Method and device for extracting large-scale silted dam based on data engineering and storage medium

Abstract

The present application relates to a wide range of silt dam extraction method, device and storage medium based on data engineering, applied to the technical field of soil and water conservation, including: through the acquisition of open data resources including GoogleMap, 12m DEM, the cost of obtaining high-precision DEM is saved, the visible light remote sensing image is divided into a plurality of specific resolution images with geographic coordinate information, a plurality of specific resolution images with geographic coordinate information are respectively input into the pre-built silt dam detection model, the geographic coordinate information of the preselected silt dam distribution is obtained, the pre-built silt dam detection model can update the image regularly, and the silt dam extraction is automatically carried out by replacing the region, the manpower and material resources cost of field investigation and remote sensing interpretation is saved, the present application detects all suspected silt dam distribution points through the image with higher resolution first, then removes the false detection point information by using hydrological analysis and deep learning algorithm, the detection range is complete, and it is suitable for various silt dams, and the detection precision is high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water and soil conservation, in particular to a large-range silt dam extraction method and device based on data engineering and a storage medium. BACKGROUND

[0002] Silt dam is an effective water and soil conservation engineering measure in the Loess Plateau region, which plays an important role in retaining mud and protecting soil, reducing mud into the Yellow River, flood control and disaster reduction, silt land reclamation, consolidating the returning farmland to forest (grass), ensuring ecological safety, promoting grain production and rational use of water resources, and stable development of economy and society. It is one of the key measures for comprehensive control of water and soil loss in the Loess Plateau, and one of the important measures that cannot be ignored in the construction of beautiful countryside and rural strategic revitalization. By 2020, a total of 163,000 silt dams have been built in seven provinces of the Loess Plateau. With the long-term silt accumulation of silt dams, lack of management, and frequent extreme precipitation events in recent years, the risk of silt dam collapse is getting higher and higher, which seriously threatens the safety of life and property of downstream residents. The development of silt dam ecological benefit evaluation and silt dam risk susceptibility evaluation requires basic data such as silt dam distribution and silt area.

[0003] In the prior art, the existing silt dam basic data is mainly obtained through field investigation, high-precision DEM data extraction, and high-resolution remote sensing image interpretation. Field investigation requires a large amount of manpower and material resources, and is low in efficiency. High-precision DEM data extraction requires high-precision DEM (resolution greater than 5 meters), which is generally obtained by unmanned aerial vehicle aerial photography or high-precision commercial satellite. The data cost is high, and it is difficult to extract on a large scale. The silt dam extraction method based on 30-meter DEM and high-precision remote sensing image needs to first determine the preselected position of the silt dam that may exist through DEM analysis, and then extract the spatial information of the silt dam through remote sensing image analysis. This process needs to manually determine the position of the silt dam, and the efficiency is relatively low. The silt dam extraction based on neural network and high-precision DEM (resolution 5 meters) needs to extract flow grids with a window size of 25*25 from high-resolution DEM data. The extracted silt dam is within a range of 125 meters*125 meters, and only large silt dams can reach this scale, which cannot identify small dam systems. SUMMARY

[0004] Therefore, the present application aims to provide a large-range silt dam extraction method and device based on data engineering to solve the problems in the prior art that it is difficult to obtain data on a large scale, the real-time performance of the data is poor, the dependence on the precision of DEM data is high, manual field investigation and remote sensing interpretation are required, the labor cost is high, and there are certain requirements for the size of the silt dam.

[0005] According to a first aspect of an embodiment of the present application, a large-range silt dam extraction method based on data engineering is provided, comprising:

[0006] acquire visible light remote sensing images within a certain area range and DEM data of a certain resolution through an open data source;

[0007] divide the visible light remote sensing images into a plurality of images with geographic coordinate information of a certain resolution;

[0008] input the images with geographic coordinate information of a certain resolution into a pre-built silt dam detection model respectively to obtain geographic coordinate information of a preselected silt dam distribution;

[0009] process the DEM data of a certain resolution to obtain river network grading data;

[0010] establish a buffer zone of each level of river network according to the river network grading data;

[0011] perform a spatial intersection between the river network buffer zone data of each level and the geographic coordinate information of the preselected silt dam distribution to eliminate first false detection data in the preselected silt dam distribution that is not in a river channel;

[0012] process the river network grading data to obtain catchment area vector data;

[0013] extract feature information in the catchment area vector data and input the feature information into a pre-built catchment area construction silt dam suitability model to obtain a catchment area suitable for construction of a silt dam;

[0014] perform a spatial intersection between the preselected silt dam distribution data after elimination of the first false detection data and the catchment area data suitable for construction of a silt dam to eliminate second false detection data in the preselected silt dam distribution data and obtain a silt dam extraction result.

