基于机器视觉的微小零件形貌检测方法

针对当前工业生产中人工对微小异形零件形貌参数测量精度低、速度慢的问题,提出了一种基于机器视觉的检测方法,并开发了一款基于开源计算机视觉库OpenCV的检测软件。该检测方法首先使用CMOS相机采集被测零件的图像,并结合频谱特征对其进行滤波、阈值分割等预处理;然后选取效率高、边缘跨度为单像素的Canny边缘检测算法对预处理之后的图像进行边缘检测;最后采用Ramer算法对零件轮廓进行递归细分,拟合出几何基元,并结合测量焦距下的系统标定系数计算出零件实际的形貌参数。实验结果表明:通过该检测方法对长、宽均为毫米量级的Ω型微小零件进行形貌检测,检测精度达到10μm以下,具有精度高、速度快的优点,可为工业化生产提供可靠依据。

Shape Parameter of Micro Part Detection Method Based on Machine Vision

Aiming at the faults of low precision and slow speed in the manual measurement of tiny special-shaped com-ponent’s shape parameters in current industrial production, a detecting method based on machine vision was proposed and a detecting software founded on open-source computer vision library OpenCV was programmed. Firstly, the object was imaged with a CMOS sensor and preprocessed with filtering and threshold by the use of spectrum analysis method. And then Canny edge detecting algorithm which is successful in extracting the edges with pixel precision and high effi-ciency was chosen to detect the edge of preprocessed image. In the end,by adopting the Ramer algorithm which per-forms a recursive subdivision of the contour to fit geometric primitives and using system calibration coefficient, the shape parameters of the measured part were obtained. The experimental result shows that through the proposed detec-tion method, the shape parameters of a micro component in the shape of Ω, the length and width of which were in millimeter level which can be acquired. And the detecting precision can reach to a level of 10μm. The method has the advantages of high precision and fast speed which can provide a reliable basis for industrialized production.