The Image Matching Technique Applied Research, Positioning And Fire

In modem war, a precise orientation and emendation system applied to artillery spying is needed, existing orientation and emendation system usually computes position of object according UAV(Unmanned Air Vehicle)'s position and pose and the pose of vidicon on UAV, but orientation isn't enough precise. Realtime video image from vidicon on UAV may be matched to digital map stored by image registration technique, which can more precisely orientate object and has higher reliability. The dissertion is dedicated to applying image registration to orientation and emendation system on artillery spying UAV, and effective algorithms of image registration are presented.In the beginning, some general ways of image pretreatment technique are introduced, including geometric correction methods of image, radiometric correction of image and filtering wave and removing noise. Video image with strong noises is well restored by a integrated filtering wave methods among some non-linear filtering wave methods, which prepare for the image registration following.In the second part of this paper, firstly, the theory of image registration and usual techniques are introduced. Secondly, a algorithm of image registration based NPROD (Normalized Product) and gauss- pyramid decomposition is presented, which based on the hypothesis that there is Aimier geometric transform among images to be matched, and the a Igorithm is proved to be precise and effective in mul-modal image registration, but it's count is great. In order to improve the algorithm, a algorithm integrating with optic flow is presented by Introducing the theory of estimating motion between video frames by optic flow.In the later part of this paper, image registration algorithm based NPROD and gauss- pyramid is performed on DSPEVM (Digital Signal Processor Evaluation Module)board, and is proved to be precise and reliable.