[0015] Preferably,

[0016] the building of the silt dam detection model comprises:

[0017] constructing a silt dam sample set based on visible light remote sensing images, each sample in the sample set comprising a 512*512 picture and an xml label file, the sample set containing four types of samples, i.e., two-side siltation type, single-side water body type, single-side siltation type and two-side mixed type, the four types of samples being uniformly distributed in the sample set;

[0018] building a silt dam detection model architecture, the silt dam detection model adopting a yolo v7 target detector, the silt dam detection model adopting an Adam optimizer, and the silt dam detection model reducing a learning rate through a cosine function and enabling a weighted image strategy;

[0019] training the silt dam detection model by taking the silt dam sample set as input to obtain the pre-built silt dam detection model.

[0020] Preferably,

[0021] The construction of the silt dam suitability model of the catchment area comprises:

[0022] Extracting feature information of each catchment area and whether there is a silt dam construction feature information sample set;

[0023] The silt dam suitability model of the catchment area is built, the silt dam suitability model of the catchment area adopts a support vector machine algorithm, a SIGMOD method is used to convert a binary classification value into a probability value, cross-validation and grid search are used for parameter adjustment, and the optimal penalty coefficient and support vector number are obtained;

[0024] The feature information sample set is taken as input, and the silt dam suitability model of the catchment area is trained to obtain the pre-built silt dam suitability model of the catchment area.

[0025] Preferably,

[0026] The DEM data of the specific resolution is processed to obtain river network classification data, which comprises:

[0027] The DEM data of the specific resolution is processed to obtain river network classification data, which comprises:

[0028] Preferably,

[0029] The river network classification data is processed to obtain catchment area vector data, which comprises:

[0030] The river network classification data is processed to obtain catchment area vector data, which comprises:

[0031] Preferably,

[0032] The feature information comprises: geomorphological information entropy, valley gradient, flow direction and flow rate of each catchment area.

[0033] The geomorphological information entropy is obtained by Strahler integration + information entropy.

[0034] According to the second aspect of the embodiment of the present application, a large-scale silt dam extraction device based on data engineering is provided, comprising:

[0035] The data acquisition module is used for acquiring visible light remote sensing images and DEM data of a specific resolution in a certain area range through an open data source.

[0036] The first preprocessing module is used for dividing the visible light remote sensing images into a plurality of specific resolution images with geographic coordinate information.

[0037] Pre-selected silt-retaining dam distribution acquisition module: Input multiple images with geographic coordinate information at specific resolutions into a pre-built silt-retaining dam detection model to obtain the geographic coordinate information of the pre-selected silt-retaining dam distribution;

[0038] The second preprocessing module is used to process DEM data at a specific resolution to obtain river network classification data.

[0039] Buffer establishment module: used to establish buffers for each level of the river network based on the river network classification data;

[0040] First error removal module: used to spatially intersect the river network buffer data at each level and the geographic coordinate information of the pre-selected silt-retaining dam distribution, and remove the first error detection data that is not in the river channel in the pre-selected silt-retaining dam distribution;

[0041] The third preprocessing module is used to process the river network classification data to obtain the catchment area vector data;

[0042] The catchment area acquisition module is used to extract feature information from the vector data of the catchment area, input the feature information into the pre-built catchment area silt-retention dam construction suitability model, and obtain the catchment area suitable for silt-retention dam construction.

[0043] The second error removal module is used to spatially intersect the pre-selected silt-retaining dam distribution data after removing the first error detection data with the catchment area data suitable for silt-retaining dam construction, remove the second error detection data in the pre-selected silt-retaining dam distribution data, and obtain the silt-retaining dam extraction results.

[0044] According to a third aspect of the present invention, a storage medium is provided, the storage medium storing a computer program, which, when executed by a host controller, implements the steps of the above-described method.

