棉花中异性纤维的图像多分辨率差分检测方法

针对棉花异性纤维难以检测的问题，提出了一种基于多分辨率尺度图像差分的检测方法。首先，通过小波多层分解实现检测图像中的不同频率信息成份的分离；然后，对不同分辨率的子图之间进行差分提高异性纤维与原棉信息的对比度；在此基础上，采用最大类间方差法对多分辨率差分子图进行二值化分割出异性纤维的信息；最后，通过对分割后的子图进行信息的融合，实现了对异性纤维检测。实验表明，该方法能够有效抑制棉花背景信息，最大限度的提取异性纤维的信息，异性纤维检测的准确率以上达到90%。

Detection of foreign fibers among cotton based on image multi-resolution difference method

Abstract For cotton foreign fiber detecting difficulty problem, the method is proposed for detection cotton foreign fiber based on the multi-resolution image difference. First of all, different frequency components of detection image are separated through the wavelet decomposition. Then, different resolution sub-graphs are used differential to improve information contrast of foreign fiber and cotton. On this basis, the Maximum between-Cluster Variance method is proposed for Bi-value image of multi-resolution difference sub-graph and anisotropic fiber information. Finally, the segmentation images are fused for the fibers information, so as to detect the complete opposite fiber information. Experimental results show that the method can effectively restrain the cotton background information, maximum extract foreign fiber information, and foreign fiber detection accuracy rate exceed above 90%.