotation Invariance Based on Associative Memory Model Cascaded WTA Model

A new method of multi-objeets rotation invariance classfieation is presented. The principle of object rotation invariance clssification is many-to-one associative memory model. In order to implement easily with optoelectronical techniques and to use the time- and space- bandwidth products sufficiently, unipolar model and a binary interconnection weights are used. To overcome the shortcoming of low storage capacity caused by unipolar model, a distributed thresholding is employed. Because of associate memory model cascaded WTA model by a suitable associate memory codes, the storage capacity and error-tolerant ability of the system are improved. The results of computer simulation are given and an opto-electronics system is presented.