A Fourier descriptor based on Zernike invariant moments in spherical coordinates for 3D pollen image recognition

This paper presents a new feature extraction method of Fourier descriptor based on the Zernike moments for pollen images recognition. Firstly, Zernike moments of the image are extracted in 3D spherical coordinates. Secondly, genetic algorithm based on probability is used to filter the Zernike moments to reduce redundant information. Finally the normalized Fourier transform coefficients are calculated as the last feature descriptor. The simulation results on Confocal dataset show that the descriptor can effectively describe the pollen images and is robust to the rotation, translation and scaling of the image.