基于小波变换的金属断口模式识别与分类

提出了一种基于小波变换的金属断口模式识别分类方法。该方法采用Daubechies四点小波和标准的金字塔结构小波变换对粘口图像进行二级小波变换，将小波变换后各个频带输出的L1范数、能量、熵作为断口分类的特征，并根据特征本身的离散程度对特征进行加权处理，采用线性最小距离分类器，对等轴塑坑、腐蚀疲劳、河流花样、拉长塑抗、韧性疲劳和韧性沿晶等六种典型的断口进行分类。实验结果表明，这种方法可以对金属断口模式进行准确的识别与分类。

RECOGNITION AND CLASSIFICATION OF METAL FRACTURE SURFACE MODELS BASED ON WAVELET TRANSFORM

A new method of recognition and classification of metal fracture surface models based on wavelet transform is presented in the paper. Two-level wavelet transform was applied to fracture surface images by selecting four-point Daubechies wavelet and adopting standard pyramided structure wavelet transform. L1 norm, energy and entropy of each output wavelet transform frequency band are regarded as characteristics of fracture surface classification, and these characteristics are weighted according to their own disperse degrees. Linear minor distance classifier was used to classify six typical fracture surface models: equiaxed dimples, eroded fatigue, stream design, elongated dimples, tenacity fatigue striations and intergranular cracking. Experiment results show that the metal fracture surface models can be properly recognized and classified by means of the proposed method.