一种新型纱线毛羽图像特征识别算法的研究

针对毛羽检测的问题,搭建了一套纱线毛羽的数字化图像采集系统,提出了一种新型的基于图像处理算法的纱线毛羽检测方法,可对连续采集的纱线图像序列进行特征分析,提取出可用于纱线毛羽质量评估的特征指标。通过对纱线图像进行预处理,包括灰度变换、背景处理、图像增强、动态阈值分割、倾斜纠正、图像去噪、图像分割,得到纱线的毛羽图像,并对毛羽图像进行细化处理,从而得到细化后的毛羽图像,以纱线轴线和主干边缘作为参考,选择合适的基准线,对毛羽像素点进行识别判断,统计得到不同长度毛羽的根数。实验表明,图像法测试的结果与目测法的结果最大偏差在5%以内,具有很好的一致性,该方法能够提高纱线毛羽检测的效率和精度。

Research of a novel method for measuring yarn hairiness based on image recognition

Aiming at hairiness measurement, a digital image acquisition system is constructed, based on which a novel algorithm is proposed in order to measure the number and length of yarn hairiness. The algorithm can be used to make feature analysis of yarn images captured continuously, and yarn hairiness features for quality evaluation of yarn hairiness are determined. With pretreatment of yarn image, including gray-scale transformation, background subtraction, image enhancement, dynamic threshold segmentation, tilt correction, image denoising and image segmentation, yarn hairiness image is obtained. Then the yarn hairiness image is treated by using thinning algorithm, so the thinned yarn hairiness image is obtained. With yarn axis and the margin of trunk as reference, and with proper baseline, yarn hairiness image skeleton is extracted at the pixel level, and the number of yarn hairiness of different length is determined by statistics. Experimental results show that the precision of the new method is consistent with the visual observation method, and its deviation is within 5%. The method can improve the efficiency and accuracy of yarn hairiness feature extraction.