Traditional method often need complicated mathematical model,and the neural network method can't tell the perspective-projection-transformation matrix between the world 3-D points and the corresponding 2-D image pixels.A new calibration method for structured light system is propoed in this paper,which using the BP neural network and the linear calibration method and utilize the multilayer feed-forward neural network to compensate the nonlinear error of structured light system brought by lens distortion,etc.The experiments results show that the method we proposed is feasible,effective,and has high anti-noise ability.