Research on multi image registration technique based on distributed aperture system
|School||University of Electronic Science and Technology|
|Course||Detection Technology and Automation|
|Keywords||Image registration distributed aperture sensor system SIFT feature cameracalibration|
A single sensor can rarely give a complete picture of its immediate surroundings. To overcome this limitation, increasing numbers of diverse sensors are being deployed on manned and unmanned platforms in order to extract as much information about the local environment as possible. Image registration and fusion process is also more in line with the visual characteristics of humans and machines, More conducive to the recognition and understanding of the target scene, Image registration techniques has been widely applied in many fields, such as:Remote sensing image processing, Automatic monitoring of industrial pipeline, Medical Image Processing and Character recognition, as the same as the hot spot in computer vision:Three-dimensional reconstruction, Stereo matching, Binocular vision and Virtual Reality and so on.For different applications, Different methods are applied according to different requirements on the registration accuracy and speed of image registration techniques. In this paper, we study the application environment of the distributed aperture system deeply, and purpose effective solutions, the main achievements are listed as follows.1. The imaging process and the geometric principles of image acquisition devices has been analyzed in depth. The camera motion model and two-dimensional image transformation model have been studied. And the traditional image registration algorithm was analyzed and evaluated.2. The image registration method based on SIFT features has been studied deeply, and the algorithm implementation process has been analyzed and Optimized.3. The camera calibration technique has been researched, especially the flexible calibration method purposed by Z. Zhang, which only requires the camera to observe a planar pattern shown at a few (at least two) different orientations. And then propose effective Measures to improve the algorithm stability.Based on the above research works, firstly, the multi-image registration system for the DAS system has been completed in two different ways, and a compare between the two strategies has been made through several group tests. The results show that the second strategy designed in this paper can maintain generators state validly and decrease the Computational complexity.