Multi-source Image Registration System And Hardware Realization

Multi-source image registration technology is a kernel component of booming multi-source image fusion technologies. A physical multi-source image registration system provides solid foundations for theory research as well as utilizations for many applications, such as military, medical imaging, remote sensing and so on. This dissertation is written according to the design and realization process of a multi-source image registration system. Furthermore, some fundamental research on image registration technologies is also presented.Firstly, the design and testing process of the image registration system are presented detailed as follows:(1) An angle sensor based on magneto resistive effect is designed for the use of angle feedback. A high resolution sensor calibration approach is proposed for the special application. Moreover, a complete calibration experiment for the sensor is introduced with angle reference of stepper motor.(2) A multiple serial ports unit based on real time embedded operating system is designed to expand computer's communication capability, and thus it enables computer to communicate with all channels of Pan&Tilt units control and measurement signals.(3) A software platform of double-channel real-time image grabbing and displaying is developed based on image grabbing card. A half-automatic camera calibration module is developed independently using Zhengyou Zhang's calibration method, and it is embedded in the overall platform.Secondly, theory research on camera calibration methods is conducted. A geometrical representation of Zhengyou Zhang's calibration method is presented, and the lens distortion model in the method is analyzed and a fault is corrected, which is verified via experiments as well.At last, on a higher level, research on infrared-visible image matching is conducted. A local feature based on Harris corner and its neighbor edge segmental orientations is addressed. An infrared-visible image matching approach using the structural feature is presented, and experiments are designed to show its effectiveness compared with area feature-based methods.