人体臀围尺寸岭回归模型的建立

Establishment of the ridge regression model of human hip girth

二维非接触式人体测量系统中,对人体围度尺寸的预测非常重要。经过对人体多个部位尺寸与臀围尺寸进行相关分析,确定TNF指数、臀宽、臀厚、人体正、侧面投影面积为建立臀围回归模型的主要部位。建立臀围的简单线性回归模型,进行共线性检验。为增强回归精度,采用逐步回归方法进一步选取自变量。最终引入岭回归方法解决多重共线性问题,获得良好的预测效果,预测平均误差值为0.52cm。

The estimation of body dimension is very important in the two-dimensional non-contact body measurement system. According to the correlation analysis to many parts′sizes of human body and hip size, it is found that the TNF index, hip width, hip depth and the projected areas of right view and side view are the main parameters for establishing the model of hip girth. Accordingly, the simple multiple linear regression model of hip girth is put forward, and its multi-colinearity is detected. In order to improve the regression precision, the independent variable is further selected using the stepwise regression method. At last, the ridge regression method is introduced to solve the multi-colinearity, which gives very good predicted results with an average error of 0.52 centimeters.