Research On Fusion And Location Of Spaceborne Photon Counting Point Cloud And High Resolution Satellite Stereoscopic Image

Now,our nation’s remote sensing and mapping satellite is capable of making worldwide observations,but the accuracy of stereo mapping is constrained due to the difficulty of obtaining high-precision control data in difficult and unreachable places.Earth observation technology’s photon counting point cloud may supply control information for satellite remote sensing images for three-dimensional mapping,which can increase the mapping precision of satellite remote sensing images without any control data.Therefore,it has become a focus of study to fuse and locate under a single spatiotemporal benchmark using optical remote sensing pictures and photon counting point cloud data.Both of these types of heterogeneous data have benefits and drawbacks.The photon counting point cloud has a sparse distribution,a low density,and a resolution that is significantly lower than that of the optical remote sensing picture data,but it has a higher altitude positioning accuracy than the optical remote sensing image.While the photon counting point cloud’s altitude positioning accuracy is greater than that of an optical remote sensing picture,it has a sparse distribution and low density,and its resolution is plainly lower than that of an optical remote sensing image.The registration and fusion processing of ICESat-2 satellite photon point cloud and image matching point cloud data is investigated in order to address the issue of satellite carried photon counting point cloud assisted highresolution remote sensing satellite Beijing 3 stereo image location.As high-precision control points,the registration points obtained from photon point cloud data may fully use their individual advantages and raise the geometric positioning accuracy of satellite pictures.Research and accomplishments comprise:(1)The current status of research into image registration fusion,satellite count point clouds assistance satellite image block adjustment is examined and assessed.In terms of registration accuracy,robustness,and automation level,both existing point cloud and image registration fusion techniques and their enhanced algorithms still have certain drawbacks.Through literature review,the current research progress was analyzed and organized,and a research plan was designed to provide guidance for subsequent research.(2)A method for combining coarse and fine registration of dense matching results between spaceborne photon counting point clouds and satellite stereo images is studied.First,use the SAC＿IA and NDT algorithms to roughly register two sets of 3D point data,count the foundational parameters for exact registration.Then,the ICP registration method is used for precise registration to obtain the optimal solution and achieve precise registration of two point clouds.Eventually,the distance threshold method is used to eliminate misregistration points.Through the above steps,an accurate geometric mapping relationship between two sets of point cloud data is established,laying a data foundation for the next step of image adjustment processing.(3)The adjustment method of Beijing-3 satellite stereo image block adjustment supported by spaceborne photon point clouds is studied.First,the feature points in the registration points of the photon point cloud data are extracted using the 3D SIFT technique as control points and check points.Then,the adjustment model corrected by affine transformation of the imageadding square of rational polynomials is used for block adjustment.Comparing the positioning accuracy of Beijing-3 satellite images under controlled and uncontrolled conditions,experiments have proven that adding ICESat-2 satellite photon point clouds as control points effectively improves the geometric positioning accuracy of Beijing-3 satellite images.(4)Harbin’s Beijing-3 satellite image data and the ICESat-2 satellite photon point cloud were used for experimental study.The 3D-SIFT method retrieved 422 feature points,of which35 were chosen as control points(such as building corners and road locations)and the rest 364 were employed as checkpoints for block adjustment modification.Under uncontrolled conditions,the plane positioning accuracy of the Beijing-3 satellite image is 4.273 m,and the elevation positioning accuracy is 4.221 m.After adding ICESat-2 point cloud,its plane positioning accuracy is 2.039 m,and its elevation positioning accuracy is 0.642 m.Compared to the uncontrolled condition,the plane positioning accuracy has been improved by 52.2%,and the elevation accuracy has been improved by 84.7%.