ABSTRACT
There appeared many kinds of sub-standard milk powder in recent years in China, which are damaged to people's life. Some illegal industrialist got huge profits by making sub-standard milk powder, which is mainly composed of cheap food ingredients like sugar, starch and milk flavor powder to reduce the cost of raw materials. Sub-standard powder doesn't look much different from good milk powder but its nutritional value was less than that of flour. Traditional chemical method to detect milk powder is timeconsuming and complex. Otherwise chemical methods need consume chemical and destroy the estimate sample. So it's unfit for the national supervise department to execute the inspect task locally.
Near infrared spectrometry is fast developed after ninety in twentieth century in China because this technology is fast, green (having no chemical-consuming) and nondestructive. The technology has good performance especially in on-line detection and locale detection. Support Vector Machines (SVM) is a new machine-learning algorithm and a new technology to data classification. SVM is built based on Vapnik-Chervonenkis (VC) Dimension and Structural Risk Minimization (SRM) of statistical learning theory. It is good at solving the problems, such as small-scale, nonlinear, high dimension and local minimum. SVM becomes a new focus in machine learning and applied successfully in face identification and text identification.
