基于机器视觉的盲孔直径测量

Blind Hole Diameter Measurement Based on Machine Vision

摘要: 针对盲孔直径测量困难、测量精度低、速度慢等问题，本文提出了一种基于机器视觉的盲孔直径测量系统。首先搭建图像采集系统采集盲孔图像，然后利用基于多尺度聚焦测度和广义随机漫步(Generalized Random Walks)的多聚焦图像融合算法进行盲孔图像融合，加深图像边缘，其次对融合后的图像利用全变分(Total variation)模型去除噪声，并提出了一种改进Canny边缘检测算子的方法对图像进行噪声滤除的同时保持了图像边缘，结合形态学方法使得到的盲孔边缘区域更加清晰、连贯，最后利用处理好的盲孔边缘图像进行多次测量求取平均值从而得到盲孔直径。实验结果表明，本文最大相对误差为0.802%，测量精度达到0.001 mm，即数据分布比较集中，证明系统稳定性较好，满足实验精度要求。

Abstract: Aiming at the problems of difficult blind hole diameter measurement, low measurement accuracy, and slow speed and so on, a blind hole diameter measurement system based on machine vision is proposed in the paper. First, set up an image acquisition system to collect blind hole images, and then use a multi-focus image fusion algorithm based on Multi-Scale Focus Measures and Generalized Random Walks to perform blind hole image fusion and deepen the edge of the image. Then, remove noise for the image using the variational (Total variation) model, and propose an improved Canny edge detection operator method to filter the image noise while maintaining the edge of the image, combined with morphological methods to make the edge area of the blind hole clearer and coherent. Finally, use the processed blind hole edge image to perform multiple measurements to obtain the average value to obtain the blind hole diameter. The experimental results show that the maximum relative error in this paper is 0.802%, and the measurement accuracy reaches 0.001 mm, that is, the data distribution is relatively concentrated, which proves that the system has good stability and meets the requirements of experimental accuracy.