Research On 3D Reconstruction For Stereo Images Of The IKONOS-Satellite

Automatic target recognition (ATR) is the key of the Imaging guidance technology.But the target not only interweaves with the complicated background, but also changes with time, season, scene, etc. The characters of the target put up in the images are unstable. It is difficult to recognise the target if only depend on the realtime image of the aerocraft, espacially the surroundings of the aerocraft is bad. It's hard for the realtime image processing system to recognise the target from the realtime image. This mode of ATR is still a problem around the world.The main trend of the ATR is that using the 3D reconstruction production of satellite images to produce the front elevation of the interested area. And then the image matching processing was employed between the front elevation and the realtime image to search the target. The most important step of this method is the 3D reconstruction of the satellite images.This dissertation focuses on the 3D reconstruction technology. Aiming at the linear CCD push-broom of high resolution IKONOS satellite imagery, the main aspects discussed are theory and application on epipolarity model, stereo image matching, and sensor geometry model.The properties of epipolar of the linear CCD push-broom images are approximate to the frame perspective images. The linear CCD push-broom image can apply the properties of the epipolar to the image matching. The shadow area was segmented out form image. The matching of the shadow area was guided by the other areas. Pyramid decomposing and bidirectional matching were adapted to resolve the big geometric deformation problem of the linear CCD push-broom image. The 3D orientation was finished. The parallel projection model was used to calculate the 3D information of the IKONOS imagery. The experiment result turn out that the accuracy achieved can fulfill the requirement.