A Novel Method for Image Recognition Based on Polynomial Curve Fitting
Chao Ji, Yang Xiao-dong, Wei Wang · 2015
Invariant feature is desirable in image recognition and computer vision, while feature matching is limited in the distance or the relative measurement between the image and the model. In this paper, we propose a novel method for image recognition based on polynomial curve fitting. Non-redundant complex moments are derived, that are invariant to the translation, constraint scale and rotation, and complex moments below 4th order are selected as the invariant feature. With the principle of polynomial curve fitting, similarity measurement between the fitting coefficients of the non-redundant complex invariants is calculated to describe the images. Finally it is applied in real ship images and its validation is analyzed. Experimental results presented show that the described method has good stability and can be used as an effective method for image recognition.