【作者】 李佳彦；

【导师】 丁永生；

【作者基本信息】 东华大学 ， 通信与信息系统， 2007， 硕士

【摘要】 我国是纺织品生产大国，纺织业是我国的传统出口行业。随着中国加入WTO，纺织产品的出口量越来越大；且随着国内生活水平的不断提高，内销产品也对纬斜指标的要求越来越高。因此，国内用户对高性能的整纬装置的需求十分迫切，前景十分广阔。本文面向摄像整纬装置中基于织物图像纹理的纬斜检测与多摄像头图像信息的融合与处理问题，综合运用数字图像处理、模式识别以及信息融合等技术，设计了多摄像头纬斜检测的架构，提出了基于图像处理的织物纹理识别算法和面向摄像整纬的多源图像融合算法，为摄像整纬装置的研制提供支持。首先，论文从图像纹理辨识技术、多传感器信息集成与融合、DSP处理器的发展等三方面，对摄像整纬装置研发的理论依据进行了分析，并提出了多摄像头纬斜检测的基本架构。其次，基于图像处理的织物纹理识别算法，以傅氏变换的自配准性质，结合基于统计的纹理自动分析方法，对织物纹理图像进行识别与特征提取，计算得到其纹理的方向特征参数。接着，针对摄像整纬装置中多摄像头图像信息的融合问题，提出了面向摄像整纬的多源图像融合算法。通过对整纬执行机构的矫正辊(直辊和弯辊)调整织物行程差的作用进行数字化处理，应用数据拟合等方法，实现对多摄像头织物纹理的方向特征参数的信息融合与处理，从而得到织物当前的纬斜状态与纬斜程度，为整纬装置的执行机构提供必要的控制依据。然后，应用基于图像处理的织物纹理识别算法，结合多摄像头纬斜检测的基本架构，提出了织物图像采集与纹理识别的实现方案，对摄像整纬装置中织物图像采集与纹理识别部分进行了初步的硬件架构与软件流程设计。最后，对本文的研究工作进行了总结，并展望了其进一步研究的方向。更多还原

【Abstract】 As a large textile producer, textile industry has become a traditional export industry in China. The volume of textile products has kept growing since China entered WTO. Meanwhile, domestic customers are paying more and more attention to weft-inclining parameters with the improvement in standard of living. Therefore, high-powered weft-straightening machines are urgently demanded within China and thus have a bright foreground.In order to solve the problem of fabric texture image identification and multi-camera image information fusion, we design an architecture for multi-CCD weft-straightening based on digital image processing, pattern recognition and information fusion We also suggest a fabric texture image identification algorithm and a multi-image fusion algorithm CCD-based weft-straightening machine, which could be a reference for further study on the research and development of this kind of machine.Firstly, we analyze the theory foundation for research and development of CCD-based weft-straightening devices based on image texture identification, multi-sensor information fusion and DSP technology. And then we propose a framework for multi-camera fabric-inclining detection. Secondly, with the auto-registration property of Fourier transformand texture automatic recognition method based on statistics, we calculate the main parameters for texture direction of a fabric and extract its characters.Thirdly, amulti-image fusion algorithm for CCD-based weft-straightening machines is proposed to implement multi-camera information fusion on them. It digitalizes the action of the straightener rolls, the erect roller and the bowed roller, of the machine’s actuator which leads a progressive error to the fabric and then fuses the direction character parameters of multiple fabric textures. It gets the weft-declining status and extends of the fabric at the moment so as to provide the actuator with necessary parameters for controlling.And then we suggest a solution for fabric image gathering and texture identification with the fabric texture image identification algorithm and the architecture for milti-camera fabric-inclining detection. We also study on an elementary hardware and software design for fabric image gathering and identification subsystem of the CCD-based weft-straightening machine.Finally, we summarize all our research and prospect its further research topics.更多还原