Application of neural network on distortion correction based of standard grid

Image interpretation must obtain accurate position of image, but those questions, brought by angle of imaging equipment placed and camera lens deviation, induce nonlinear geometry distortion of image. The paper proposes the method that utilizes neural network to achieve the correction of geometry distortion of standard grid on the basis of grid image for extracting available exact information of image. The method makes use of least square procedure to obtain measured data and distortion data on the basis of grid plate centre, and uses BP neural network to gain correcting model and then obtain true value of optional point on image. It overcomes the shortcoming of interpolation that can not describe nonlinear distortion, and precision is less than 0.001 mm.