基于DSP的棉花异纤识别系统研究与实现

发布时间：2018-05-05 14:56

本文选题：异性纤维 + 数字信号处理器（DSP）　；参考：《中原工学院》2013年硕士论文

【摘要】：在纺织工业中，纺织品的质量会受到多种因素的影响，其中很重要的一点就是纺织原材料棉花在生产和运输的过程中混入的异性纤维。异性纤维通常称作“异纤”，主要包括植物的叶茎、人和动物的毛发、化学纤维制品等，由于这些物质在特性上与棉花纤维有较大的差异，如果纺织品中混杂有这些异纤就会造成纺线的断裂、着色不均匀等一系列危害。为了解决这些问题，提高纺织品质量，国外多家企业与机构都研发了针对异性纤维的清除设备，，并且价格不菲。而国内同类型的装备在技术上也主要依赖于国外的技术和设备。因而研制拥有自主核心技术与知识产权的异纤清除设备变得十分必要。本系统的研究主要包括基于DSP的硬件系统设计和基于支持向量机（SVM）的分类算法设计两个部分。系统采用DSP+FPGA的结构来完成对高速棉花流图像的实时处理。首先采用CCD相机采集目标图像，将获得的RGB数据送入到FPGA的数据缓冲区，DSP配合外部存储器SDRAM完成对数据的搬运和数据的运算处理，对目标图像进行识别。另一方面通过分析异性纤维的颜色特征，针对异纤的多样性，线性不可分等特点，在传统的阈值分割方法的基础上，对采集的图像进行先期的学习，利用SVM分类算法，对含有多种类型异纤的图像进行先期的训练，获得性能优异的分类器，克服异纤图像在RGB空间中线性不可分割的特点，提高系统对异性纤维识别的准确率。
[Abstract]:In the textile industry, the quality of textiles will be affected by many factors, one of which is the textile raw materials cotton mixed in the process of production and transportation of heterosexual fibers. Heterosexual fibers, commonly known as "hetero-fibrils," mainly include plant leaves, human and animal hair, chemical fiber products, etc., because of their characteristics, which differ greatly from cotton fibers. If the textile is mixed with these fibers, it will lead to the breakage of spinning thread, uneven coloring and a series of hazards. In order to solve these problems and improve the quality of textiles, many foreign enterprises and institutions have developed anti-heterosexual fiber cleaning equipment, and the price is high. The same type of equipment in the domestic technology is also mainly dependent on foreign technology and equipment. Therefore, the development of independent core technology and intellectual property rights of foreign fiber clearance equipment has become very necessary. The research of this system mainly includes two parts: hardware system design based on DSP and classification algorithm design based on support vector machine (SVM). The system adopts the structure of DSP FPGA to complete the real-time processing of high-speed cotton flow image. Firstly, the CCD camera is used to collect the target image, and the obtained RGB data is sent to the FPGA data buffer to complete the data handling and data processing together with the external memory SDRAM to identify the target image. On the other hand, by analyzing the color characteristics of heterosexual fibers, aiming at the diversity and linear inseparability of hetero-fibers, based on the traditional threshold segmentation method, the collected images are studied in advance, and the SVM classification algorithm is used. In order to overcome the linear inseparability of heterogeneous images in RGB space and improve the accuracy of heterosexual fiber recognition, the image with various types of fibers is trained in advance to obtain a classifier with excellent performance.
【学位授予单位】：中原工学院
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TS102.211;TP391.41

【相似文献】