Background The Unmanned Aerial Vehicle(UAV)remote sensing technology has provided a new technical mean for soil and water conservation monitoring in construction projects, especially in terms of the calculation of the area and volume of disposal ground, also greatly improved the efficiency and accuracy of monitoring. However, the accuracy of different models processing UAV remote sensing data varies a lot. Based on the Yellow River-to-Baiyangdian Water Transfer Project, this study selected 5 disposal grounds in Puyang, a city located in the plain area, as the research object. Since these 5 disposal grounds are similar in location and natural conditions, it is convenient for conducting comparative experiments and comparing the calculation accuracy of different models processing remote sensing data.

Methods In this study, PhotoScan and Pix4D were used to process the UAV remote sensing data to obtain DOM and DSM images of each disposal ground. Global Mapper, LocaSpace Viewer and Context Capture were used to extract information from DOM and DSM images to calculate the area and volume of disposal grounds. Then 6 sets of models processing UAV remote sensing data were structured:Pho-Glo, Pho-Loc, Pho-Con, Pix-Glo, Pix-Loc and Pix-Con. Based on the actual value of construction organization, we quantified the errors of the area and volume of each disposal ground calculated by different models.

Results 1) Using these 6 models:Pho-Glo, Pho-Loc, Pho-Con, Pix-Glo, Pix-Loc and Pix-Con, the calculation errors of disposal area were 5.57%, 5.05%, 4.84%, 1.69%, 3.06% and 1.23% respectively, and the errors of disposal volume were, 9.06%, 10.28%, 4.76%, 5.73%, 6.52% and 2.97% respectively. 2) When calculating the disposal area, using Pix4D for preliminary processing significantly reduced the error. There was no significant difference among Global Mapper, LocaSpace Viewer and Context Capture as for the information extraction. 3) When calculating the disposal volume, using Pix4D for preliminary treatment significantly reduced the error. There was no significant difference between Global Mapper and LocaSpace Viewer to calculate the volume of disposal grounds, while the accuracy of Context Capture was significantly higher than that of the others. 4) PhotoScan processed images more accurately when there was water on the surface and the DSM images were more consistent with the actual situation.

Conclusions The accuracy of the 6 models are quite different, though all of them meet the requirements of relevant regulations. UAV has a bright application prospect in soil and water conservation monitoring for construction projects, which is more efficient and accurate than traditional monitoring methods when calculating the area and volume of disposal grounds. It is suggested that Pix-Con model processing UAV remote sensing data should be popularized in monitoring of construction projects.