基于循环频谱特征提取的路面裂缝分类方法

Pavement cracks classification method based on cyclic spectrum feature extraction

提出基于循环频谱特征提取的路面裂缝分类方法.首先从图像像素获得2个1维信号,然后通过累积傅里叶变换计算每个信号的频谱相关函数,最后根据不同区域频谱相关函数的能量和标准差对路面图像进行分类.实验结果表明:与基于Gabor小波、离散小波、双树复小波及双树旋转复小波变换的特征提取分类方法相比,所提方法得到了较高的分类正确率。

Pavement cracks classification method based on cyclic spectrum feature extraction was proposed. Firstly,two one-dimensional signals were obtained by pixels from each image. Then,the spectral correlation function of each signal was calculated by the accumulated Fourier transform. Finally,the pavement crack images were classified by the energy and standard deviation of the spectral characteristics of different regions of the correlation function. Compared with the classification method of feature extraction based on Gabor wavelet,discrete wavelet,double tree complex wavelet and double tree rotating complex wavelet transform,the experimental results showed that the proposed method could get high classification accuracy.