基于逻辑回归的测量图像亚像素边缘定位  

Sub-pixel Edge Location of Measurement Image Based on Logistic Regression Regression

作　　者：段振云 庞文琦 张静 赵文珍 赵文辉 杜坡 DUAN Zhen-yun;PANG Wen-qi;ZHANG Jing;ZHAO Wen-zhen;ZHAO Wen-hui;DU Po(School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学机械学院,沈阳110870

出　　处：《组合机床与自动化加工技术》2021年第12期97-100,共4页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然基金项目(52005345)。

摘　　要：针对现有的高精度零件图像中亚像素边缘定位算法精度不高,光强标定复杂等问题,提出一种基于逻辑回归的边缘定位算法。视觉测量图像中的点分别属于前景区域或者背景区域,图像边缘定位实质上就是高精度提取前景区域与背景区域的分界,采用人工智能等领域经典的逻辑回归算法可以有效解决前景和背景分离的二分类问题。根据像素点的灰度值,利用逻辑回归算法对图像中的点进行分类,其边界即为亚像素边缘。在一定实验条件下,用量块直线边缘和轴承的外圆边缘和内圆边缘进行实验,并与现有的亚像素算法进行比较,证明本文算法简单且精度较高。算法中设有光强补偿系数,简化了光强标定方法,可以有效补偿光源强度造成的边缘定位误差。Logistic regression is a classical classification algorithm.Aiming at the problems of low precision and complicated calculation of existing sub-pixel edge location algorithms in high-precision part images,an edge location algorithm based on logistic regression is proposed.Each point in the image can only be foreground or background,and separating foreground and background is a binary classification problem.According to the gray value of pixels,the pixels in the image are classified by using logistic regression algorithm,and the classification boundary is sub-pixel edge.Under certain experimental conditions,experiments are carried out with the linear edge of the block and the outer and inner edges of the bearing,and compared with the existing sub-pixel algorithm,it is proved that this algorithm is simple and has high accuracy.In this paper,the sub-pixel location algorithm is equipped with light intensity compensation coefficient,which can effectively compensate the edge location error caused by light source intensity.

关 键 词：视觉测量 亚像素 逻辑回归 分类 

分 类 号：TH161[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]