摘要
建立了空间二次曲线和平面二次曲线坐标概念 ;提出了一种适合于光流计算的空间二次曲线表示方法 ;提出了一种较为简单的二次曲线求解方法 ;给出了一种效果较好的边缘检测新方法 ,即把模板为 3× 3的
The interpretation of three dimensional motion and structure of a moving object from a sequence of images is a key step in computer vision tasks of robotic navigation, obstacle avoidance, recognition of solid objects, and surveillance. One of the fundamental problems is the measurement of motion in a time varying image, frequently called optical flow which is an intermediate representation widely used for recovering three dimensional information. In this paper, a new method for representing a conic with uniqueness in three dimensional space is proposed that is convenient to compute and infers optical flows from a sequence of images. The concept of the coordinates of the conics in two dimensional and three dimensional space is defined. The coordinates space of a conic in three dimensional space is a nine dimensional space. The Sobel's edge detector that is the best for a synthetic image or a real image with noise is used to detect edge curve from digital images. Furthermore, an advanced and effective method of 3×3 edge detector that combines the Sobel's edge detector with edge thinning and contour following is proposed. A new algorithm for extracting edges (conics) from a primitive image, namely the projections of conics in three dimensional space, is established. All of the algorithms such as edge detection, edge thinning, contour following, edge linking, and edge representation have been implemented on artificial images and images with real world data.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2003年第2期203-206,共4页
Journal of Northwestern Polytechnical University
基金
国家自然科学基金 (6 0 2 75 0 37)
航空科学基金 (99F5 30 6 5 )
南昌航空工业学院测试技术与控制工程研究中心开放基金 (2 0 0 1- 15 )
关键词
二次曲线
边缘检测
边缘细化
轮廓跟踪
光流
conics, edge detection, edge thinning, contour following, optical flow