[0045] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0046] The application saves the cost of obtaining high-precision DEM by obtaining open data resources including GoogleMap and 12-meter DEM, can extract silt dams on a large scale and regularly, divides visible light remote sensing images into a plurality of specific resolution images with geographic coordinate information, inputs the plurality of specific resolution images with geographic coordinate information into a pre-built silt dam detection model respectively, obtains geographic coordinate information of pre-selected silt dam distribution, and can regularly update images through the pre-built silt dam detection model, replace the region to automatically extract silt dams, save the manpower and material resources cost of field investigation and remote sensing interpretation, and provide dynamic data support for soil and water conservation benefit evaluation and silt dam risk evaluation. The scheme first detects all suspected silt dam distribution points through high-resolution images, and then removes false detection point information by using hydrological analysis and deep learning algorithm, has full detection range, is suitable for various silt dams of large, medium and small types, and has high detection precision.

[0047] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0048] The drawings incorporated into the specification and forming a part thereof show embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.

[0049] Figure 1 is a flowchart of a large-range silt dam extraction method based on data engineering according to an exemplary embodiment;

[0050] Figure 2 is a schematic diagram of pre-selected silt dam distribution in the river network buffer zone of the Jiayuangou watershed according to an exemplary embodiment;

[0051] Figure 3 is a schematic diagram of catchment area distribution of whether the Jiayuangou watershed is suitable for construction of silt dams according to another exemplary embodiment;

[0052] Figure 4 is a system schematic diagram of a large-range silt dam extraction device based on data engineering according to another exemplary embodiment;

[0053] In the drawings: 1-data acquisition module, 2-first preprocessing module, 3-pre-selected silt dam distribution acquisition module, 4-second preprocessing module, 5-buffer establishment module, 6-first error removal module, 7-third preprocessing module, 8-catchment area acquisition module, 9-second error removal module. DETAILED DESCRIPTION

[0054] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The description below refers to the accompanying drawings, which show, by way of example, specific embodiments with which the inventive concept can be practiced. The following detailed description is not intended to limit the scope of the invention, as claimed, but is intended to be illustrative thereof. Rather, the following detailed description is intended to describe devices and methods in accordance with some aspects of the invention as detailed in the appended claims.

[0055] Embodiment One

[0056] Figure 1 is a flowchart of a large-scale silt dam extraction method based on data engineering according to an exemplary embodiment, as shown in Figure 1 , the method comprises:

[0057] S1, obtaining visible light remote sensing images of a certain area range and DEM data of a certain resolution through open data sources;

[0058] S2, dividing the visible light remote sensing images into a plurality of images with geographic coordinate information of a certain resolution;

[0059] S3, inputting the images with geographic coordinate information of a certain resolution into a pre-built silt dam detection model respectively, to obtain geographic coordinate information of preselected silt dam distribution;

[0060] S4, processing the DEM data of a certain resolution to obtain river network grading data;

[0061] S5, establishing a buffer zone of each level of river network according to the river network grading data;

[0062] S6, performing a spatial intersection between the river network buffer zone data of each level and the geographic coordinate information of the preselected silt dam distribution, to eliminate first false detection data in the preselected silt dam distribution which is not in the river channel;

[0063] S7, processing the river network grading data to obtain catchment area vector data;

[0064] S8, extracting feature information in the catchment area vector data, inputting the feature information into a pre-built catchment area construction silt dam suitability model, and obtaining catchment areas suitable for construction of silt dams;

[0065] S9, performing a spatial intersection between the preselected silt dam distribution data after elimination of the first false detection data and the catchment area data suitable for construction of silt dams, to eliminate second false detection data in the preselected silt dam distribution data, and obtaining a silt dam extraction result;

