| In recent years,the earth observation technology has developed rapidly and the earth observation system has been increasingly completed.Foreign high-resolution earth observation remote sensing satellite technology represented by IKONOS,WorldView,and Pleiades has been very mature.China's high-resolution civil mapping satellite headed by ZY-3 has also reached the world's advanced level.The above-listed high-resolution line array satellites can form stereoscopic observations(such as the ZY-3 of the three-line array sensor can provide three-dimensional stereo,namely front view,bottom view and rear view),thus providing an effective and reliable source of data for the extraction of three-dimensional information.The application of three-dimensional information extraction of high-resolution linear array satellite stereoscopic observation data includes,but is not limited to,digital terrain model generation,urban three-dimensional modeling,natural disaster assessment,and target location and tracking.A core technology for extracting three-dimensional information from satellite imagery is stereo matching(dense matching)technology,and the epipolar image serves as input data for dense matching,and its accuracy directly affects the reliability of the matching results,thereby affecting the accuracy of the three-dimensional information.Therefore,correct epipolar image input is very important for dense matching.Different from the frame image,the epipolar line of the linear array image presents a hyperbola,which poses a great challenge to the generation of the epipolar image.The traditional epipolar image generation method is suitable for a single stereoscopic observation mode(such as the along-track or cross-track)and lacks generality.In addition,with the increasing number of earth observation satellites,the possibility of forming stereoscopic observation between different platforms is becoming more and more likely.Therefore,it is a need and trend to study a generic method of generating the epipolar image which can be applied simultaneously to the along-track,cross-track and multi-source stereoscopic images.This article focuses on two major aspects:improving the generality of the epipolar image generation method for linear array image and the improvement of epipolar line fitting accuracy.The main contents are as follows:(1)The characteristics of the epipolar line of the linear array image were reviewed,and the relevant characteristics were verified one by one.Based on this,the differences in the curvature of different experimental data were studied,which laid the foundation for the next epipolar line rectification.(2)The characteristics of the existing epipolar image generation methods are summarized,and the advantages and disadvantages of two types of methods(the image-space based method and the object-based method)are summarized.Based on these two methods,we provide a theoretical support for the next generation of generic epipolar image generation methods,and take them as experimental comparison methods.(3)The definition of generality is given,and a hybrid image-object space-based epipolar line generation method based on the combination of the image-space based and the object-based is proposed.This method combines both advantages.In order to facilitate the post-processing of epipolar images,an improved method for generating RPC(Rational Function Coefficient)for epipolar images was proposed based on the original method,which is a generic epipolar image generation method.Several groups of epipolar experiments show that the proposed method has the same level of vertical parallax,positioning accuracy,single epipolar line rearrangement accuracy,and the ability to maintain the linear relationship between horizontal parallax and elevation,and is even better than the comparison method.This method has strong generality and high robustness. |
PDF Full Text Request