基于图像的棉花异性纤维识别算法研究

发布时间：2018-03-24 00:15

本文选题：棉花　切入点：异性纤维　出处：《郑州大学》2017年硕士论文

【摘要】：棉花中的异性纤维主要是指棉花在采摘、晾晒、打包、加工和运输的过程中混入的非棉性纤维,如头发丝、各类丙纶丝、塑料薄膜、破布片、彩色线等,也俗称“三丝”。异性纤维在棉花中的含量虽少,但由它造成的危害却非常大。主要表现在:在清梳工序中,异性纤维极易被粉碎,混入纱线后容易导致纱线断头;在印染过程中,因为异性纤维和棉纤维的着色率不同,会造成染色不均匀,出现大量的印染瑕疵,进而影响布匹的外观和质量;由异性纤维造成的残次品严重影响了商家信誉、纺织品的质量以及市场竞争力。因此,研究一种满足高识别率和实时性要求的棉花异性纤维识别算法具有十分重要的意义。本课题主要研究了棉花中异性纤维的识别算法,首先对棉花异纤图像进行预处理,然后分别采用迭代全局阈值分割法和改进二维最大类间方差法(Otsu)对棉花异纤图像进行识别实验,并对两种算法的识别效果进行对比,最后利用Matlab完成了异纤识别算法的软件设计,具体的工作如下:(1)在棉花异纤检测设备上通过高速CCD相机完成棉花异纤图像的采集。(2)选择RGB颜色空间进行实验,首先分析了异纤图像的灰度直方图,根据深色异纤和白色异纤的不同特点,在后续识别时深色异纤直接利用图像的灰度直方图进行分析,而白色异纤利用图像的蓝色分量直方图进行分析。然后通过对背景的灰度值进行变换,使背景的灰度值范围落在棉花的灰度值范围内,减少了背景对异纤识别的影响,最后使用中值滤波来消除图像中的噪声。(3)分析了几种常用的图像分割方法,选取迭代全局阈值分割法识别异纤图像,并对多幅图像进行实验分析,统计了各种异纤的识别率。(4)利用改进的二维Otsu方法识别异纤图像,该方法在传统二维Otsu算法的基础上,对目标和背景的分块策略进行了调整,减少了算法的运行时间。实验部分对该方法与迭代阈值法在运行时间和异纤识别率上做了对比分析。(5)用Matlab的图形用户界面(GUI)设计了异纤识别算法界面,并实现了异纤图像处理的主要功能,方便用户操作,提高了交互性。
[Abstract]:Foreign fibers in cotton mainly refers to cotton picking, drying, packaging, non cotton fiber mixed in the process of processing and transportation in the hair, all kinds of polypropylene fiber, plastic film, rags, color line, also known as the "three wire". The content of foreign fibers in cotton is less however, the harm caused by it is very large. Mainly in: in the process of carding, the fibers can easily be crushed, easily lead to yarn breakage after mixed yarn; dyeing process, because straight fiber and cotton fiber dyeing rate, will cause uneven dyeing, dyeing a large number of defects then, affecting the appearance and quality of the cloth; the defective products caused by the fibers seriously affect the business reputation, textile quality and market competitiveness. Therefore, a high recognition rate and meet the real-time requirements of the cotton fiber recognition algorithm is very important The significance of this topic mainly studies the recognition algorithm of foreign fibers in cotton, the cotton fiber image, then using iterative global thresholding method and improved 2D maximum between class variance (Otsu) experiments on cotton fiber image, and compared the effect of two kinds of recognition algorithm finally, using Matlab to complete the software design of the fiber recognition algorithm, the specific work is as follows: (1) in the cotton fiber detection equipment through the high-speed CCD camera to complete the cotton foreign fiber image acquisition. (2) RGB color space experiment, firstly analyzes the histogram of image fiber, according to the different the characteristics of dark and white fiber fiber, in the subsequent recognition of dark fiber directly using image histogram analysis, and the use of white foreign fibers in blue color histogram of the image was analyzed. Then by On the background of the gray value transform, the background gray value range in cotton gray value range, reduce the influence of background on the identification of different fiber, finally use the median filter to eliminate the noise in the image. (3) analyzed several commonly used methods of image segmentation, selection of iterative global threshold segmentation recognition different fiber images, and the images were analyzed, the statistical identification of different fiber rate. (4) using two-dimensional Otsu method to identify different fiber improved image, the method based on the traditional 2D Otsu algorithm, block strategy of target and background were adjusted to reduce the running time of the algorithm the experimental part of the method and the iterative threshold method to do a comparative analysis on the running time and different fiber recognition rate. (5) using Matlab graphical user interface (GUI) design of the different fiber recognition algorithm interface, and realizes the main function of different fiber image processing Able to facilitate user operation and improve interactivity.

【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TS111.9;TP391.41

【相似文献】