应用显微视觉的跨尺度几何参数测量方法

A Micro Vision-Based Method for Multi-Scale Geometric Parameters Measurement

结合数控机床,提出了基于显微立体视觉的工件多尺度测量方法。首先,针对显微镜视场窄、景深小、非线性因素多的特点,建立了基于多种非线性畸变修正的双目显微视觉成像模型。以此为基础,提出基于张氏标定法和变倍率法的主参数标定方法以及基于光束平差法的主参数和畸变参数优化方法。其次,针对直线、圆、自由曲线等多种特征,用阈值分割算法从背景中识别出待测工件,并采用Robert算子、Hough变换、多项式拟合等方法提取不同特征轮廓和轮廓关键特征点。最后,通过跨尺度标准件的关键尺寸测量,验证测量方法的测量精度。结果表明,系统的平均测量误差为3.15μm,满足测量精度要求。

A method based on micro vision is proposed in this paper, and the vision inspection system combining CNC machine tools is established for measuring a variety of the size of one parts. First, in consideration of narrow view field, small depth of focus and too many nonlinear distortion factors, a binocular micro vision imaging model based on the non-linear corrections is established. Therefore, the calibration method for scale factor based on Zhang＇s method, the principal point calibration method based on the variable-magnification-method, and the optimization for the main parameters and distortion parameters based on bundle adjustment are studied. Second, in view of some common features including straight line, round, free curve, etc, the linear feature extraction based on Robert operator and the Hough transform, the circular feature extraction method based on connected area, as well as the free curve extraction method based on least square method are explored respectively. At last, in order to verify the accuracy of the measuring system, a standard work-piece with multi-scale structures are measured. The results show that the average measurement error of the system is 3.15 μm and it satisfies the requirements of measurement accuracy.