【作者】 张小锋；

【导师】 徐鸿钧；

【作者基本信息】 南京航空航天大学 ， 机械制造及其自动化， 2007， 博士

【摘要】 长期以来,砂轮地貌的研究历来是磨削界的研究热点。因为在磨削加工中,当磨削用量组合一定时,砂轮表面的有效磨粒间距、等高性、磨粒磨损平台面积等地貌参数,对砂轮的加工性能和工件的表面加工质量有着重大影响。若能通过测量量化这些砂轮地貌参数,便可以掌握砂轮的工作状况,进而预测整个磨削过程。因此砂轮地貌的检测对深入研究磨削过程有着重要的意义。由于砂轮地貌本身和工况的复杂性,使得砂轮地貌的检测比较困难,目前能用于指导实际生产的科研成果鲜见报导。本文在深入分析以往砂轮地貌检测方法的基础上,提出采用计算机双目视觉技术,对砂轮地貌的定量检测开展基础研究。论文完成了以下几项具有创意的基础性研究工作:(1)根据光轴平行法,研制出砂轮地貌双目视觉检测系统,该系统采用单摄像头,通过被测物体的精确平移,可实现双目图像的采集。(2)研究数字图像预处理算法。主要包括数字图像的灰度化算法、图像缩放算法和边缘检测算法,分析各种算法的适用场合和优缺点,并对各种算法进行了编程验证及处理效果对比。(3)研究双目视觉的灰度模板匹配算法。对匹配模板大小进行优化,优化的匹配算法,具有匹配精度高、抗噪声能力强、运算速度高等的综合优势。(4)采用自制的标准试块,对砂轮地貌双目视觉检测系统进行了标定。分析了系统的误差源并相应提出了改良办法,高度测量误差可控制在5%以下。(5)改进了基于角点响应函数的快速角点提取算法,该算法不仅省去了改变图像分辨率的步骤,而且保证了高度估算精度;通过测量磨粒表面多个特征点的高度,重构出磨粒的三维形状。(6)将砂轮地貌双目视觉检测系统应用于单层钎焊金刚石端面砂轮表面形貌的测量中,实现了砂轮形貌主要参数(磨粒高度、磨粒间距)的检测。这种直接检测的方法具有非接触、检测精度较高、检测结果直观等优点。更多还原

【Abstract】 The research on the grinding wheel topography has been regarded as one of the important fields in the grinding industry for a long time. When the grinding wheel topography parameters are acquired, such as the effective distance between two adjacent grains, uniform grain height and wear-flat areas, the grinding parameters can be optimized, the wear situation can be controlled, and the working condition can be forecasted. In the grinding process, the grinding wheel topography has both great impact on machinability of the grinding wheel and the surface finished quality of the product, so the detection of the grinding wheel topography is significant to reveal further grinding principles.It is difficult to detect the grinding wheel topography because of its complexity and varying machining conditions, and there are few reports on the production. On the basis of analyzing former methods, the binocular vision technology is used to inspect quantificationally the grinding wheel topography in this paper.The main creative contents in this paper are as follows:(1)According to the optical axis parallel approach, the binocular vision detection system of grinding wheel topography is developed. By using a single camera, the binocular images acquisition is realized with the precise translation of the object.(2)The preprocessing algorithms of digital image are studied. It mainly includes the gray degree processing, scaling and edge extraction algorithm. The characteristic and application scopes of each algorithm are analyzed, and the programming verifications are carried out and the processing results are compared.(3)The template matching algorithm of binocular images based on the light intensity information is studied. By optimizing the template size, there are integrated advantages of high matching precision, the increased noise resistance performance and high speed calculation course of this algorithm.(4)By using self-made standard sample, the binocular vision detection system of grinding wheel topography is calibrated. The system is improved by analyzing error source. The measuring error is less than 5%.(5)The fast corner extraction algorithm based on corner response function is improved. Not only the processing of the resolution of an input image is left out, but also the estimation accuracy of height is guaranteed. By measuring the heights of feature points in the grain surface, the 3D shape of grain is reconstructed.(6) The brazed diamond monolayer grinding wheel topography is detected with the binocular vision detection system. The height of grains, the distance between two adjacent diamond grains are both obtained. This technique has advantages of non-contact, high precision, and intuitionistic measuring results.更多还原