Fast Extraction and Matching of Straight Lines on Sequential Images
Zhou Xing-chun · Nanjing Hangkong Hangtian Daxue xuebao · 2005
Aimed at the real-time application, such as air vehicle visual navigation, a scheme for fast extracting and matching line segments on the aerial sequential images is proposed. In straight line extraction, a fast connected-component labeling algorithm is employed to improve the extraction speed. And when tracking line segments, a three-step method and a metric robust to the noise are proposed. Firstly, three line segments are selected as a reference model. Subject to the orientational, positional and physical constraints, the candidate model is constructed. Then the exact matching model corresponding to the reference model is determined based on the eigenvector technique. Based on the reference and matching model, other straight line correspondence is determined. Experiment on the real images demonstrates that the proposed matching method is more than 30 times than the existing method at computation speed. It can meets the need of the real-time application.