Rock-soil mass data analysis method based on multi-source satellite data and geoscience data

Abstract

The invention belongs to the technical field of geological exploration, and provides a rock-soil mass data analysis method based on multi-source satellite data and geoscience data. Comprising the steps of acquisition of data to be analyzed, construction of a high-resolution multispectral overground three-dimensional model, construction of an original underground model, optimization of the original underground model, construction of a regional total-space model, region segmentation and data fusion, and calculation of a rock-soil mass data analysis model. According to the method, various collected data are fused, the drilling data and the basic geological data are effectively utilized, and the estimation precision of the rock-soil mass data is improved; by performing space filling processing on the original underground model, the data density of the underground model is improved, finer data is provided for model estimation, and interference of noise on a result caused by low data volume is avoided; through fusion of the upper model, the lower model and the region segmentation mask, the region feature analysis effect is improved, the boundary extraction speed and accuracy are improved, and the estimation precision of conventional rock-soil mass data is improved.

Description

Technical Field

[0001] The present invention relates to the field of geological exploration technology, and in particular to a rock and soil body data analysis method based on multi-source satellite data and geological information. Background Art

[0002] Currently, geotechnical data in engineering geology is primarily acquired through field surveys and testing, resulting in limited data collection for large areas. Methods for rapidly acquiring large-scale geotechnical data are urgently needed for intelligent engineering site selection, roadless off-road navigation, and emergency rescue command. Researchers published a study on the application of data fusion technology in railway geological remote sensing interpretation, highlighting the use of multi-source data fusion for geological remote sensing interpretation. This method is primarily used to interpret undesirable geological volumes in engineering geology.

[0003] However, the original technology mainly relies on field surveys, and the method for quickly obtaining rock and soil related data over a large area is mainly based on remote sensing interpretation and judgment. It does not effectively utilize geological data, drilling data, etc., and the main target is unfavorable geological bodies. There is no basic data refinement method for rock and soil. The existing methods lack the fusion analysis process of above-ground and underground models, and often analyze the above-ground model alone, resulting in a lack of credibility of the analysis results. In addition, after constructing the regional model, the existing methods only obtain conclusions through subjective analysis by researchers, resulting in low reliability and low efficiency. Summary of the Invention

[0004] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide a rock and soil data analysis method based on multi-source satellite data and geological information, so as to solve the problems of the existing technology such as lack of drilling data and basic geological data utilization, low data density, lack of physical model pre-analysis and high model calculation amount.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A geotechnical data analysis method based on multi-source satellite data and geological data includes:

[0007] Collect high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, DEM data, borehole data and basic geological data of the target analysis area; the basic geological data includes: geological plan map, geological cross-section map, stratigraphic and lithological information, and geological structure information;

[0008] fusing the high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data, and the DEM data to obtain a high-resolution multispectral ground three-dimensional model;

[0009] fusing the drilling data and the basic geological data to obtain an original underground model;

[0010] spatially filling the original subsurface model to obtain a three-dimensional model of a subsurface geological body;

[0011] registering and fusing the high-resolution multispectral three-dimensional model of the ground and the three-dimensional model of the subsurface geological body to obtain a regional full-space model;

[0012] segmenting the regional full-space model to obtain a three-dimensional segmentation mask, and fusing the regional full-space model and the three-dimensional segmentation mask to obtain to-be-analyzed data;

[0013] inputting the to-be-analyzed data into a pre-trained rock-soil data analysis model for analysis to obtain a target rock-soil data analysis result.

[0014] Preferably, high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, hyperspectral satellite remote sensing image data, DEM data, drilling data, and basic geological data of the target analysis region are collected, including:

[0015] defining the boundary of the target analysis region through satellite images and topographic mapping to obtain a regional range;

[0016] dividing the regional range into regular longitudinal and transverse grids to obtain a regional grid;

[0017] labeling the grid intersection points of the regional grid in the target analysis region using GPS, and determining a line segment composed of the grid intersection points located in the same row or the same column as a longitudinal or transverse edge.

[0018] Preferably, high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, hyperspectral satellite remote sensing image data, DEM data, drilling data, and basic geological data of the target analysis region are collected, including:

[0019] conducting stratigraphic outcrop observation, structure identification, overburden investigation, hydrogeological phenomenon recording, and geophysical data collection on the region where each longitudinal or transverse edge is located to obtain the geological profile;

[0020] drilling at each grid intersection point to obtain collected samples;

[0021] conducting core and soil sample analysis on the collected samples to obtain basic physical parameters;

[0022] conducting geological data collection on each grid intersection point using geophysical logging technology and in-situ testing technology to obtain stratigraphic parameters and mechanical and hydrological parameters;

[0023] Integrate the basic physical parameters, the formation parameters, the mechanical and hydrological parameters to obtain the drilling data.