[0066] It can be understood that the visible light remote sensing image in a certain regional range and the DEM data with a specific resolution are obtained through an open data source, the open data source includes Google Earth image, 12-meter DEM, etc., compared with the demand for high-precision DEM data in the prior art, the cost of obtaining high-precision DEM is saved, and the silt dam can be extracted in a large range and regularly. The data source can also be replaced with other high-resolution visible light remote sensing images such as Gaofen-2; then the visible light remote sensing image is segmented into a plurality of 4096*4096 images with geographic coordinate information, the 4096*4096 images with geographic coordinate information are input into a pre-built silt dam detection model, the confidence of the pre-built silt dam detection model is set to 0.5, the IoU threshold is set to 0.45, and the output is the distribution of the preselected silt dam in longitude, latitude, width information; then the 12-meter DEM data is processed to obtain river network grading data, and a buffer zone of each level of river network is established according to the river network grading data. In the present application, the 1-3 level buffer zone is set to 60, the 4-5 level buffer zone is set to 80, and the 6-7 level buffer zone is set to 100; the river network buffer zone data and the preselected silt dam distribution detection data are fused, the spatial intersection of point and surface elements is calculated, and the first false detection data in the silt dam distribution data that is not in the river channel is removed; the river network grading data is processed to obtain catchment area vector data, feature information in the catchment area vector data is extracted, the feature information is input into a pre-built catchment area construction silt dam suitability model, and a catchment area suitable for constructing a silt dam is obtained; the silt dam distribution detection data point and the catchment area suitable for constructing a silt dam are fused, the spatial intersection of point and surface elements is calculated, the second false detection data in the silt dam distribution data is removed, and the final result is obtained; through the pre-built silt dam detection model, the image can be updated regularly, the silt dam extraction can be automatically performed in the region, the cost of manpower and material resources for field investigation and remote sensing interpretation is saved, dynamic data support is provided for soil and water conservation benefit evaluation and silt dam risk evaluation, and the scheme has the advantages that all suspected silt dam distribution points are detected through the image with high resolution, then the false detection point information is removed by using hydrological analysis and deep learning algorithm, the detection range is full, the scheme is suitable for various silt dams of large, medium and small types, and the detection precision is high.

[0067] Preferably,

[0068] The silt dam detection model is built, including:

[0069] A silt dam sample set is constructed based on visible light remote sensing images, each sample in the sample set includes a 512*512 picture and an xml label file, the sample set includes four types of samples, i.e., two-side siltation type, single-side water type, single-side siltation type and two-side mixed type, and the four types of samples are uniformly distributed in the sample set;

[0070] The silt dam detection model is built, the silt dam detection model adopts a yolo v7 target detector, the silt dam detection model adopts an Adam optimizer, and the silt dam detection model reduces the learning rate through a cosine function and enables a weighted image strategy;

[0071] The silt dam detection model is trained with the silt dam sample set as input, and a pre-built silt dam detection model is obtained.

[0072] It can be understood that the silt dam sample set is constructed based on high-resolution visible light remote sensing images. Each sample includes a 512*512 picture and an xml label file, containing four types of samples of two-side siltation type, single-side water type, single-side siltation type, and two-side mixed type, and the samples are uniformly distributed. The model adopts yolo v7 (You Only Look Once), the optimizer selects Adam, the learning rate is reduced through a cosine function, and a weighted image strategy is enabled. The silt dam sample set is used as input to train the model, and the epochs in the training process is set to 10000 and the batch size is 16.

[0073] Preferably,

[0074] The silt dam suitability model for the catchment area construction includes:

[0075] Feature information of each catchment area and whether there is a silt dam construction feature information sample set are extracted.

[0076] The silt dam suitability model for the catchment area construction is built, the silt dam suitability model for the catchment area construction adopts a support vector machine algorithm, and a SIGMOD method is used to convert a binary classification value into a probability value. Cross-validation and grid search are used for parameter adjustment to obtain the optimal penalty coefficient and support vector number.

[0077] The silt dam suitability model for the catchment area construction is trained with the feature information sample set as input, and a pre-built silt dam suitability model for the catchment area construction is obtained.

[0078] It can be understood that the feature information of each catchment area and whether there is a silt dam are extracted to construct a sample set. The silt dam suitability model for the catchment area construction is built, the sample set is used as input to train the model, the model adopts a SVM (Support Vector Machine) algorithm, a SIGMOD method is used to convert a binary classification value into a probability value, cross-validation (cross-validation) and grid search (grid-search) are used for parameter adjustment, and the optimal penalty coefficient (c) and support vector number (gamma) are obtained.

