Watermarking is an important technical way to realize copyright protection of intellectual property. The traditional video watermarking can cause distortion to host video in a certain extent, and has a weak robustness against strong geometric attacks. Firstly, a Nonnegative Matrix Factorization with Sparseness Constraints on Parts (NMFSCP) method is proposed in this paper. Secondly, the NMFSCP is utilized to obtain coefficient matrix with geometric invariance. Finally, the coefficient matrix is used to generate robust zero-watermarking. The experiments indicate that, the proposed zero-watermarking not only can resist conventional video signal processing such as adding noise and filtering, etc., but also shows strong robustness against strong geometric attacks such as rotation, scaling, translation, and cropping, etc., and moreover, can play the role in copyright protection of intellectual property without causing any visual perceptible distortion.