基于压缩感知的精密轴高精度视觉测量方法研究

Precision axes are widely used in various fields, which are the core components of all kinds of transmission equipment. It is significant to study the method of high precision vision measurement for the reliability and service life of the equipment. Compressed sensing can be sampled and compressed at the same time with low rate, which can lower the requirements for hardware, reduce the amount of data needed to process and speed up the processing speed, which is one of the most attractive research field in current non-contact measurement. The project first proposed the method of high precision vision measurement based on compressed sensing and machine vision technology for slender type of precision axes. The research is carried out as follows: (1) Establishing the conversion model between the reference image and the image waiting to mosaic, and determining the high precision global calibration method. (2)Research on the high precision image mosaic algorithm based on compressed sensing and Non-Subsampled Shearlet Transform(NSST). (3) Study on sub pixel level segmentation of the edge of precision axis based on interpolation method and the conics fitting method in accordance with different edge types. (4) Analyzing the mechanism of error, and realizing the multi-parameters high precision vision measurement for precision axes with the distortion correction and error compensation.The research will explore new ideas and methods for the non-destructive and non-contact measurement of modern machines, and have important significance to improve the level of modern equipment.

精密轴的应用非常广泛，是各种传动设备的核心组件，研究此类零件的高精度视觉测量方法，对提高设备运行的可靠性与使用寿命，具有重要意义。压缩感知能以低速率对信号同时进行采样和压缩，具有可降低对硬件的要求、大幅减少需处理的数据量、加快图像处理速度等特点，是当前非接触测量中最具吸引力的研究方向。本项目首次提出采用压缩感知原理结合机器视觉，研究细长型精密轴的高精度视觉测量方法。课题开展以下工作：1）建立待拼接图像与参考图像之间的转换模型，确定高精度全局标定方法；2）研究基于压缩感知与非下采样剪切波变换（NSST）的高精度图像拼接算法；3）研究对于不同光源形成的不同目标边缘类型，采用不同的边缘检测算子并融合插值及二次曲线拟合法，以实现精密轴图像边缘的亚像素级定位分割；4）分析误差形成机理，建立畸变补偿函数对检测数据进行畸变校正和误差补偿处理，实现精密轴多参数的高精度视觉测量。

针对传统接触式测量效率低、误检率高、难以实现机械加工流程自动控制的现状，课题组研究融合压缩感知与机器视觉的细长型精密轴高精度视觉测量方法。本课题主要研究内容如下：建立了待拼接图像与参考图像之间的转换模型，确定了高精度全局标定方法；提出了基于压缩感知与非下采样剪切波变换（NSST）的高精度图像拼接算法；针对于不同光源形成的不同目标边缘类型，采用不同的边缘检测算子并融合插值及二次曲线拟合法，实现精密轴图像边缘的亚像素级定位分割算法；通过分析误差形成机理，建立畸变补偿函数对检测数据进行畸变校正和误差补偿处理，实现精密轴多参数的高精度视觉测量。本项目的研究成果为精密轴高精度视觉测量提供了可借鉴的理论和技术。研究成果:以第一作者或通讯作者发表研究论文13篇，其他相关论文3篇，其中以第一作者或通讯作者发表SCI检索论文7篇（二区论文3篇），申请发明专利5项（已授权4项），实用新型专利授权1项，授权软件著作权3项，培养硕士生6名，顺利完成项目的研究任务和目标。