基于延时数字摄影测量的积雪过程4D监测技术研究

Snow surface monitoring from 4D structure from motion photogrammetry

为实现4D(时间+空间)多目标、高精度的积雪监测,本次试验研究采用单台相机延时拍摄结合运动结构重建算法(Structure from motion,SfM),分别获取了祁连山黑河上游站裸露山坡坡面尺度单次降雪的雪深、逐日积雪空间分布和面积,以及祁连山八一冰川1.5m×1.5m的斑块尺度全年雪深及雪面特征数据。坡面尺度积雪观测研究表明:本方法可以准确获取积雪分布信息,但其雪深空间分布获取精度较差。斑块尺度雪深监测研究表明:本方法能够很好地获取连续的雪面特征信息和雪深,且获取雪深与SR50观测雪深的绝对误差小于3.4cm。在不同季节,本方法对积雪监测能力略有差异:春季快速积累期雪面纹理少,照片组对齐并获取点云数据和DEM数据的成功率较低,而冬季和消融季雪面纹理丰富,相应的对齐成功率比例和精度较高。本研究表明基于单台相机的4D摄影测量方法能够实现小范围、连续、高精度、多目标的积雪监测,未来应用前景广泛。

In order to achieve4 D(time+space)multi-objective and high-precision snow monitoring,a singlecamera time-lapse Structure-from-Motion(Sf M)photogrammetry setup was build-up and tested at two different places of Qilian Mountains. The one test was performed to estimate snow depth,snow cover area and their distribution on slope scale at Qilian Alpine station. Another experiments was carried out next to the August-one glacier to monitor snow-surface depth and snow-surface features at plot-scale. At slope scale,the4 D Sf M photogrammetry is capable to acquire snow cover area with high accuracy. Yet the accuracy of4 D Sf M photogrammetry derived snow depth was poor at slop scale. At plot-scale,the4 D Sf M photogrammetry can obtain continuous snow surface characteristic information and snow depth well. The absolute error between the4 D Sf M photogrammetry estimated and the SR50 observed snow depth was less than3.4 cm. The4 D Sf M photogrammetry performance varies with the variation of surface condition in different season. The best performance was reached with snow surface features were abundant in winter and in melt season. It is hard for4 D Sf M photogrammetry to capture high precision and alignment achievement in spring. Our results suggest that4 D Sf M photogrammetry can achieve long-term,continual,multi-objective and high-precision monitor of plot scale snow processes.