Analysis of Blind Image Registration — On the Application of Automatic Navigation

• Main Author: 李正一
• Other Authors: 蘇順豐
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/62762971805725172175

碩士 === 國立臺灣科技大學 === 電機工程系 === 89 === The goal of an assistant system in an inertial navigation system is to provide alternative navigation information for the navigation system to update navigation parameters and to align the fly trajectory. In some confidential applications, navigation systems are demanded to be self-contained and should not communicate with external facilities. Thus assistant systems must gather navigation information by themselves. In this situation, assistant systems based on image matching become desirable. In the thesis, we extend conventional phase-correlation-based image registration techniques to solve the problem, and an assistant system based on such an image registration technique is proposed. This approach can obtain necessary navigation information, such as location and direction, by matching sensed images to a pre-stored map. Two types of systems are studied in the research. They are the basic model and the improved hierarchical model. The latter introduced the coarse-to-fine concept to reduce the computational time. Both models provide the blind registration ability and can estimate the location and direction of an observation point. A confident value is also proposed to evaluate the reliability of registration result and can be adopted as a decision factor for fusing navigation data. Complete architectures and algorithms of the assistant system is developed in the thesis, and a large number of tests that simulate real-world cases are conducted to demonstrate the performance and reliability of proposed system. Simulation results show that the proposed method has great registration ability in most situations, and the system is also very reliable.