基于特征矩阵和神经网络的铝型材整体壁板外形识别  

Shape Recognition of Aluminum Wallboard Profile Based on Neural Net and Feature Matrix

作　　者：王匀[1] 刘全坤[2] 

机构地区：[1]燕山大学机械工程学院 [2]合肥工业大学学术委员会

出　　处：《中国机械工程》2003年第15期1326-1328,共3页China Mechanical Engineering

基　　金：国家自然科学基金资助项目 ( 5 0 0 75 0 2 2 )

摘　　要：针对铝型材整体壁板的外形特征和设计要求 ,提取出 9种典型壁板并构建出对应的壁板特征矩阵和目标矩阵。采用神经网络分别对带与不带噪声的整体壁板特征识别进行了研究、训练和测试 ,测试结果显示 ,采用矩阵构造的神经网络对壁板特征识别继承了神经网络的优点 ,可扩充性好 ,解决了整体壁板挤压模具According to shape features of wallboard profile and design specifications, nine typical wallboards are extracted, and the feature matrix and target matrix are constructed respectively. Then, by means of neural net, wallboard feature recognition with and without noise is studied, trained and tested step by step. The test results show that neural net combined with matrix can be applied conveniently and effectively in feature recognition of aluminum wallboard profile, with the advantages of neural net and extensibility.Furthermore, the bottleneck of pre-processing occurred in extrusion die CAD expert system for wallboard can be solved by this method.

关 键 词：带筋整体壁板 特征矩阵 目标矩阵 神经网络 特征识别 噪声 

分 类 号：TG316[金属学及工艺—金属压力加工]