Research On The Characteristics Of Collapsing Erosion Based On Structure From Motion Photogrammetry

In order to study the accuracy of the application of structure from motion photogrammetry in analyzing the characteristics of collapsing erosion,this research analyzed the measurement accuracy of SFM-MVS(Structure from Motion-Multi View Stereo)photogrammetry under different conditions through the indoor scouring experiment,improved and analyzed the measuring method and the source of error of SFM-MVS photogrammetry in the process of monitoring soil erosion.This research also analyzed the measurement accuracy of SFM-MVS low-altitude oblique photogrammetry through field in-situ test,conducted characteristic analysis of the process of soil erosion within the collapsing hill research area and analyzed the impact of rainfall intensity on the collapsing erosion degree.The following conclusions are drawn from this experiment:(1)The average measurement accuracy(without threshold value)of Pix4D and Smart 3D are 85.64%and 84.94%respectively,so the Pix4D has a better matching effect and a higher accuracy.The minimum threshold value can significantly correct the errors of DSM generated by SFM-MVS close-range photogrammetry in soils with little erosion.After using the minimum threshold value in the complex soil physiognomy,the total average measurement accuracy of DOD(Digital Surface Model of Differences)in the river sand scouring experiment is 88.59%while the undisturbed colluvial soil reaches 93.25%.The setting of minimum threshold could improve the measurement accuracy of DOD and make the measurement result obtained more stable.Whether or not the threshold value is used,the DOD measurement accuracy generated by the triangulation method is higher than that of the inverse distance weighing,and the result is more stable.(2)In the river sand catchment scouring experiment,it is found that the reason of large error of DOD is that the hang surface generated by the lateral erosion of trench wall can lead to the partial information distortion of DSM,which can decrease the measurement accuracy by about 9%.This is the drawback of DSM data,which cannot be avoided even with high precision LIDAR.However,the results of the experiment show that even though the problem exists,the affected average measurement accuracy of the DOD can reach around 84%,and the DOD average accuracy of all sequences can reach 88.59%.Through the catchment scouring experiment of the undisturbed colluvial soil,it is found that DSM enjoys a high accuracy with a minimum error of only 0.14 mm.The offset of repeated DSM is only 1 mm,and the stability is high.(3)The image processing results of unmanned aerial vehicle show that the SFM low-altitude oblique photogrammetry enjoys a good matching effect,the total average re-projection error is 0.00047 mm and the total root mean squared error at the ground control point is 13.6 mm.The average difference between the repeated point cloud is only 0.137 mm,which presents that the SFM low-altitude oblique photogrammetry has an excellent repeatability.The mean absolute value of the vertical error between the calculated values of SFM close-range DODLOD and SFM low-altitude DOD is 5.6 mm.After the application of threshold value,the mean absolute value of the error decreases to 2.6 mm,while the measurement accuracy of SFM low-altitude DODLOD increases by 53.57%.The vertical error between the calculated DODLOD values of SFM close-range DSM and SFM low-altitude DSM at the same time is 4.4 mm.Under the condition of this experiment,the measurement error of SFM low-altitude oblique photography is about 5 mm.The result shows that an appropriate minimum threshold value can significantly improve the measurement accuracy of unmanned aerial vehicle low-altitude oblique photogrammetry,which can be applied to the monitoring of soil erosion in a wide range.(4)Within the whole monitoring period:Almost all erosion gullies on the collapsing wall in the west side of the research area have undergone different degrees of erosion,among which the most severe erosion is found in the big erosion gullies,and only a small part of the collapsing wall experiences the process of sedimentation.Two small collapsing walls in the central and northeast of the research area have undergone severe erosion in each erosion gully,the erosion thickness of which is over 10 cm.The accumulative total soil loss amount of the collapsing wall in the west side of the research area within the monitoring period reaches 108.6488 m3.Although generally speaking,the loss of soil increases with the increase of rainfall,there is no significant linear relationship between them,which is mainly related to the self-property of the collapsing soil and the eroding force of single rainfall.