Wavelet modulus maximum method can be used to recognize the features and restrai n the noise. However, some important local features would disappear in the fines t scales when using this method to recognize the whole crosssection data. The concept of characteristic scale factor was proposed to find out the differences of the features contained in the crosssection data. When the characteristic sc ale factor was larger, the numbers of the decomposable scales would become more and more, and vice versa. The crosssection data can be divided into several se ctions to be recognized respectively when the features of the crosssection dat a differed considerably. And the feature points of each section would be integra ted finally. The experimental results show that the shapescale factor can refl ect the information contained in the signal well and guide features recognition. This recognition method can ensure the integrity of the feature numbers effecti vely. 