Object Recognition Based on Improved Zernike Moments and SURF

Abstract

Since single global or local features can only describe objects partly or unilaterally that may lead to a low recognition rate, object recognition algorithm based on improved Zernike moments and Speeded-up Robust Features (SURF) is proposed. Firstly, the seven improved Zernike moments and SURF descriptor of objects are extracted, and then the two features are fused together with the weights in term of their contribution to the recognition. Secondly, Euclidean distance is calculated to determine the recognition result. Finally, the performance of algorithm is tested by some image data. Experimental results show that the proposed method is robust to scaling transformation, rotation change and noise variation. Compared with the other three ones, the results show that the proposed method has better recognition performance.