Abstract: Aiming at the problem that location precision of existing subpixel edge location algorithm is not high, based on computer simulation and common physical experiments, this paper compares edge precision and ability of noise canceling for three kinds of representative sub-pixel location algorithm such as 1D gray moment method, Gauss fitting method and polynomial interpolating method. Experimental result shows that edge locating repeatability is stable of three kinds of algorithms when SNR is beyond or equal 40 dB, error of edge locating repeatability is lower than 0.01 pixel.

Key words: machine vision, edge detection, operator, stability, repeated location

摘要： 针对现有亚像素边缘定位算法定位精度不高的问题，基于计算机模拟和普通物理实验，比较一维灰度矩法、高斯拟合法和二次多项式插值法3种典型亚像素定位算法的边缘定位精度和抗噪能力。实验结果表明，3种算法在信噪比SNR≥40 dB时，边缘定位稳定性较好，边缘定位重复性误差小于0.01个像素。

关键词: 机器视觉, 边缘检测, 算子, 稳定性, 反复定位

CLC Number: 

• TP391.4