函数链神经网络在移动人体身高测量中的应用  

Application of functional link neural network in height measurement of moving person

作　　者：罗金富[1] 张全法[1] 王庆峰[1] 

机构地区：[1]郑州大学物理工程学院,郑州450001

出　　处：《电子测量技术》2011年第7期91-94,共4页Electronic Measurement Technology

摘　　要：利用视频测量移动人体的身高时,人体图像上单位像素所代表的实际高度与很多因素有关,如摄像机的内部、外部参数以及人体到摄像机的距离等,是1个非常复杂的非线性关系。为了避免标定摄像机参数,提出利用函数链神经网络来逼近这种关系。让已知身高的人在视场中走过,通过视频图像处理得到1组训练样本。利用训练样本对函数链神经网络进行训练,之后即可用来实时地测量移动人体的身高。实验结果表明,该方法不仅易于实现,而且具有较高的精度。When measuring moving person＇s height based on video,the actual height of per pixel on image of moving person is influenced by many factors,such as camera＇s inner and outer parameters,distance from person to camera,etc,and the relation is a very complex nonlinear function.To avoid complicated calibration of camera parameters,functional link neural network is used to determine an approximating function of this relation.Let a height-known person walk through the view field and obtain a group of training samples by video image processing.After training with these samples,the functional link neural network can be used to measure moving person＇s height at real time.Experiment results demonstrate that this method can be realized easily with high accuracy.

关 键 词：函数链神经网络 移动人体 身高测量 视频测量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]