Identification of wool and cashmere extremely similar fibers is always an important topic in the textile industry. In order to solve this problem much better, a novel fiber identification method based on the extraction and analysis of the morphological features was proposed in this paper. Firstly, the original fiber images were captured by the self-developed system including the optical microscope and digital camera. The influence of the acquisition process may lead to the low contrast and impurities, so the original fiber images needed to be processed by the image enhancement and de-noise to obtain the available fiber images with a better quality. Then the hessian matrix of processed images was put into the Frangi filter to detect the edge of the fiber scales, and the binary images of filter output images were processed to obtain the signal-pixel scale skeleton. The connected region labeling algorithm can be adopted for the scale skeleton images to mark and extract every scale from the whole fiber according to the different color information. Next, the three morphological features including scale height, fiber diameter and their ratio can be calculated by the self-defined vertical line rotation analysis method, and the mean value of five different scales was calculated as the final features to describe one fiber. In the experiment, 500 fiber cashmere and 500 wool fiber images were collected for the whole research, and a Bayesian classification model for identifying wool and cashmere fibers was established based on the statistical assumptions of three morphological characteristics. The results show that the identification accuracy of the method proposed in this paper could reached the 94.2%. It also proves that this novel method can be used for the identification of cashmere and wool extremely similar animal fibers.

羊毛和羊绒极其相似的纤维的识别一直是纺织工业中的重要主题。为了更好地解决该问题，本文提出了一种基于形态特征提取和分析的新型纤维识别方法。首先，原始的纤维图像是由包括光学显微镜和数码相机在内的自行开发的系统捕获的。采集过程的影响可能导致对比度低和杂质少，因此需要对原始纤维图像进行图像增强和去噪处理，以获得质量更好的可用纤维图像。然后将已处理图像的粗麻布矩阵放入Frangi滤镜中，以检测纤维鳞片的边缘，对滤波器输出图像的二值图像进行处理，以获得信号像素尺度骨架。鳞片骨架图像可以采用连通区域标记算法，根据不同的颜色信息标记和提取整个纤维中的每个鳞片。接下来，可以通过自定义垂直线旋转分析方法计算出鳞片高度，纤维直径及其比率这三个形态特征，并计算出五个不同鳞片的平均值作为描述一根纤维的最终特征。在实验中，为整个研究收集了500个纤维羊绒和500个羊毛纤维图像，并基于三个形态特征的统计假设，建立了用于识别羊毛和羊绒纤维的贝叶斯分类模型。结果表明，该方法的识别准确率可达94.2％。还证明了该新方法可用于鉴定羊绒和羊毛极相似的动物纤维。