[0079] Preferably,

[0080] The processing of the DEM data of the specific resolution to obtain the river network classification data comprises:

[0081] The river network classification data is extracted from the DEM data of the specific resolution through filling and digging, flow direction analysis, flow statistics, flow division, and river classification operation, wherein the flow direction analysis operation adopts a D8 algorithm;

[0082] It can be understood that the river network classification data is extracted from the DEM data through filling and digging, flow direction analysis, flow statistics, flow division, and river classification, the flow direction analysis adopts a D8 method, the threshold of the catchment area is 0.5 km2, and the D8 algorithm is a mathematical algorithm used to calculate the distance weight drop, assuming that the water flow in a single grid can only flow into 8 adjacent grids. It uses the steepest slope method to determine the direction of water flow, that is, on a 3x3 DEM grid, the distance weight drop between the center grid and each adjacent grid (i.e., the grid center point drop divided by the distance between the grid center points) is calculated, and the grid with the largest distance weight drop is taken as the outflow grid of the center grid. The principle of the steepest slope method is to assume that the ground is impermeable and the rainfall is uniform. Then the water flow on the watershed unit always flows to the lowest place. The calculation unit is taken as the center, and a window is formed by combining several adjacent units. The window is taken as the basic calculation element, and the final result is obtained by pushing the entire DEM.

[0083] Preferably,

[0084] The processing of the river network classification data to obtain the catchment area vector data comprises:

[0085] The river network classification data is processed through river linking, raster river network vectorization, drawing of inclined points, and raster watershed extraction to obtain the catchment area vector data.

[0086] It can be understood that the above-mentioned processing of the river network classification data through river linking, raster river network vectorization, drawing of inclined points, and raster watershed extraction to obtain the catchment area vector data is a common technical means for those skilled in the art, and will not be described in detail here.

[0087] Preferably,

[0088] The feature information comprises the geomorphic information entropy, the valley slope, the flow direction, and the flow of each catchment area.

[0089] The geomorphic information entropy is obtained through Strahler integration + information entropy.

[0090] It can be understood that the geomorphic information entropy represents the geomorphic development information of the catchment area, the valley slope represents the energy information of the terrain, and the flow direction and flow represent the catchment process and flow level. These parameters comprehensively represent the characteristics of the topography, energy, flow, etc. of the silt dam construction area; Strahler integral is an area-elevation analysis method calculated by A. N. Strahler, a geomorphologist in the United States, which is used to quantitatively analyze the method of geomorphic development stage.

[0091] In order to facilitate the understanding of the present application, the present application demonstrates the present application through the following specific cases:

[0092] The experimental area of the typical concentrated area of silt dam in Suide County, Shaanxi Province is Jiayuangou. Jiayuangou watershed is a first-order tributary on the left bank of the middle reaches of Wudinghe River, belongs to the first sub-region of loess hilly and gully region, the watershed area is 70.7km2, the main ditch is 18km long, there are 337 branch ditches above 200m, the average slope of the ditch is 1.2%, the gully density is 5.34km / km2, and the watershed elevation is between 820-1180m. As of the end of 2017, there were a total of 206 silt dams in the whole watershed, including 24 backbone dams, 35 medium-sized dams and 147 small-sized dams;

[0093] Collect 19-level Google Earth images (resolution 0.47 meters) and 12-meter DEM data in Suide County. The data is spliced and cut;

[0094] Based on high-resolution visible light remote sensing image, a silt dam sample set is constructed. Each sample includes a 512*512 picture and an xml label file, including four types of samples: two-sided silt type, single-sided water type, single-sided silt type, and two-sided mixed type. The samples are evenly distributed, and a total of 1000 samples are labeled; the model is trained by the sample set to obtain a pre-built silt dam detection model;

[0095] The high-resolution remote sensing image of Suide County is segmented into 4096*4096 images with geographic coordinate information, and the pre-built silt dam detection model is input to detect silt dams one by one. The confidence is set to 0.5, the IoU threshold is set to 0.45, and the output is the distribution of the pre-selected silt dam latitude, longitude, width information, a total of 881 pre-selected silt dams are detected;

[0096] Because terraces, bridges and other similar shapes to silt dams, when using image detection technology to identify silt dams, there will often be errors in identifying terraces and other images as silt dams, so the present application uses hydrological analysis and other supplementary means to eliminate the incorrectly identified images. First, silt dams must be distributed in the ditch, so the present application first extracts the river network classification data. Then, according to the width range of the silt dam, a buffer zone of the river network data is established. Remove the detection data not contained in the river network buffer zone. The specific process is as follows:

[0097] The collected DEM data is filled by DEM data, flow direction analysis, flow statistics, flow division, river network classification data is extracted by river classification. The flow direction analysis adopts the D8 method, and the threshold of the catchment area is 0.5 km2. According to different levels of the river network, the buffer zone of each level of the river network is established. The buffer zone of 1-3 levels is set to 60, the buffer zone of 4-5 levels is set to 80, and the buffer zone of 6-7 levels is set to 100. The output is a river network buffer zone vector data;

