Application Study On Image Correction Of Fisheye Lens Used In Unmanned Aerial Vehicle

Camera as one of the main sensors on unmanned aerial vehicles(UAVs)has been widely used in the scenario of small cluster UAVs formation flying in order to achieve different purposes.For example,the camera captures visual information as an important alternative method to ensure the acquisition and transmission of information when other sensors or transmission methods between UAVs do not work well.However,traditional pinhole cameras carried by each UAV have limited Field Of View(FOV).The FOV often ranges from 40 degrees to 50 degrees,which results in a smaller visual range of each camera,hinderring the acquisition of information and the blind area between cluster UAVs will endanger the flight safety.At the same time,the fisheye lens can be used to expand the field of view and reduce the blind area.In this dissertation,the research and application of fisheye lens used in samll cluster UAV are studied,including:Comparing with the imaging model of traditional pinhole camera and analyzing the causes of distortion by introducing the optical structure and general model of fisheye lens.Summarizing four design models of fisheye lens and some distortion types of fisheye images.Taking Spherical Positioning Correction algorithm as an example,this dissertation illustrates the necessity of contour extraction.Comparing the effect of Area Statistics algorithm and Scanning Line algorithm through experiments,which shows the importance of contour extraction.In scenario of UAVs carry fisheye lens,there is often a halo outside the imaging area caused by the strong light source whose position is changing constantly.Currently,various extraction algorithms can not accurately avoid the influence of halo.Hence,this dissertation proposes a contour extraction algorithm,which reduces the voting space of the Traditional Hough Transform(THT)from three-dimensional to one-dimensional.The design of voting avoids the influence of halo or occlusion on the edge of the lens,and is suitable for the halo outside the imaging area which shape and position often changes.It takes into account the real-time performance of the algorithm and has a good extraction effect.The influence of mapping order of distortion correction on interpolation algorithm is summarized.By analyzing the principle of plane correction algorithm and space correction algorithm,the characteristics of existing correction algorithm are compared.And discussing whether it is suitable for the correction application of fisheye lens in UAV formation flight.Based on the expansion of Semi-cube Correction algorithm,a region partition method is proposed.Through the recognition and location of UAV in distorted image,the relationship between UAV’s position and segmentation area is further discussed.Under ideal assumptions,the local correction of UAV’s area is realized.