中文摘要：

分析了线阵卫星影像投影轨迹法计算局部范围近似核线影像的基本原理，提出利用分段直线拟舍近似核线；基于RPC和物方经纬度的线阵卫星近似核线影像重采样方法，以便在密集影像匹配生成高精度数字高程模型（DEM）时缩小匹配搜索范围，并降低误匹配率。试验结果表明，利用投影轨迹法可以确定立体像对局部范围内的近似同名核线，且分段直线拟合方法精度更高；基于物方经纬度方法进行核线影像重采样后，非常有利于同名点的高可靠性匹配，且核线影像上的同名点坐标到原始影像坐标的反算精度优于0．04像素，验证了重采样方法的可行性。

英文摘要：

The basic principle of projective locus method to calculate local approximate epipolar image of linear array satellite stereo is discussed. A Line segment fitting approach is proposed to improve the precision of epipolar image. A new object space longitude and latitude based approximate epipolar image re-sampling approach is put forward. The proposed approach has well potential in dense image matching for generating high precision digital elevation models. Both searching range and mismatch possibility can be decreased. The experimental results show that the projective locus method is qualified for calculating local epipolar lines. The precision of line segment fitting is superior to that of traditional line fitting method. There is no y-parallax on re-sampled epipolar image with the proposed approach, which is advantageous for dense image matching. The precision of reverse calculation from epipolar image coordinates to original image coordinates is better than 0.04 pixel, which verifies the correctness of the proposed re-sampling approach.