Abstract: The accuracy of image measurement is heavily dependent on sub-pixel edge detection.This paper analyzes the disadvantages of traditional edge detection method based on Zernike moments that is widely used now.And it also improves the performance in many fields，such as gray-value function modeling around the edge，sub-pixel localization at curved edge，real-time performance and so on.Then this paper proposes an improved sub-pixel edge localization algorithm based on rotate-invariant moments and uses this method to measure the thickness of cable sheath.The experimental results demonstrate that the proposed method is feasible and immensely improves the accuracy of localization.

摘要： 图像精密检测精度很大程度上依赖于亚像素边缘检测精度，目前应用较广的是基于Zernike旋转不变矩的边缘检测技术。分析了传统Zernike矩边缘检测不足之处，分别在边缘灰度值函数建模、曲率较大处亚像素定位和实时性能等方面给出改进与创新，进而提出一种改进的基于旋转不变矩亚像素定位算法思想。并用此算法做了电缆护套厚度测试实验。从实验结果看，该算法定位精度有了改进，算法思想可行。