Research On Key Technologies For Reconstruction Of Three-dimensional Model Based On Hyperspectral Imaging

The reconstruction of the object model is based on the characteristics of the object features,and the virtual simulation is matched with the real object.The geometric model and optical model of the object are constructed by computer graphics and remote sensing optics,which makes the simulation model correspond to the real object in terms of geometric and optical characteristics,and saves the feature information of the target.This topic helps to deepen the research of optical remote sensing target characteristics,promotes the development of remote sensing multidimensional information extraction and fusion technology,and provides analysis and verification methods for the development of new types of instruments.For high-resolution remote sensing scenes with a wide variety of features,variable geometry,and high spectral characteristics,the traditional methods of spectral reconstruction of ground objects has the characteristics of heavy workload,and inevitably there is a classification error and material mapping error.So it is difficult to accurately correspond to the spatial spectrum and it is impossible to achieve the requirements of fine modeling.The resulting model is not practical.Aiming at the need for high-precision and high-resolution measurement of the new hyperspectral remote sensor and the current target situation that the data measurement of reconstructed models has a large workload and cannot be precisely modeled,this paper researched the simulation principles and key technologies of the ground model based on hyperspectral imaging.Based on this,using the characteristics of each pixel in the hyperspectral image data combined with its spectral data,it is proposed to extend the bidirectional reflectivity function to spatially varying bi-directional reflectivity distributed by pixels.Combining feature detection and matching methods of invariable scale,the feature matching points of the target in the spectral image sequence are determined.The three-dimensional coordinates of the space points corresponding to all the feature matching points and the shooting positions of the spectrometer were calculated by the method of structure from motion.The results were optimized by using random sampling consistency estimation and beam adjustment method to obtain the target geometric model.Using the characteristics of the spectral image,the bidirectional reflectivity factor value of each feature pixel point is obtained,and the direction vector between each feature space point and the photographing position of the spectrograph is calculated to determine the bidirectional reflectivity direction of the feature point.The bidirectional reflectivity factor values and directions of all the feature points are collected to obtain the bidirectional reflectivity distribution of the target sparse spatial variation,and finally the three-dimensional model reconstruction of the feature is completed.Finally,aiming at the characteristics of the physical parameters of the instrument and the particularity of the outer space's operating environment for the on-orbit operation of satellite-based remote sensors,the paper designed and completed the simulation of instrument line shape function updating and pointing error correction for on-orbit instruments in space.Experimental results show that the method based on hyperspectral image sequences can quickly obtain the target three-dimensional geometric model and spatially varying bidirectional reflectivity distribution,also make the modeling of the target optical features fine to the unit of pixels.This method has good practicality,automatically rebuilds the three-dimensional model of the observed object,and enables the geometric model of the object to fuse with the optical model reflecting the physical properties of the object.In addition,the proposed method for the instrument line shape function updating and the pointing error correction method works well,which solves the problem of in-orbit parameter calibration of the actual remote sensor.The results of the dissertation provide important parameters and evaluation methods for the design of major new remote sensing instruments and the application of remote sensing for accurate detection of ground targets.