[0024] Preferably, the high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data and the DEM data are fused to obtain a high-resolution multispectral terrestrial three-dimensional model, comprising:

[0025] The high-resolution panchromatic satellite remote sensing image data and the multispectral satellite remote sensing image data are respectively subjected to radiation correction, geometric correction and image registration.

[0026] The correlation of each multispectral band of the multispectral satellite remote sensing image data and the high-resolution panchromatic satellite remote sensing image data is calculated to obtain a correlation coefficient; the calculation formula of the correlation coefficient is: wherein, w i is the i-th correlation coefficient; Corr(·) represents a correlation coefficient calculation function; P is the high-resolution panchromatic satellite remote sensing image data; MS i is the i-th multispectral band;

[0027] The multispectral bands are weightedly summed with the correlation coefficient as the weight to obtain a simulated panchromatic image.

[0028] The high-resolution panchromatic satellite remote sensing image data and the simulated panchromatic image are subjected to difference calculation to obtain original difference data, and the original difference data is subjected to three times standard deviation threshold filtering and interpolation processing to obtain optimized difference data.

[0029] The multispectral bands and the projection components of the multispectral bands on the simulated panchromatic image are subjected to difference calculation to obtain high-frequency detail components.

[0030] The optimized difference data is proportionally replaced by the high-frequency detail components to obtain a high-resolution multispectral image; the expression of the high-resolution multispectral image is: MS′ i = Proj(MS i , P ′ ) + k·D; wherein, MS′ i is the i-th optimized band data of the high-resolution multispectral image; Proj(·) is a projection formula; P' is the simulated panchromatic image; k is a scaling coefficient; D is the optimized difference data.

[0031] Preferably, the high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data and the DEM data are fused to obtain a high-resolution multispectral terrestrial three-dimensional model, further comprising:

[0032] converting the DEM data into a digital terrain model using a GIS tool;

[0033] mapping the high-resolution multispectral image onto the surface of the digital terrain model to obtain a high-resolution multispectral above-ground three-dimensional model.

[0034] Preferably, the borehole data and the basic geological data are fused to obtain an original underground model, comprising:

[0035] constructing a grid model according to the regional grid and the depth data of the geological profile, and setting the resolution of the grid model;

[0036] taking the upper left corner of the grid model as an origin, taking the two edges where the origin is located as an X-axis and a Y-axis, and taking the vertical direction depth as a Z-axis to obtain a space coordinate system;

[0037] injecting the borehole data and the basic geological data into the grid model according to a coordinate scaling ratio based on the space coordinate system to obtain the original underground model.

[0038] Preferably, the original underground model is subjected to a space filling process to obtain an underground geological body three-dimensional model, comprising:

[0039] dividing the original underground model into a plurality of cuboids; the attribute data of the vertical edges of the cuboids correspond to the borehole data; and the attribute data of the vertical sides of the cuboids correspond to the basic geological data;

[0040] mapping the cuboids into unit cubes;

[0041] fitting the borehole data according to the unit cubes to obtain an edge line segment function;

[0042] fitting the basic geological data according to the unit cubes to obtain a boundary surface function;

[0043] constructing a control equation according to the unit cubes and the boundary surface function; the expression of the control equation is: F(u=0,v,w)=F front (v,w);F(u=1,v,w)=F back (v,w);F(u,v=0,w)=F left (u,w);F(u,v=1,w)=F right (u,w);F is an attribute function; u, v, and w are respectively a horizontal axis, a vertical axis, and a vertical axis of the unit cube; F front (v,w)、F back (v,w)、F left (u,w)、Fright (u, w) correspond to the boundary surface functions of the front side, back side, left side and right side of the unit cube, respectively;

[0044] Discretize and iteratively solve the control equation by using finite difference method from the edge function, to obtain a primary solution result;

[0045] Perform tensor product spline fitting and inverse mapping on the primary solution result to obtain an optimized solution result;

[0046] When the local first derivative difference value of the optimized solution result on both sides of the vertical direction of the side of the cuboid is greater than a preset abnormal difference threshold, perform secondary solving on the two cuboids corresponding to the local first derivative difference value.

