Research Of A New Method Of The Curves Similarity Judgment

Curves similarity judgment in 2D and 3D spaces is a central problem in computer vision, pattern recognition, and protein structure prediction. We often use the similar function definition law and the characteristic value law to judge the similarity of curves. Due to the popularity of computer and Internet, as a kind of behavioral feature, signature has advantages of easy acquirement and sharing. It turns out one of the most popular features in identity recognition and verification.In this paper, we apply the new curve similarity judgment to the online handwriting signature verification. We pick up the key characters vertices and rock bottom from the special discrete points in signature curves, and then according to the new definition of the curve similarity, we do researches on vertices and rock bottom via the discrete Fr'echet as the distance measure. At last we use the new method of the curve similarity judgment to judge the similarity of the signature curves, and then we can identify the real signature and the spurious signature. The main tasks of this paper are listed below:1: A new curves similarity definition is proposed. We often use the similar function definition law and the characteristic value law to judge the similarity of curves. The characteristic value law mainly uses the neural network or wavelet analysis to compare the characteristics of the parameters, while the similar function definition law uses a similarity judgment theorem or a similarity definition. The effort of the characteristic value law is not very satisfactory, while the similar function definition law need express the curves into a function. The new curves similarity definition do researches on discrete points, it doesn't need curve fitting and implies translation and the stretching to the curves.2: Discrete Fr'echet distance as the distance measure. The efforts of the judgment usually depend on the distance measure of the similar function definition law. The normal distance measures include the Hausdorff distance which used for matching two point sets and the Fr'echet distance which used for the curves similarity, but they have little efforts on the curves that are constituted with discrete points. We have significantly effects on using the discrete Fr'echet distance as the distance measure.3: One kind of new multinomial solution algorithm is proposed. There is an Fr'echet alignment definition in the discrete Fr'echet distance definition, the solution of it is one kind of NP difficult problem, and actually it reduces the searching space. We discuss the reduction of the searching space in detail.The fruits achieved in this paper will apply new ideas on the online signature verification, and it promotes the development of the online signature verification field. Compared with the other signature verification algorithms, it has distinct advantage. It has lower degree of redundancy in computer running compared with the method that matching the whole curves. Experiments show that this method improves the efficiency of the algorithm, achieved good results, which also confirms the validity in the signature verification. Finally we choose 540 text signatures, the successful matching rate is 91.3 % ,the wrong reacceptance is 5.92 % and the wrong reject rate is 2.78%.