联合LoFTR与自适应IRLS的立体测绘卫星高程误差检查

Elevation error checking of stereo mapping satellite based on LoFTR and adaptive IRLS

立体测绘卫星产品精度受多种因素的影响,采用一定的技术手段对立体像对接边的平面高程误差进行准确和全面的量测,不仅可以有效保障测绘产品的生产质量,还有助于检查影像内部可能存在的畸变和异常。本文提出了一种立体测绘卫星高程接边误差检查方法,首先通过LoFTR算法对具有重叠的异轨立体卫星像对进行特征匹配后获取大量的高精度匹配点;然后找出能够分别在两个像对上进行前方交会的匹配点,并计算这一位置的高程接边误差;最后通过自适应迭代再加权最小二乘法(IRLS)方法对单一区域中的匹配点对应的高程接边差的平均值进行稳定性估计,剔除检查结果中的异常值,得到可信结果。通过国产资源三号立体像对和高分七号立体像对的试验,证明了该方法达到的精度水平足以满足立体测绘卫星高程误差检查的精度需求,并验证了该方法的有效性。

The mapping accuracy of stereoscopic surveying and mapping satellite is affected by many factors. Quality of the surveying and mapping products can be effectively ensured by adopting precise and comprehensive measurements of the plane elevation error on the butt side of the stereoscopic images using certain technical means, while possible image distortions and anomalies inside can be filtered. In this paper, a method for checking the elevation edge error of stereo mapping satellites is proposed. Firstly, a large number of high-precision matching points can be obtained by feature matching of overlapping out-of-orbit stereo satellite image pairs through LoFTR, and screen for the matching point of forward intersection simultaneously. Figure out the elevation edge error of this position. Then, use the adaptive IRLS method to robustly estimate the average value of the elevation edge difference corresponding to the matching point in a single area. Finally, the reliable result can be obtained by screening out the inspection result outliers. By evaluating the experiment results of the ZY-3 and Gaofen-7 stereo pair, it is proved that the accuracy level achieved by the method is sufficient to meet the accuracy requirements of the elevation error inspection of stereo mapping satellites, and the effectiveness of the method has been validated too.