[0047] Preferably, the high-resolution multispectral terrestrial three-dimensional model and the underground geological body three-dimensional model are registered and fused to obtain a regional full-space model, including:

[0048] Convert the spatial coordinate system of the underground geological body three-dimensional model into a global coordinate system of the high-resolution multispectral terrestrial three-dimensional model by using a coordinate transformation algorithm;

[0049] Extract the positioning data of the grid intersection points of the underground geological body three-dimensional model;

[0050] Match the position data of the grid intersection points in the underground geological body three-dimensional model and the position data of the positioning data in the high-resolution multispectral terrestrial three-dimensional model to obtain a mapping pair;

[0051] Construct a rigid transformation model according to the mapping pair, and match the underground geological body three-dimensional model to the high-resolution multispectral terrestrial three-dimensional model by using the rigid transformation model to obtain the regional full-space model; the expression of the rigid transformation model is: wherein, is the original position data of the underground geological body three-dimensional model; is the mapped position data; R is a rotation matrix; T is a translation vector.

[0052] Preferably, the regional full-space model is regionally segmented to obtain a three-dimensional segmentation mask, and the regional full-space model and the three-dimensional segmentation mask are fused to obtain analysis data, including:

[0053] Determine each node of the drill hole position in the regional full-space model as an initial seed point; the initial seed point includes rock-soil body attribute data;

[0054] Calculate attribute difference data of the initial seed point and adjacent points by using the Euclidean distance formula to obtain attribute distance values;

[0055] Add adjacent points with attribute distance values less than a heterogeneous point threshold to a same-region point set of the initial seed point;

[0056] Determine the newly added adjacent point in the same-region point set as a new initial seed point, and return to the step of calculating attribute difference data of the initial seed point and adjacent points by using the Euclidean distance formula to obtain attribute distance values for iterative growth to obtain the same-region point set after propagation is completed;

[0057] Remove the same-region point set with a point quantity less than a region volume minimum value;

[0058] Label the points corresponding to the same-region point set after removal, and assign 0 to the points in the region full-space model except the same-region point set to obtain the three-dimensional segmentation mask.

[0059] Preferably, the training process of the rock-soil body data analysis model comprises:

[0060] Select a three-dimensional convolutional neural network, and add several classification output heads to the three-dimensional convolutional neural network;

[0061] Set target outputs, and assign the same target outputs to the same classification output head; the target outputs comprise rock and soil body boundary lines, rock and soil body types, rock body fragmentation degrees, rock body weathering degrees, rock body overburden soil thicknesses, rock body overburden soil types, soil body deposition types, and soil body underlying bedrock types;

[0062] Construct a sub-loss function according to Dice loss, cross-entropy loss, and mean square error loss, and perform weighted fusion on all sub-loss functions to obtain a total loss function;

[0063] Pre-obtain several groups of the to-be-analyzed data and actual measurement results corresponding to the to-be-analyzed data;

[0064] Take the pre-obtained to-be-analyzed data as input of the three-dimensional convolutional neural network, and perform iterative training on the three-dimensional convolutional neural network by using the total loss function according to the actual measurement results;

[0065] When the proportion of the classification output head with a loss value less than a preset loss threshold exceeds a preset freezing ratio, freeze the network layer parameters of the three-dimensional convolutional neural network, and perform iterative training on the classification output head with a loss value greater than the preset loss threshold by using the sub-loss function to obtain the trained rock-soil body data analysis model.

[0066] The present application discloses the following technical effects:

[0067] The present application provides a kind of rock-soil data analysis method based on multi-source satellite data and geosciences data, by fusing multiple acquisition data data, it solves the defects of lack of drilling data and basic geological data utilization of prior art, realizes the construction and analysis of underground geological body three-dimensional model;By space filling processing to original underground model, it solves the defect of low data density of conventional construction physical model, realizes single data filling and the sum analysis of multi-body boundary data;By upper and lower model and regional segmentation mask fusion, it solves the problem of lack of physical model pre-analysis and high model calculation of conventional technology, realizes the fusion analysis of upper and lower model and the pre-extraction of regional information. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0069] Figure 1 The rock-soil data analysis process diagram based on multi-source satellite data and geosciences data provided for the embodiments of the present application is shown in the figure.

[0070] Figure 2 The drilling data and basic geological data acquisition process diagram provided for the embodiments of the present application is shown in the figure.

[0071] Figure 3 The high-resolution multispectral ground three-dimensional model construction process diagram provided for the embodiments of the present application is shown in the figure.

[0072] Figure 4 The original underground model construction process diagram provided for the embodiments of the present application is shown in the figure.

[0073] Figure 5 The underground geological body three-dimensional model construction process diagram provided for the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0075] The application aims to provide a rock-soil data analysis method based on multi-source satellite data and geosciences data, and solve the problems of lack of utilization of drilling data and basic geological data, low data density, lack of physical model pre-analysis and high model calculation amount in the prior art.