[0098] The river network buffer zone data and the preselected silt dam distribution data are used to calculate the spatial intersection of point and surface elements, and the incorrect detection data outside the river channel is removed, and the detection result is as shown in Figure 2

[0099] The silt dam and the slope ridge, the relatively flat channel slope are also similar on the remote sensing image, so there is a situation that the slope ridge, the channel slope and the like are identified as the silt dam in the river network buffer zone. The application further removes the error-identified data such as the channel slope by analyzing the catchment area characteristics suitable for constructing the silt dam.

[0100] The specific process is as follows:

[0101] The extracted river network classification data is subjected to the steps of river linking, raster river network vectorization, drawing of inclined points and raster watershed extraction, and vector data of the catchment area is extracted. The threshold of the catchment area is 0.5 km2, and the Jiuyuangou watershed has 496 catchment areas;

[0102] The geomorphologic information entropy, the valley gradient, the flow direction, the flow and whether there is a silt dam of each catchment area in the whole Jiuyuangou watershed are extracted, and a sample set is constructed. The geomorphologic information entropy represents the geomorphologic development information of the catchment area, the valley gradient represents the energy information of the terrain, the flow direction and the flow represent the catchment process and the flow level, and these parameters comprehensively represent the geomorphology, energy, flow and the like of the silt dam construction area. The geomorphologic information entropy is calculated by using Strahler integration + information entropy. The valley gradient is calculated by using the Johnstone-Kross method. 120 samples are randomly selected, including 83 positive samples and 37 negative samples;

[0103] Based on the above sample set, a SVM (support vector machine) algorithm is used to construct and train a catchment area construction silt dam suitability model. The characteristic values are normalized, cross-validation (cross-validation) and grid search (grid-search) are used for parameter adjustment, the optimal penalty coefficient (c) and support vector number (gamma) are obtained, and the SIGMOD method is used to convert the binary classification value into a probability value;

[0104] ​The geomorphic information entropy, gully gradient, flow direction and flow information of 496 catchments in the whole Jiyeungou watershed are input into the model trained in the previous step, and the probability of the suitability of constructing the check dam in each catchment is output. The catchment with a suitability probability threshold greater than 65% is determined as a catchment with a check dam, and the catchment vector data suitable for constructing the check dam is output, and 352 catchments are output as shown in the accompanying Figure 3 ;

[0105] The output preselected check dam distribution data and the catchment area element suitable for constructing the check dam are intersected in space, and the error detection data such as river channel side slope is removed to obtain the final result, and 226 check dams are identified. The output result is compared and verified by historical data and manual interpretation, and it is proved that the accuracy of the method can reach 91.2%, and the Kappa coefficient reaches 0.862, which can be applied to large-scale check dam identification.

[0106] Example two

[0107] Figure 4 is a system schematic diagram of a large-scale check dam extraction device based on data engineering according to another exemplary embodiment, comprising:

[0108] The data acquisition module 1 is used for acquiring visible light remote sensing images and DEM data of a certain area range through an open data source;

[0109] The first preprocessing module 2 is used for segmenting the visible light remote sensing images into a plurality of specific resolution images with geographic coordinate information;

[0110] The preselected check dam distribution acquisition module 3 is used for inputting the plurality of specific resolution images with geographic coordinate information into the pre-built check dam detection model respectively, and obtaining the geographic coordinate information of the preselected check dam distribution;

[0111] The second preprocessing module 4 is used for processing the specific resolution DEM data to obtain river network classification data;

[0112] The buffer establishment module 5 is used for establishing the buffer zone of each level of river network according to the river network classification data;

[0113] The first error removal module 6 is used for performing spatial intersection between the river network buffer zone data of each level and the geographic coordinate information of the preselected check dam distribution, and removing the first error detection data in the preselected check dam distribution which is not in the river channel;

[0114] The third preprocessing module 7 is used for processing the river network classification data to obtain catchment vector data;

[0115] The catchment area acquisition module 8 is used for extracting feature information in the catchment area vector data, inputting the feature information into a pre-built catchment area construction silt dam suitability model, and acquiring a catchment area suitable for construction of a silt dam.

