基于机器视觉的车灯定位误差检测方法研究

本文对图像处理技术的概念、特点、构成进行了概述,对现有的经典图像分割和识别方法进行了介绍,着重讨论了曲线拟合的方法在车灯面形定位误差检测中的应用。利用相关系数在真实匹配点附近的统计特性,通过对相关系数进行曲线拟合,根据其一阶导数确定最大相关系数所在的位置,快速得到亚像素匹配点。与逐步搜索法相比,该方法大大减少了运算次数,提高了数字图像相关运算的速度。

Detection Method Study Of The Shape Of Vehicle Headlight Fixed Position Error

This paper simply describes the conception,character, composition of the image processing, also analyzes the methods of image partition and recognition that has already exist. It is emphasis discussed that a curve approximation method is applied in the measurement of the shape of the vehicle headlight fixed position error. By using the statistic properties of the correlation coefficient, a curve correlation coefficient is calculated, and then the matching point is obtained through evaluating it's derivative function. Compared to sub-pixel progressive searching method, this method enhances the speed of calculation in correlation recognition remarkably.