[0076] In order to make the above-mentioned objects, features and advantages of the application more obvious and easy to understand, the application will be further described in detail below with reference to the drawings and specific embodiments.

[0077] Figure 1 A rock-soil data analysis process based on multi-source satellite data and geosciences data provided by the embodiments of the application is shown in FIG. Figure 1 The application provides a rock-soil data analysis method based on multi-source satellite data and geosciences data, which comprises the following steps:

[0078] Step 100: collecting high-resolution panchromatic satellite remote sensing image data, multi-spectral satellite remote sensing image data, DEM data, drilling data and basic geological data of a target analysis area; the basic geological data comprises a geological plan, a geological profile, stratum and lithology information and geological structure information;

[0079] Step 200: fusing the high-resolution panchromatic satellite remote sensing image data, the multi-spectral satellite remote sensing image data and the DEM data to obtain a high-resolution multi-spectral ground three-dimensional model;

[0080] Step 300: fusing the drilling data and the basic geological data to obtain an original underground model;

[0081] Step 400: performing spatial filling processing on the original underground model to obtain an underground geological body three-dimensional model;

[0082] Step 500: performing registration fusion processing on the high-resolution multi-spectral ground three-dimensional model and the underground geological body three-dimensional model to obtain a regional full-space model;

[0083] Step 600: performing regional segmentation on the regional full-space model to obtain a three-dimensional segmentation mask, and fusing the regional full-space model and the three-dimensional segmentation mask to obtain to-be-analyzed data;

[0084] Step 100: inputting the to-be-analyzed data into a pre-trained rock-soil data analysis model for analysis to obtain target rock-soil data analysis results.

[0085] Reference is made to Figure 2 , collecting high-resolution panchromatic satellite remote sensing image data, multi-spectral satellite remote sensing image data, hyperspectral satellite remote sensing image data, DEM data, drilling data and basic geological data of a target analysis area, which comprises the following steps:

[0086] Step 101: defining the boundary of the target analysis area by satellite images and topographic mapping, obtaining the area range;

[0087] Step 102: dividing the area range into a regular longitudinal and transverse grid, obtaining the area grid;

[0088] Step 103: using GPS to mark the grid intersection points of the area grid in the target analysis area, and determining the line segment composed of the grid intersection points located in the same row or column as a longitudinal or transverse edge.

[0089] Reference Figure 2 , collecting high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, hyperspectral satellite remote sensing image data, DEM data, drilling data and basic geological data, also including:

[0090] Step 104: conducting stratigraphic outcrop observation, structure identification, overburden investigation, hydrogeological phenomenon recording and geophysical data collection in the area where each longitudinal or transverse edge is located, obtaining the geological profile;

[0091] Step 105: drilling at each grid intersection point to obtain collected samples;

[0092] Step 106: analyzing the rock core and soil samples to obtain the basic physical parameters;

[0093] Step 107: using geophysical logging technology and in-situ testing technology to collect geological data at each grid intersection point, obtaining stratigraphic parameters and mechanical and hydrological parameters;

[0094] Step 108: integrating the basic physical parameters, stratigraphic parameters, mechanical and hydrological parameters to obtain the drilling data.

[0095] Reference Figure 3 , fusing the high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data and the DEM data to obtain a high-resolution multispectral three-dimensional model on the ground, including:

[0096] Step 201: performing radiation correction, geometric precision correction and image registration on the high-resolution panchromatic satellite remote sensing image data and the multispectral satellite remote sensing image data, respectively;

[0097] Step 202: calculating the correlation of each multispectral band of the multispectral satellite remote sensing image data with the high-resolution panchromatic satellite remote sensing image data to obtain a correlation coefficient; the calculation formula of the correlation coefficient is: where w iis the i-th correlation coefficient; Corr(·) represents the correlation coefficient calculation function; P is the high-resolution panchromatic satellite remote sensing image data; MS i is the i-th multispectral band;

[0098] Step 203: performing weighted summation on the multispectral bands using the correlation coefficient as a weight to obtain a simulated panchromatic image;

[0099] Step 204: performing difference calculation on the high-resolution panchromatic satellite remote sensing image data and the simulated panchromatic image to obtain original difference data, performing triple standard deviation threshold filtering and interpolation processing on the original difference data to obtain optimized difference data;

[0100] Step 205: performing difference calculation on the multispectral band and the projection component of the multispectral band on the simulated panchromatic image to obtain a high-frequency detail component;

[0101] Step 206: The high-frequency detail components are replaced by the optimized difference data in proportion to obtain a high-resolution multispectral image; the expression of the high-resolution multispectral image is: MS′ i =Proj(MS i ,P′)+k·D; where MS′ i is the i-th optimized band data of the high-resolution multispectral image; Proj(·) is the projection formula; P′ is the simulated panchromatic image; k is the scaling factor; and D is the optimized difference data.

