基于小波变换及投影分析的织物倾斜度检测算法

Fabric skew detection based on wavelet transform and projection profile analysis

为实现对织物倾斜度的精确检测,基于小波变换及投影分析提出一种有效的倾斜度检测算法。首先,待检织物图像经过一层小波分解,并提取出纬线边缘;然后利用形态学滤波方法中的腐蚀操作消除边缘噪声;最后将处理好的边缘图像在一定角度范围内以一定的步长进行旋转,并求出在每个角度下的投影向量,通过判定函数确定织物图像的倾斜角度。实验结果证明:对于平纹图像,可获得大于95%的准确率,即使对于纹理较为复杂的花纹图像,仍能获得大于88%的准确率;所提算法具有较高的准确率和较好的实用性。

In order to precisely detect the skew of fabrics, a novel skew detection scheme is proposed based on wavelet transform and projection profile analysis. Firstly, the test image is decomposed by wavelet transform, and the weft boundary is generated. Then morphological method is adopted to reduce serrated noise around weft boundary. In the end, a projection profile is computed at each angle, and the angle that maximizes a criterion function is regarded as the skew angle. Experimental results show that the skew angles detected by the proposed detection algorithm are of higher accuracy （over 95% accuracy is obtained for the image of plain weave, and over 88% accuracy in the case of image of more complex patterns ）.