Bootstrap model order selection of Zernike polynomial expansion for classification of rice

Zernike moment is used as image representation for rice grain image due to its rotational invariant property. However, the computation of Zernike moments takes a lot of time. Therefore an optimal modelling of rice grain image using Zernike polynomials is applied to reduce the computation time while maintaining the high accuracy. Accurate modelling of rice grain image with Zernike polynomials involves the selection of polynomial expansion order based on the captured image. Bootstrap model order selection method based on the resampling residuals is used to select the optimal model order of rice grain image according to a prediction criterion. The proposed method is easy to implement and detail knowledge of the distribution of rice grain image and modelling error are not necessary during selection process.