首页|航空重力梯度地形改正

航空重力梯度地形改正

扫码查看
地形改正是航空重力梯度数据处理的重要环节,其改正精度不仅取决于地形精度和分辨率,还与改正模型有关.本文基于棱柱积分公式研究了地形精度和分辨率、观测点高程误差对地形改正的影响,导出了 一种量化评估模型.为了在不降低改正精度的前提下提高计算效率,设计了一种基于CUDA的棱柱积分并行算法,在GPU端实现了地形改正分量的快速计算.分别在地形起伏平缓和剧烈区域进行模型验证,结果表明,当测量高度大于40 m时,精度优于0.5 m的10 m分辨率地形数据可使地形改正精度优于1 E,验证了本文量化评估模型的有效性.棱柱积分并行算法在普通GPU显卡上实现了 15倍以上的效率提升,在NVIDIA TiTan V专业计算显卡上效率提升了 150倍以上,完全能够满足大范围高精度地形改正应用需求.本文的量化评估模型和并行算法可为航空重力梯度测量任务设计和数据处理提供参考.
Terrain corrections for airborne gravity gradiometry
Terrain correction is a critical part for airborne gravity gradient data processing,the quality of which is not only de-pends on elevation resolution and accuracy,but also related to the correction model.Based on the prism integration method,this paper studies the effects of terrain accuracy and resolution,and survey height error on the terrain correction results,then derived an evaluation model.To accelerating calculation of terrain correction without any approximations that may lead to a loss of accuracy,the prism method was parallelized on Nvidia's GPU card based on CUDA interface.Our model and paralleled algo-rithm were validated in both moderate and rugged terrain.The result shows that a 10 m resolution terrain dataset with accuracy better than 0.5 m can guarantee the terrain correction accuracy better than 1 E when survey altitude is higher than 40 m.Mean-while,the parallel algorithm achieves speedup of a factor of 15 on consumer GPU and a factor of 150 on professional GPU,which helps to quickly accomplish terrain corrections even in large survey areas.We confirm that our model,a simple analytic formula,presents a clear guideline for both position and terrain requirements in gravity gradient survey,our parallel algorithm proves to be practical and dramatically reduce the calculation cost while retaining the accuracy.

airborne gravity gradiometryterrain correctionselevation accuracyevaluation modelGPU parallel computing

黄佳喜、边少锋、纪兵

展开 >

海军工程大学电气工程学院,湖北武汉 430033

中国地质大学(武汉)地质探测与评估教育部重点实验室,湖北武汉 430074

航空重力梯度 地形改正 高程精度 量化评估模型 GPU并行计算

国家自然科学基金国家自然科学基金国家自然科学基金

420740104197141642122025

2024

测绘学报
中国测绘学会

测绘学报

CSTPCD北大核心
影响因子:1.602
ISSN:1001-1595
年,卷(期):2024.53(8)