Research On DEM Establishment Methods For Gully Of Different Scales Based On Point Cloud Data

Gully is one of the main manifestations of soil erosion,and the development of gully has a serious impact on local soil erosion and ecological environment.With the advancement of measurement methods,UAV airborne LiDAR to obtain point cloud data has become the main means of gully data collection.In this paper,five gullies in Xide County,Liangshan Yi Autonomous Prefecture,Sichuan Province were selected as the research area,and they were divided into small gully(width less than 20 m),mediumsized gully(width 20～50 m),and large gully(width greater than 50 m)according to the different widths of the gully development scale.The CSF filtering algorithm,TIN filtering algorithm and morphological filtering algorithm were used to filter the point cloud data,and the accuracy of the point cloud data processed by different filtering algorithms was compared,and the elevation error method and the comparative gully morphological parameter method were used to explore the accuracy of the DEM of different scale gully established after different point cloud filtering processing.The main findings are as follows:(1)The point cloud filtering error of the area where the gully is located at different scales is different under different filtering algorithms.The error of CSF filtering algorithm in small gully area is the smallest,and the error of TIN filtering algorithm in medium and large gully is the smallest.After filtering and classifying the eroded trench point cloud data by CSF filtering algorithm,TIN filtering algorithm and morphological filtering algorithm,the original point cloud data of the area where the gully is located of different scales were intercepted and ground points were manually classified as standard data.After comparing and analyzing the error results of the three point cloud filtering algorithms in gully at different scales,it is found that the filtering errors of the three groups of filtering algorithms in large,medium and small gully are low,and the accuracy of point cloud data ground points is high.For small gully,the accuracy of CSF filtering algorithm was the highest,with Class Ⅰ error of 3.05%and Class Ⅱ.error of 2.45%;for medium-sized erosion grooves,the accuracy of TIN filtering algorithm was the highest,of which ClassⅠ error was 1.48%and Class Ⅱ.error was 3.69%;for large erosion grooves,the accuracy of TIN filtering algorithm was the highest,of which Class Ⅰ error was 3.25%and Class Ⅱ.error was 6.41%.(2)Using the elevation error method,the error between the DEM elevation value and the measured elevation value of RTK was compared,and the DEM accuracy of different scale gully established based on the point cloud data processed by different filtering methods was analyzed.The elevation statistical error of small gully was large,the RMSE value was higher overall,and the data dispersion degree was high.Comparing the small gully elevation errors established by the three groups of filtering algorithms,it is found that the DEM elevation error values established by the CSF filtering algorithm are the lowest.For medium-sized gully and large gully,the DEM error values established by TIN filtering algorithm are better,and the DEM elevation error RMSE established by the TIN filtering algorithm in medium-sized gully is only 0.07 m,and the data accuracy is high.(3)Based on the different scale gully DEM established after processing by different filtering algorithms,four morphological parameters of gully in the study area,including volume,average width,average depth and length,were extracted respectively,and the standard length of gully was obtained by visual interpretation of high-definition remote sensing images.Comparative analysis showed that the volume error of small erosion grooves was between 0.02%and 0.1%,the volume error of medium erosion grooves was between 0.03%and 0.3%,and the volume error of large erosion grooves was between 0.05%and 0.4%.Comprehensively comparing the differences in DEM morphological parameters of different filtering algorithms and gully of different scales,it is concluded that the CSF filtering algorithm is better than the morphological filtering algorithm and TIN filtering algorithm for small gully.For medium and large gully,the TIN filtering algorithm is superior to the morphological filtering algorithm and the CSF filtering algorithm.(4)Through the cross-section comparison analysis of the gully cross-section,the detail accuracy of the DEM cross-section of the gully of different scales after the point cloud filtering treatment of different filtering algorithms,and the comparison map of each cross-section is basically the same in terms of morphology,elevation,terrain undulation trend,etc.,indicating that the three filtering algorithms have reliability in the filtering processing of gully point cloud data and the high accuracy of establishing DEM.The polygon area formed by the intersection of each cross-sectional line was used to measure the error difference characteristics of different DEM,and the polygon area was statistically analyzed.The results show that the difference in cross-section in the gully is smaller than that in the trench head and the groove mouth cross-section,and for the small gully,the intersection area of CSF filter and TIN filter cross-section line is the smallest,and the intersection area of morphological filtering and CSF filter cross-section line is the largest.For medium and large gully,the intersection area of TIN filter and morphological filter cross-section line was the smallest,and the intersection area of morphological filter and CSF filter cross-section line was the largest,and the verification results were consistent with the comparison results of morphological parameters.(5)DEM error analysis of gully of different scales was established after treatment by different filtering methods,and the DEM accuracy established by CSF filtering method was used for small gully areas,and the DEM accuracy established after treatment by TIN filtering method was used in medium and large gully areas.