The Research Of Aerial Image Panoramic Generation And Its Geographical Registration

Panoramic generation and geographic registration of aerial images is a hot research direction in the field of computer vision.It not only has a wide range of application requirements in civil fields such as industrial manufacturing,transportation planning,geographic survey,disaster assessment,etc.,but also plays an important role in the field of military reconnaissance and surveillance.However,the aerial image of UAVs generally suffers from geometric distortion,uneven distribution of feature points,and interference of moving targets.It is prone to ghosting and parallax when generating panoramic images,and the accuracy of georeferencing is difficult to guarantee.Therefore,based on the analysis of existing panoramic stitching and visual positioning technologies,the techniques of orthorectification,panoramic generation and georeferencing of aerial image of unmanned system are studied,and the feasibility and effectiveness of the proposed algorithm are verified.Firstly,the principle of geometric deformation of aerial images is analyzed.Based on this,the image distortion correction algorithm based on camera calibration and the orthorectification algorithm based on digital elevation model and ground control point are studied.Aiming at the shortcomings of high computational complexity and poor real-time performance of the traditional orthorectification algorithm,this paper proposes an orthorectification algorithm for aerial images based on flight attitude parameters: the corresponding relationship between image points and actual object points of aerial images is calculated by using the pitch angle,yaw angle,roll angle and flight height of unmanned aerial vehicles,and a geometric correction model based on flight attitude is established,thereby realizing rapid correction of aerial images.Then,an image registration algorithm based on block feature detection is proposed to reduce the computational complexity of the registration algorithm.In view of the fusion ghost in panoramic stitching,this paper proposes an image fusion algorithm based on multiresolution fusion and suture line search,and simultaneously removes the motion region of aerial images to be registered,thus improving the display quality of aerial panoramic images.Finally,the traditional aerial image positioning technology mainly relies on GPS position information.This paper studies the feasibility of aerial image matching with map from the perspective of feature matching.This paper extracts aerial image roads based on connected region features,generates shape context descriptors of aerial image roads and electronic map roads respectively,and then matches them.In order to solve the problem that it is difficult to extract the connected region features of large-field panoramic images,this paper proposes an algorithm that uses fast Hough transform algorithm to extract straight line segments,and then uses RANSAC method to robustly estimate the image transformation parameters.Finally,the effectiveness of the proposed algorithm is illustrated by experiments.