Research Of Image Rectification Technology For Omni-directional Vision Navigation

Omni-directional vision system can one-time obtain the 180 degree image of the scene in the three-dimensional angle. As with the notable features of large visual perspective, the application of omni-direction vision system to military and civilian areas increasingly widespread. But due to its special optical imaging system, images generated by omni-directional vision equipments exist serious distortion. Therefore, if we want to put omni-directional vision into the real application, restoration of the scene information from the omni-directional image is needed. This paper studies the distortion correction problem of omni-directional vision image taken by fisheye lens.The geometric distortion correction of an image is an operation to transform an image to an ideal image by the means of special transformation. This technology plays an important role in the omni-directional vision system. It lays a good foundation for target identification, tracking and positioning in the system. The omni-directional vision system broadens the applicability of the recognition system, optimize the performance of the system and increase the system's real-time performance and reliability. The major distortion error of fisheye lens is divided into three categories: radial distortion, eccentric distortion and thin prism distortion. Under the premise of assuming that there is no or small distortion in the lens centre and near the centre , we make standard calibration board to get the ideal and the actual value of the calibration reference points pixel coordinates. At the same time, we apply the Support Vector Machine to image processing area, which is used to generate coordinate mapping table by the means of the radial distance regression. We use coordinate mapping table in the correction procedure to correct image distortion, so we can get distortion-free images.This paper firstly introduces the omni-directional vision technology, the current research status and the specific application status of this technology. Then it introduces camera imaging theory and camera calibration theory. It gives the method of the calibration for fisheye lens on the basis of the analysis of the fisheye lens imaging distortion model. This paper describes the related theory of image processing, especially the correction method for the geometric distortion produced in the images. This paper designs and implements an application software. This software achieves the effective and real-time correction for the distortion images and the handled images can be a single frame or Continuous frames. It lays the foundation for omni-directional vision navigation objective positioning and identification.