[0116] The second error elimination module 9 is used for performing a spatial intersection between the preselected silt dam distribution data after elimination of the first error detection data and the catchment area data suitable for construction of a silt dam, eliminating second error detection data in the preselected silt dam distribution data, and obtaining a silt dam extraction result.

[0117] It can be understood that the application acquires visible light remote sensing images in a certain regional range and DEM data of a specific resolution through the data acquisition module 1 by using an open data source; the visible light remote sensing images are segmented into a plurality of specific resolution images with geographic coordinate information through the first preprocessing module 2; the plurality of specific resolution images with geographic coordinate information are respectively input into a pre-built silt dam detection model through the preselected silt dam distribution acquisition module 3, and geographic coordinate information of the preselected silt dam distribution is obtained; river network classification data is obtained by processing the DEM data of the specific resolution through the second preprocessing module 4; the buffer zone of each level of river network is established according to the river network classification data through the buffer establishment module 5; the geographic coordinate information of the preselected silt dam distribution is subjected to a spatial intersection with the river network buffer zone data of each level through the first error elimination module 6, and the first error detection data outside the river channel in the preselected silt dam distribution is eliminated; the river network classification data is processed through the third preprocessing module 7 to obtain catchment area vector data; feature information in the catchment area vector data is extracted through the catchment area acquisition module 8, the feature information is input into a pre-built catchment area construction silt dam suitability model, and a catchment area suitable for construction of a silt dam is acquired; the preselected silt dam distribution data after elimination of the first error detection data is subjected to a spatial intersection with the catchment area data suitable for construction of a silt dam through the second error elimination module 9, the second error detection data in the preselected silt dam distribution data is eliminated, and a silt dam extraction result is obtained. The pre-built silt dam detection model can periodically update the image, and the silt dam extraction is automatically performed in the replacement area, thereby saving the manpower and material resources cost of field investigation and remote sensing interpretation, providing dynamic data support for water and soil conservation benefit evaluation and silt dam risk evaluation, and the scheme detects all suspected silt dam distribution points through high-resolution images, then eliminates the error detection point information by using a hydrological analysis and a deep learning algorithm, has a full detection range, is suitable for various silt dams of large, medium and small types, and has high detection precision.

[0118] Embodiment three

[0119] The embodiment provides a storage medium, the storage medium stores a computer program, and the computer program is executed by a host computer to realize each step in the above method.

[0120] It can be understood that the storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0121] It can be understood that the same or similar parts in the above embodiments can be mutually referred to, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0122] It should be noted that, in the description of the present application, the terms "first", "second", etc. are only for the purpose of description, and cannot be understood as indicating or implying relative importance. In addition, in the description of the present application, unless otherwise specified, the meaning of "a plurality of" is at least two.

[0123] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for performing specific logic functions or steps in the process, and that the various embodiments of the present application include additional implementations in which the functions described are performed in a different order, including substantially simultaneously, or in reverse order, and that the described implementations are merely examples.

[0124] It should be understood that the parts of the present application can be realized by hardware, software, firmware or their combination. In the above embodiments, a plurality of steps or methods can be realized by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if realized by hardware, and as in another embodiment, it can be realized by any one or their combination of the following technologies known in the art: discrete logic circuit with logic gate circuit for implementing logic function on data signal, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA), etc.

[0125] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the above-mentioned embodiment method can be completed by a program instructing the relevant hardware, and the program can be stored in a computer readable storage medium, and the program includes one or a combination of the steps of the method embodiment when executed.

[0126] In addition, each functional unit in each embodiment of the present application can be integrated in one processing module, or each unit can exist physically independently, or two or more units can be integrated in one module. The above integrated module can be realized in the form of hardware or in the form of a software functional module. When the integrated module is realized in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0127] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0128] In the description of the present specification, the description referring to the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Also, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in an appropriate manner.

[0129] Although the embodiments of the present application have been shown and described above, it is understood that the above-described embodiments are exemplary, and are not to be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-described embodiments within the scope of the present application.