[0102] refer to Figure 3 , fusing the high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data and the DEM data to obtain a high-resolution multispectral ground three-dimensional model, further comprising:

[0103] Step 207: using GIS tools to convert the DEM data into a digital terrain model;

[0104] Step 208: Mapping the high-resolution multispectral image to the surface of the digital terrain model to obtain the high-resolution multispectral ground three-dimensional model.

[0105] refer to Figure 4 , fusing the drilling data and the basic geological data to obtain an original underground model, including:

[0106] Step 301: constructing a grid model based on the regional grid and the depth data of the geological profile, and setting the resolution of the grid model;

[0107] Step 302: taking the upper left corner of the grid model as the origin, taking the two edges where the origin is located as the X-axis and Y-axis, and taking the vertical direction depth as the Z-axis to obtain a space coordinate system;

[0108] Step 303: according to the space coordinate system, the drilling data and the basic geological data are injected into the grid model according to the coordinate scaling ratio to obtain the original underground model.

[0109] Reference Figure 5 The original underground model is subjected to a space filling process to obtain a three-dimensional underground geological body model, comprising:

[0110] Step 401: dividing the original underground model into a plurality of cuboids; the attribute data of the vertical edges of the cuboids correspond to the drilling data; the attribute data of the vertical sides of the cuboids correspond to the basic geological data;

[0111] Step 402: mapping the cuboids into unit cubes;

[0112] Step 403: fitting the drilling data according to the unit cubes to obtain an edge line segment function;

[0113] Step 404: fitting the basic geological data according to the unit cubes to obtain a boundary surface function;

[0114] Step 405: constructing a control equation according to the unit cubes and the boundary surface function; the expression of the control equation is: wherein, F(u=0,v,w)=F front (v,w);F(u=1,v,w)=F back (v,w);F(u,v=0,w)=F left (u,w);F(u,v=1,w)=F right (u,w);F is an attribute function; u, v, and w are the horizontal axis, vertical axis, and vertical axis of the unit cube, respectively; F front (v,w)、F back (v,w)、F left (u,w)、F right (u,w) correspond to the boundary surface functions of the front side, back side, left side, and right side of the unit cube, respectively;

[0115] Step 406: using finite difference method to discretize and iteratively solve the control equation from the edge line segment function to obtain an original solution result;

[0116] Step 407: performing tensor product spline fitting and inverse mapping on the original solution result to obtain an optimized solution result;

[0117] Step 408: When the difference of the local first-order derivatives on both sides of the vertical direction of the side surface of the cuboid is greater than a preset abnormal difference threshold, the two cuboids corresponding to the difference of the local first-order derivatives are solved again.

[0118] Specifically, the high-resolution multispectral terrestrial three-dimensional model and the underground geological body three-dimensional model are registered and fused to obtain a regional full-space model, including:

[0119] The spatial coordinate system of the underground geological body three-dimensional model is converted into the global coordinate system of the high-resolution multispectral terrestrial three-dimensional model by using a coordinate transformation algorithm;

[0120] The positioning data of the grid intersection points of the underground geological body three-dimensional model are extracted;

[0121] The position data of the grid intersection points in the underground geological body three-dimensional model and the position data of the positioning data in the high-resolution multispectral terrestrial three-dimensional model are matched to obtain a mapping pair;

[0122] A rigid transformation model is constructed according to the mapping pair, and the underground geological body three-dimensional model is matched to the high-resolution multispectral terrestrial three-dimensional model by using the rigid transformation model to obtain the regional full-space model; the expression of the rigid transformation model is: wherein, is the original position data of the underground geological body three-dimensional model; is the mapped position data; R is a rotation matrix; and T is a translation vector.

[0123] Further, the regional full-space model is regionally segmented to obtain a three-dimensional segmentation mask, and the regional full-space model and the three-dimensional segmentation mask are fused to obtain analysis data, including:

[0124] Each node of a drill hole position in the regional full-space model is determined as an initial seed point; the initial seed point includes rock-soil body attribute data;

[0125] The attribute difference data of the initial seed point and adjacent points are calculated by using a Euclidean distance formula to obtain an attribute distance value;

[0126] Adjacent points with an attribute distance value less than a heterogeneous point threshold are added to a same-region point set of the initial seed point;

[0127] The newly added adjacent points in the same-region point set are determined as new initial seed points, and the step of “calculating the attribute difference data of the initial seed point and adjacent points by using a Euclidean distance formula to obtain an attribute distance value” is returned for iterative growth to obtain a same-region point set with completed propagation.

