【作者】 熊四昌；

【导师】 程耀东；

【作者基本信息】 浙江大学 ， 机械制造及其自动化， 2003， 博士

【摘要】 基于计算机视觉的刀具状态监测技术具有传统刀具状态监测方法所没有的优点。本文旨在研究基于计算机视觉的刀具状态监测方法和刀具在各种磨损状态下所表现出的视觉特征。研究以工件表面纹理特征的提取和刀具磨损图像的分割算法为主。工件表面纹理特征的提取以随机图像纹理模型为基础，分别研究了基于分数布朗运动理论和马尔可夫随机场理论的图像纹理特征参数的提取方法和应用。刀具磨损图像的分割算法建立在马尔可夫随机场理论基础上，推导出合理的计算方法，并在此基础上计算了刀具磨损VB值及其它参数。本研究采用的基于计算机视觉的方法为刀具状态监测技术的开发提供了一种新思路。本文各章的主要内容如下： 第一章首先阐述了刀具状态监测技术的研究背景和重要性，对刀具状态监测技术的发展状况作了深入阐述。综合比较研究了传统刀具状态监测方法的特点，阐明了计算机视觉的原理和其在刀具状态监测中的应用与优点。 第二章分析了刀具磨损形态的特征、磨损过程和磨钝标准，从理论上描述了利用后刀面磨损图像进行刀具状态监测的合理性和可行性。同时分析了影响工件表面纹理的主要因素，为后续章节研究基于工件表面纹理的刀具磨损状态监测奠定了基础。 第三章对纹理学原理与其主要研究方法进行了分析，然后结合工件表面图像特征，研究了纹理学特征分析的经典方法—灰度共生矩阵法，并结合工件表面图像，分析了工件表面纹理在归一化灰度共生矩阵中所表现出的特点。 第四章采用分数布朗运动(FBM)的基本理论，对切削工件表面纹理的图像进行分析，提取纹理特征参数，根据分形维数和图像上对数功率谱的拟合曲线的平均斜率来判断刀具的磨损状态。并研究了图像噪声对该方法的影响，实验表明，该方法对车削工件表面纹理具有较好的识别能力。 第五章引入了马尔可夫随机场理论，建立了马尔可夫随机场纹理模型，对工件表面纹理图像进行了分析。提出了基于Gibbs分布和基于近邻像元的两种纹理模型参数估计方法以提取工件表面纹理特征，并利用纹理特征参数对刀具的磨损浙江大学博士学位论文状态进行判别。研究了小角度的旋转和图像噪声污染对马尔可夫随机场模型的纹理特征的影响。 第六章以马尔可夫随机场理论为基础，提出刀具磨损区域分割算法，推导了松弛迭代更新公式，并用该方法对图像分割问题进行求解，获得了MAP准则下将刀具图像的分割结果。在此基础上，使用峰值带宽法从磨损区图像中分割出目标区，并通过链码边界搜索法得到了目标区的边界描述。针对边界中存在的分割噪声使用圆形形态算子对边界进行完整化，最终获得了完整的刀具磨损区图像，为实现通过刀具磨损区大小来评判刀具磨损程度的方法奠定基础。 第七章研究了基于计算机视觉的刀具状态监测技术在车削监测中的应用。首先介绍了刀具状态监测系统的基本组成，说明了刀具磨损状态的监测步骤。然后基于车削实验所得到的数据，使用前述章节的各种纹理分析方法对工件表面纹理进行分析和讨论。最后利用基于马尔可夫随机场的图像分割模型对刀具后刀面图像的磨损区进行分割，分析了分割结果，取得了满意的效果。 第八章对全文工作进行了总结，并归纳了论文的创新点，对进一步的研究工作提出了设想和展望。更多还原

【Abstract】 Cutting t ool c ondition monitoring b ased one omputer v ision h as a dvantage w ith which traditional monitoring methods aren’t provided. In this paper, cutting tool condition monitoring method based on computer vision is researched. The research is composed of feature extraction of workpiece surface texture and segment of tool wear image, which are based on the model of Markov Random Field. Cutting tool condition monitoring based on computer vision has provided a new method for the development of TCMS. Each chapter research contents are as follows:The first chapter talks about the research significance and background. Some traditional tool condition monitoring methods are researched. The theory and advantage of Computer vision are expounded.The second chapter analyzes the characters of tool wearing surface and the factors effecting workpiece surface texture. A reasonable experiment scheme is designed. Based on the experiment datum, the rationality of the selection about cutting parameters is analyzed.The third chapter gives the brief expatiation about the concept of texture. The features of workpiece surface texture are extracted with gray co-occurrence matrix and the disadvantage of this method is pointed out.The fourth chapter introduces the theory of Fractional Brown Movement (FBM). The image texture of turning workpiece is analyzed using FBM model. After abstracting the texture character, the cutting tool wear status can be determined according to fractal dimension and average slope of the fitting curve of the logarithm power spectrum.Compared with the experimental facts, the result of this method is satisfying.The fifth chapter introduces the theory of Markov Random Field. Markov Random Field texture model is established to analyze workpiece surface texture image. Two parameter estimation methods of the texture model based on the Gibbs distribution and neighbor pixels are adopted to extract the features of workpiece surface texture. Tool wear condition is recognized by the feature parameters of texture. The experiment datum indicates that the feature parameters of MRF model have the rotative invariability for the texture image with small angle rotation.* The sixth chapter presents a tool wear area image segmentation algorithm based on Markov Random Field m odel. The refresh formula of relaxation labeling is deduced. With the algorithm, the segmentation result composed of wear area, background area and tool body area is obtained according to the MAP(maximum a posterior) criterion. By using peak-band algorithm, the aim area (B area) is segmented from the wear area image. Aiming at the edge noise of image, mathematical morphology theory is applied to integrate edge. Test results show that the segmentation algorithm is effective in identifying tool wear degree.The seventh chapter researches the application of tool condition monitoring technology based on computer vision in the turning monitoring. First, the basic structure of tool condition monitoring system and the monitoring approach of tool wear status are presented. Secondly, based on the obtained data of cutting experiment, several texture analysis techniques are used to analyze workpiece surface texture. Lastly, the image segmentation model based on MRF is applied to acquire the wear area of flank face of tool and the segmentation result is analyzed.The eighth chapter summarizes the main fruits and prospects the future of the research.更多还原