摘要
开发了一个从灰度图象实时、精确和鲁棒地提取直线和二次曲线的新技术通过利用由Zernike正交矩(ZOM)亚象素边缘检测算子提供的边缘位置、方向和曲率信息,该技术有效地融合了随机Hough变换(RHT)、异常点与伪曲线鉴别和最优曲线拟合等,并已作为低层视觉模块集成进我们的用于导引智能水下机器人(AUV)的实时水下三维视觉系统实践证明我们所开发的此技术是高速。
We developed a new technique of fast,robust and exact extraction of straight lines and conics from a gray-level image .By using the subpixel information with respect to the position,direction and curvature of an edge provided by Zernike Orthogonal Moment (ZOM)edge detection operators,the technique efficiently fuses Random Hough Transform (RHT),outier and pseudo-curve discrimination and optimal curve fitting .It has been integrated as a low-level vision module into the real-time 3D underwater vision system of our Autonomous Underwater Vehicle (AUV) Zhishui-Ⅱ to guide it in its underwater missions.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
1997年第4期69-77,共9页
Journal of Harbin Engineering University
关键词
计算机视觉
直线提取
二次曲线提取
鲁棒性
computer vision
geometric primitive extraction
subpixel edge detection
random Hough transform
outlier discrimination
curve fitting