[0128] The same-area point set whose point quantity is less than the minimum value of the area volume;

[0129] The points corresponding to the same-area point set after the elimination are labeled, and the points in the region full-space model except the points in the same-area point set are assigned a value of 0, to obtain the three-dimensional segmentation mask.

[0130] Preferably, the training process of the rock-soil body data analysis model comprises:

[0131] A three-dimensional convolutional neural network is selected, and a plurality of classification output heads are added to the three-dimensional convolutional neural network;

[0132] Target outputs are set, and the same target outputs are assigned to the same classification output heads; the target outputs include rock and soil body boundary lines, rock and soil body types, rock body fragmentation degrees, rock body weathering degrees, rock body overlying soil thicknesses, rock body overlying soil types, soil body deposition types, and soil body underlying bedrock types;

[0133] Sub-loss functions are constructed according to Dice loss, cross-entropy loss, and mean square error loss, and all the sub-loss functions are weighted and fused to obtain a total loss function;

[0134] A plurality of groups of the to-be-analyzed data and actual measurement results corresponding to the to-be-analyzed data are pre-acquired;

[0135] The pre-acquired to-be-analyzed data are taken as inputs of the three-dimensional convolutional neural network, and the three-dimensional convolutional neural network is iteratively trained according to the actual measurement results and the total loss function;

[0136] When the proportion of the classification output heads whose loss values are less than a preset loss threshold value exceeds a preset freezing ratio, network layer parameters of the three-dimensional convolutional neural network are frozen, and the classification output heads whose loss values are greater than the preset loss threshold value are iteratively trained using the sub-loss functions, to obtain the trained rock-soil body data analysis model.

[0137] Specifically, the Gao Fen 2 satellite is a civilian optical remote sensing satellite with a spatial resolution better than 1 meter, and is equipped with two high-resolution 1-meter panchromatic and 4-meter multispectral cameras, which can provide high-resolution panchromatic satellite remote sensing image data with a resolution of 1 meter and multispectral satellite remote sensing image data with a resolution of 4 meters for the embodiment. The Gao Fen 7 satellite can generate stereo images and can provide high-precision DEM data for the embodiment.

[0138] Preferably, the geographical boundary of the target analysis area is determined through satellite images and topographic surveying techniques, and the area range is defined; the defined area range is divided into regular longitudinal and transverse grids to form area grids, facilitating the standardization and systematization of subsequent data collection; the intersection of each area grid in the target analysis area is marked by using GPS positioning technology, and the grid intersections located in the same row or column are connected into line segments, which are defined as longitudinal and transverse edges, providing positioning reference for subsequent data collection.

[0139] Further, multi-dimensional field observation and data collection are performed on the area where each longitudinal and transverse edge is located to obtain basic geological data; the geological structure in the area is identified; the type, thickness, distribution range and engineering properties of the surface cover layer are investigated; the position, water quantity, water quality of the underground water outcrop, and the recharge relationship between surface water bodies and underground water are recorded; geophysical prospecting techniques are used to obtain information of underground geological structure and lithological interface, assisting in constructing a geological profile.

[0140] Further, drilling is performed at the intersection of each grid to extract rock cores and soil samples, etc. Then, the samples are analyzed in a laboratory to obtain basic physical parameters. In addition, resistivity logging and acoustic logging are used to measure the resistivity and acoustic velocity of the stratum. Finally, the basic physical parameters, stratum parameters, mechanical parameters and hydrological parameters of the intersections are integrated to form drilling data

[0141] Specifically, a simulated panchromatic image is constructed, the simulated panchromatic image is subjected to weighted summation to generate a simulated panchromatic image P' having the highest correlation with the panchromatic image; the correlation coefficient between the multi-spectral bands and the panchromatic image is usually determined, and the formula is as follows:

[0142]

[0143] The difference between the panchromatic image P and the simulated panchromatic image P' is calculated:

[0144] D = P - P'

[0145] The difference contains high-frequency details in the panchromatic image that are not captured by the multi-spectral bands.

