Research On Image Positioning Optimization Algorithms For UAV Platform

As a powerful supplement to aerospace and aviation remote sensing,UAV remote sensing platform is characterized by flexibility,low cost,high safety and strong timeliness,and has been widely used in environmental monitoring,surveying and mapping,emergency relief and other scenarios.When the UAV platform is used for aerial photogrammetry,it is usually required to carry out three-dimensional reconstruction of the scene and orthophoto image generation in the flight area.UAV image positioning is one of the key technologies.The UAV image positioning can obtain the attitude information when the camera is shooting,and further calculate the real three-dimensional model of the scene,which provides a good foundation for the subsequent information analysis and management.However,the UAV platform is susceptible to wind and rain and other environmental factors,and the acquired images have problems such as large attitude change,irregular overlap,and small base to height ratio,leading to large image feature matching error and high matching error rate.At the same time,the UAV image usually only has GPS information,but does not have attitude information,which cannot completely eliminate the accumulated error when using the traditional method of beam adjustment,resulting in low image positioning accuracy or even completely wrong positioning results.In order to locate the UAV image quickly and accurately,this paper studies how to accelerate the UAV image positioning and improve its positioning accuracy based on the image mismatching elimination method based on visual characteristics and the theory and method of computer vision motion recovery structure.This paper mainly studies the optimization method of UAV image positioning based on motion recovery structure,and uses motion grid estimation method,hierarchical positioning method and GPS assisted bundle adjustment method to realize the acceleration and improvement of positioning accuracy of UAV image.Based on the constraint of smoothness of image motion,the method of grid motion statistics is used to distinguish correct and wrong matching,which effectively guarantees the model estimation accuracy of RANSAC method and the accuracy of image matching.Based on the idea of divide and conquer,the scene map is constructed and divided by image matching and correlation information,and each sub-scene is processed by hierarchical motion restoration structure method,so as to accelerate the positioning of large-scale UAV image.Based on the GPS information of UAV image,a uniform cost function that takes into account both the re-projection error and the spatial distance error is constructed by fusing the visual feature constraint and spatial geographic information constraint of the image to realize the GPS assisted adjustment optimization of UAV image positioning.At the same time,this paper verifies in multiple experimental data sets,and obtains good experimental results,fully proving the effectiveness of relevant algorithms for UAV image positioning acceleration and positioning accuracy improvement.