Font Size: a A A

Ephemeral Gully Recognition And Temporal-spatial Distribution Based On Deep Learning And Remote Sensing Images ——A Case Study Of Zhoutungou Watershed

Posted on:2022-12-11Degree:MasterType:Thesis
Country:ChinaCandidate:B Y LiuFull Text:PDF
GTID:2518306776490284Subject:Automation Technology
Abstract/Summary:PDF Full Text Request
Ephemeral gully erosion is a linear erosion mode on steep slope cultivated land,which plays a connecting role in the slope gully erosion system of the Loess Plateau.At present,the research on ephemeral gully mostly focuses on indoor simulation test and field slope unit positioning test.With the continuous development of remote sensing technology,remote sensing images have been gradually applied to shallow ditch research.However,the methods of shallow ditch spatial position recognition based on remote sensing images are mostly visual interpretation and machine learning.There are few reports on ephemeral gully recognition based on deep learning.Taking high-resolution remote sensing images as the data source,this study uses the deep learning image semantic segmentation model for ephemeral gully recognition.On this basis,ephemeral gully eigenvalues such as ephemeral gully number,gully length,gully width,distribution spacing,critical slope and critical catchment area are extracted,and the temporal and spatial distribution law of ephemeral gully eigenvalues is analyzed,Explore the influence of terrain factors(slope,aspect and altitude)on the distribution of ephemeral gully eigenvalues,clarify the temporal and spatial distribution law and influencing factors of ephemeral gully in zhoutungou watershed in hilly and gully area of the Loess Plateau,provide a concise and efficient method for ephemeral gully identification in hilly and gully area of the Loess Plateau,and provide reference and guidance for slope ephemeral gully treatment in hilly and gully area of the Loess Plateau.The main conclusions of this study are as follows:(1)The feasibility of depth learning in ephemeral gully recognition in Hilly and gully areas of the Loess Plateau is verified.The six evaluation indexes of Accuracy,Precision,Recall,F1 value,ROC curve and AUC value are used to comprehensively quantify the advantages of U-Net,SegNet and R2U-Net image semantic segmentation models in identifying ephemeral gull IES.The image semantic segmentation models are suitable for ephemeral gully recognition,among which segnet model performs best and has good integrity and accuracy in ephemeral gully recognition,It can be widely applied to ephemeral gully recognition and ephemeral gully eigenvalue extraction in hilly and gully areas of the Loess Plateau.(2)The temporal and spatial distribution law of each eigenvalue of ephemeral gully is clarified.From 1999 to 2009,the number of ephemeral gullies decreased sharply from 4495to 1153,but the number of ephemeral gullies rebounded slightly in 2009?2021,which may be mainly due to the frequent occurrence of extreme rainfall events and the insufficient consolidation of the results of returning farmland to forest.From 2009 to 2021,the spatial distribution of ephemeral gullies is mainly concentrated in the southern region of the basin,which is mostly hilly,complex landform and large slope gradient;The average growth rate of the number of ephemeral gullies is 97/a.The length of ephemeral gullies is mainly distributed between 25?40 m,and the average growth rate is 1.66 m/a.The width of ephemeral gullies is mainly distributed between 0.5?1.5 m,and the average growth rate is 0.04 m/a.The distribution spacing between ephemeral gullies is mianly between 14?35 m,the critical catchment area of ephemeral gullies is mainly between 75?225 m~2 and the average growth rate is 1.8 m~2/a.The critical slope of ephemeral gullies is mainly concentrated between 23°?39°and the average growth rate is 1.34°/a.The characteristic values of ephemeral gullies show a decreasing trend year by year,indicating that the ephemeral gullies in small watersheds in hilly and gully areas of the Loess Plateau have been effectively controlled.(3)The influence of slope gradient and slope aspect on the distribution of ephemeral gully eigenvalues is explored.Ephemeral gullies are concentrated on the slope above 16°,especially on the slope above 26°,and the length of ephemeral gullies will gradually decrease with the increase of slope;The gully width of 0.5m?1.5m on each slope grade is evenly distributed,and there is no ephemeral gully with a gully width of more than 2m on the slopes of flat gentle slope I and medium slope II;The mean value of ephemeral gully critical catchment area at different slope grades has little difference.The number of ephemeral gullies in different slope directions is semi sunny slope>semi shady slope>shady slope>sunny slope;The length of ephemeral gully is sunny slope>semi sunny slope>semi shady slope>shady slope in different slope directions;There is no significant difference in the distribution of critical slope and critical catchment area of ephemeral gully in each slope direction.(4)The influence of altitude on the distribution of ephemeral gully eigenvalues was explored.ephemeral gullies are mainly distributed at an altitude of more than 1250 m,and the proportion of 0.5?1.5 m gully width,critical slope and critical catchment area gradually decrease with the increase of altitude.
Keywords/Search Tags:Deep learning, Hilly and gully region of the Loess Plateau, Ephemeral gully recognition, Ephemeral gully eigenvalues, Terrain factor
PDF Full Text Request
Related items