[0146] The high-frequency details are injected into the multi-spectral bands, and each multi-spectral band MS i is processed as follows:

[0147] The projection component of the band in the P' direction is calculated;

[0148] The high-frequency detail component of the band is separated by orthogonalization calculation;

[0149] The difference D of the panchromatic image is injected into the high-frequency component of each band in proportion to generate a new band:

[0150] MS'i = Proj(MS i , P') + k · D

[0151] The processed multi-spectral band MS′ i is combined into a high-resolution multi-spectral image, the spatial resolution of which is consistent with that of the panchromatic image, to obtain a high-resolution multi-spectral image

[0152] Further, the pre-processed high-resolution multi-spectral image is spatially registered with DEM data. A model is generated by using the three-dimensional modeling function of GIS or remote sensing software through the following steps:

[0153] A digital terrain model is generated based on DEM;

[0154] The high-resolution multi-spectral image is mapped to the DTM surface as texture data, to give the terrain real colors and object details;

[0155] A geographic coordinate system and a projection coordinate system are specified, to obtain a high-resolution multi-spectral three-dimensional model on the ground.

[0156] Optionally, a grid model is constructed according to the regional grid and the depth data of the geological profile, and the resolution of the grid model is set; the upper left corner of the grid model is taken as the origin, the two edges where the origin is located are taken as the X-axis and Y-axis, and the vertical direction depth is taken as the Z-axis, to obtain a spatial coordinate system; the drilling data and the basic geological data are injected into the grid model according to the coordinate scaling ratio of the spatial coordinate system, to obtain an original underground model.

[0157] Further, the three-dimensional space is divided:

[0158] The line segment extending along the Z-axis is located at the uniform grid nodes (x i ,y j ) in the X-Y plane, denoted as L i ;

[0159] The planes are uniformly distributed along the X direction, and the plane number is corresponding to x = x i+1 ;

[0160] The planes are uniformly distributed along the Y direction, and the plane number is corresponding to y = y j ;

[0161] Cuboid definition: composed of adjacent planes x j+1 → x k , y k+1 → y i,j,k and height interval z i,j → z i+1,j , denoted as C i,j+1 , the four vertical edges of which are: L i+1,j+1 , Li+1,j , L i,j+1 , L i+1,j+1 ;

[0162] Known attribute data is F(x i ,y j ,z k );

[0163] Boundary condition constraints:

[0164] Four sides (front, back, left, right) of the cuboid C i,j,k all belong to vertical planes, and the attribute values thereof are directly provided by the plane nodes;

[0165] The attribute values of the four vertical edges are provided by the line segment nodes.

[0166] Specifically, the cuboid C i,j,k is mapped to a unit cube, and local coordinates (u, v, w) are defined; the side attribute functions are: the front side is F front (v, w), the back side is F back (v, w), the left side is F left (u, w), and the right side is F right (u, w). The control equation is constructed:

[0167]

[0168] The boundary condition is defined as: F(u=0, v, w)=F front (v, w); F(u=1, v, w)=F back (v, w); F(u, v=0, w)=F left (u, w); F(u, v=1, w)=F right (u, w).

[0169] The above control equation is discretized by using the finite difference method, and the solution is smoothed by using the tensor product spline interpolation method;

[0170] When the difference value of the local first-order derivatives of the optimization solution on both sides of the vertical direction of the side of the cuboid is greater than a preset abnormal difference threshold value, the two cuboids corresponding to the local first-order derivative difference value are solved again, and finally the three-dimensional model of the underground geological body is obtained.

[0171] Preferably, the local coordinate system of the subsurface model is converted into a global coordinate system consistent with the surface model by using a coordinate transformation algorithm. The borehole positions have a clear spatial correspondence in the surface model and the subsurface model, and the surface coordinates are extracted from the borehole data as the common control points of the surface model and the subsurface model; a rigid transformation model is used to adjust the subsurface model, so that the projection of the borehole control points in the surface model is aligned with the coordinates of the subsurface model, and the formula is as follows:

[0172]

[0173] Finally, the regional full-space model is obtained.

[0174] Further, first, each borehole position in the regional model is determined as an initial seed point, and the geotechnical attribute data thereof is extracted. Then, the attribute difference between the initial seed point and the adjacent points is calculated by using the Euclidean distance formula to obtain the attribute distance value. If the attribute distance value of a certain adjacent point is lower than a set threshold value, the adjacent point is added to the same regional point set of the initial seed point. Then, the newly added adjacent point is set as a new initial seed point, and the previous step is returned to continue the calculation and iteration until the expansion of the same regional point set is completed. Next, the points with a number less than the minimum volume of the region are removed. Finally, the same regional point set is numbered, and the points not in the set are assigned a value of 0, so as to generate a three-dimensional segmentation mask.

