基于图像能量的布匹瑕疵检测方法

为了满足布匹瑕疵检测的准确性和实时性要求,通过对几种纹理分析方法的讨论,提出了一种基于能量的瑕疵检测方法;该方法利用图像极坐标系下的傅立叶频谱,得到图像纹理频谱能量的分布情况;由于纹理的空间周期性,其图像的能量在对应的频率处会有峰值出现;如果将这些频率处集中的能量置零,通过傅立叶逆变换得到的图像纹理会被模糊甚至去除,瑕疵与纹理的对比度得到增强;这时再进行边缘检测和二值化处理,即可将瑕疵从背景中提取出来,从而完成瑕疵的检测;实验证明,该方法检测速度快,准确率高,能够满足检测要求。

Fabric Defect Detection Method Based on Image Energy

To meet accuracy and real-time requirements of the fabric defect detection,this article proposed a detection method based on image energy,through discussions of several texture analysis methods.The distribution of texture spectrum energy can be obtained,using Fourier spectrum under polar coordinate.And there are peaks in corresponding frequencies due to the spatial period of texture.If resetting energy values in the peaks,the image texture from IFFT will be blurred even removed.Consequently,the contrast of texture and defect will be strengthened.Then edge detection and binary process are successive and defects can be detected.Experiments showed that the method is fast,accurate and can meet the detection requirements.