基于二维连续小波变换的织物疵点检测

同正常织物纹理比较,疵点区域由于其纹理不规则及变形而导致不同的局部纹理特征。利用二维连续小波特征,能在时域和频域上对织物图像同时实现任意尺度和旋转角度的变换。通过纹理模型和频谱分析,确定出最优的变换尺度和旋转角度,并由预先确定的全局阈值从小波变换系数的模中进行疵点的分割。实际疵点的检测结果表明该方法是可行的。

Fabric Defect Detection with Two Dimensional Continuous Wavelet Transform

The defects have different local texture characterization due to their irregularities and deformities in the textile fabrics. In order to defect the local texture variation, a inspection method based on feature extraction using two - dimensional continuous wavelet transform(2D CWT)is proposed. The 2D CWT is employed to extract local features from texture images with defects both in the frequency and in the spatial domain, in which the optimum scales and rotation angles are produced automatically through Fourier spectrum analysis. By a pre - determined global threshold, the defect region are segmented from the modulus of 2D CWT. Experiments with real image data show that the proposed method is robust in fabric defect detection.