摘要
图像中心子像素定位是视觉测量的关键技术之一。为满足多领域现场精确标定和测量要求,针对现有子像素定位方法过程复杂、定位精度受初始边缘提取效果影响大,对噪声抵抗能力差、定位精度低等问题。提出一种遥感多视点图像中心子像素定位方法。采用Canny算法对遥感图像进行边缘提取,以增强的Snake方法对边缘提取结果进行全局最优搜索。引入邻域贡献权值系数细化遥感多视点图像边缘,在图像中心子像素边缘点的梯度方向对灰度值进行高斯拟合,确定多视点图像中心子像素边缘位置,针对图像中心子像素边缘突变点提出利用最小二乘法来拟合图像中心。实验结果表明,上述方法在有噪声影响下仍能给出精确的图像中心定位,定位精度较高。
The central sub-pixel positioning is one of key technologies in visual measurement. In order to meet requirements of accurate on-field calibration and measurement in many fields, a positioning method for central sub- pixel in remote sensing multi-view image is proposed. Canny algorithm is used to extract the edge of remote sensing image, and the enhanced Snake method is used to search the global optimum solution of edge extraction result. The neighborhood contribution weight coefficient is introduced to refine the edge of multi-view remote sensing image. Gaussian fitting is performed on gray value in the gradient direction of sub-pixel marginal point in image center, so as to determine the position of central sub-pixel edge in multi-view image. Finally, for the mutation point of central sub -pixel in image, the least square method is used to fit the center of image. Experimental results show that the proposed method can give accurate localization of image center under noise, which has high precision in localization.
出处
《计算机仿真》
北大核心
2018年第3期377-380,共4页
Computer Simulation
基金
广西自然科学基金项目(2015GXNSFAA139295)
2017年度广西高校中青年教师基础能力提升项目(2017KY0628)
梧州学院2016年校级科研项目(2016B009)
关键词
遥感
多视点图像
中心子像素
定位
Remote sensing
Multi-view image
Central sub-pixel
Localization