Claims

1. A method for large-scale extraction of levees based on data engineering, characterized in that, The method comprises the following steps: acquiring visible light remote sensing images within a certain area range and DEM data with a certain resolution; segmenting the visible light remote sensing images into a plurality of images with geographic coordinate information and a certain resolution; inputting the images with geographic coordinate information and a certain resolution into a pre-built silt dam detection model respectively to obtain geographic coordinate information of preselected silt dam distribution; processing the DEM data with a certain resolution to obtain river network grading data; establishing a buffer zone of each level of river network according to the river network grading data; performing spatial intersection between the river network buffer zone data of each level and the geographic coordinate information of the preselected silt dam distribution to remove first false detection data in the preselected silt dam distribution which is not in a river channel; processing the river network grading data to obtain catchment area vector data; extracting feature information in the catchment area vector data and inputting the feature information into a pre-built catchment area construction silt dam suitability model to obtain catchment areas suitable for silt dam construction; performing spatial intersection between the preselected silt dam distribution data after removing the first false detection data and the catchment area data suitable for silt dam construction to remove second false detection data in the preselected silt dam distribution data and obtain silt dam extraction results.

2. The method of claim 1, wherein the silt dam detection model is built by: constructing a silt dam sample set based on visible light remote sensing images, each sample in the sample set comprising a 512*512 picture and an xml label file, the sample set containing four types of samples, i.e., two-side siltation type, single-side water body type, single-side siltation type and two-side mixed type, the four types of samples being uniformly distributed in the sample set; building a silt dam detection model architecture, the silt dam detection model using a yolo v7 target detector, the silt dam detection model using an Adam optimizer, and the silt dam detection model reducing the learning rate by using a cosine function and enabling a weighted image strategy; training the silt dam detection model by taking the silt dam sample set as input to obtain the pre-built silt dam detection model.

3. The method of claim 2, wherein the catchment area construction silt dam suitability model is built by: extracting feature information of each catchment area and whether there is a silt dam construction feature information sample set; building a catchment area construction silt dam suitability model architecture, the catchment area construction silt dam suitability model using a support vector machine algorithm, converting a binary classification value into a probability value by using a SIGMOD method, and using cross-validation and grid search to adjust parameters to obtain an optimal penalty coefficient and support vector number; training the catchment area construction silt dam suitability model by taking the feature information sample set as input to obtain the pre-built catchment area construction silt dam suitability model.

4. The method of claim 3, wherein the processing of the DEM data with a certain resolution to obtain the river network grading data comprises: extracting the river network grading data from the DEM data with a certain resolution by filling and digging, flow direction analysis, flow statistics, flow division and river grading operation, the flow direction analysis operation using a D8 algorithm.

5. The method of claim 4, wherein ​ ​ ​ The processing of the river network hierarchical data to obtain the catchment area vector data comprises: The river network hierarchical data is subjected to river linking, raster river network vectorization, drawing of tilt points, and raster catchment extraction operations to obtain the catchment area vector data.

6. The method of any one of claims 1 or 3, wherein the feature information comprises landform information entropy, valley gradient, flow direction, and flow rate of each catchment area. The landform information entropy is obtained by Strahler integration + information entropy. The device comprises:

7. A large scale embanked land extraction apparatus based on data engineering, characterized by, a data acquisition module configured to acquire visible light remote sensing images and DEM data of a certain area range and at a certain resolution; a first preprocessing module configured to divide the visible light remote sensing images into a plurality of images with geographic coordinate information at a certain resolution; a preselected silt dam distribution acquisition module configured to input the plurality of images with geographic coordinate information at a certain resolution into a pre-built silt dam detection model to obtain geographic coordinate information of a preselected silt dam distribution; a second preprocessing module configured to process the DEM data at a certain resolution to obtain river network hierarchical data; a buffer establishment module configured to establish a buffer zone of each level of river network according to the river network hierarchical data; a first error elimination module configured to perform spatial intersection between the buffer zone data of each level of river network and the geographic coordinate information of the preselected silt dam distribution to eliminate first error detection data in the preselected silt dam distribution that is not in a river channel; a third preprocessing module configured to process the river network hierarchical data to obtain catchment area vector data; a catchment area acquisition module configured to extract feature information from the catchment area vector data, input the feature information into a pre-built catchment area suitability model for construction of silt dams, and obtain catchment areas suitable for construction of silt dams; a second error elimination module configured to perform spatial intersection between the preselected silt dam distribution data after elimination of the first error detection data and the catchment area data suitable for construction of silt dams to eliminate second error detection data in the preselected silt dam distribution data and obtain silt dam extraction results. The storage medium stores a computer program, and when the computer program is executed by the host controller, each step in the large-range silt dam extraction method based on data engineering according to any one of claims 1-6 is implemented.

8. A storage medium, characterized by ​

Images