[0175] Preferably, the embodiment selects a three-dimensional convolutional neural network (hereinafter referred to as 3D-CNN) as the basic training architecture of the geotechnical data analysis model. The input data is a three-dimensional regional full-space model and a three-dimensional segmentation mask, and the target output includes a classification task and a regression task. According to the Dice loss, the cross-entropy loss and the mean square error loss, a plurality of sub-loss functions are constructed, and then the sub-loss functions are weighted together to form a total loss function; the training samples are pre-collected and labeled with actual target attribute data or image mask data; the 3D-CNN multi-task network is constructed and the weights are initialized; the training data is input into the 3D-CNN multi-task network in batches to alternately optimize the task loss. When the loss value of the classification output head is lower than a preset threshold value and exceeds a certain proportion, the parameters of the network layer are frozen, and then the classification output head with a loss value higher than the threshold value is further trained by using the sub-loss function, so as to finally complete the training of the geotechnical data analysis model.

[0176] The beneficial effects of the present application are as follows:

[0177] The application effectively utilizes the drilling data and the basic geological data by fusing various collected data, improves the estimation accuracy of the rock-soil data, improves the data density of the underground model by spatial filling processing on the original underground model, provides more fine data for model estimation, avoids the interference of noise on the result caused by low data volume, improves the regional feature analysis effect by fusing the upper and lower models and the regional segmentation mask, improves the boundary extraction speed and accuracy, and improves the estimation accuracy of the conventional rock-soil data.

[0178] The various embodiments are described in a progressive manner in the specification, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the various embodiments can be mutually referred to.

[0179] The principles and implementation manners of the application are described by applying specific examples herein, and the above description of the embodiments is only used to help understand the method of the application and the core idea thereof; meanwhile, for the general technical personnel in the field, the specific implementation manners and application ranges will be changed according to the idea of the application. In conclusion, the content of the specification should not be understood as the limitation of the application.

Claims

1. A rock and soil data analysis method based on multi-source satellite data and geological data, characterized in that: include: Collect high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, DEM data, borehole data and basic geological data of the target analysis area; The basic geological data include: geological plan map, geological cross-section map, stratigraphic and lithologic information, and geological structural information; fusing the high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data, and the DEM data to obtain a high-resolution multispectral ground three-dimensional model; fusing the drilling data and the basic geological data to obtain an original underground model; Performing space filling processing on the original underground model to obtain a three-dimensional model of the underground geological body; Performing registration and fusion processing on the high-resolution multispectral ground three-dimensional model and the underground geological body three-dimensional model to obtain a regional full-space model; Performing regional segmentation on the regional full-space model to obtain a three-dimensional segmentation mask, and fusing the regional full-space model and the three-dimensional segmentation mask to obtain data to be analyzed; The data to be analyzed is input into a pre-trained rock and soil body data analysis model for analysis to obtain target rock and soil body data analysis results.

2. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 1, characterized in that: Collect high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, hyperspectral satellite remote sensing image data, DEM data, borehole data and basic geological data of the target analysis area, including: Defining the boundaries of the target analysis area through satellite imagery and topographic mapping to obtain the area scope; Dividing the region into regular vertical and horizontal grids to obtain a regional grid; The grid intersections of the regional grid are marked in the target analysis area using GPS, and a line segment formed by the grid intersections located in the same row or column is determined as a vertical and horizontal edge.

3. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 2, characterized in that: Collect high-resolution panchromatic satellite remote sensing image data, multispectral satellite remote sensing image data, hyperspectral satellite remote sensing image data, DEM data, borehole data and basic geological data of the target analysis area, including: Conducting stratigraphic outcrop observation, structural identification, cover layer investigation, hydrogeological phenomenon recording, and geophysical data collection in the area where each of the vertical and horizontal edges is located to obtain the geological profile; Drilling is performed at each grid intersection to obtain a sample; Performing core and soil sample analysis on the collected samples to obtain basic physical parameters; Using geophysical logging technology and in-situ testing technology to collect geological data at each grid intersection to obtain formation parameters as well as mechanical and hydrological parameters; The basic physical parameters, the formation parameters, the mechanical and hydrological parameters are integrated to obtain the drilling data.

4. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 3, characterized in that: The high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data, and the DEM data are integrated to obtain a high-resolution multispectral ground three-dimensional model, including: performing radiation correction, geometric precision correction, and image registration on the high-resolution panchromatic satellite remote sensing image data and the multispectral satellite remote sensing image data, respectively; The correlation between each multispectral band of the multispectral satellite remote sensing image data and the high-resolution panchromatic satellite remote sensing image data is calculated to obtain a correlation coefficient; the calculation formula of the correlation coefficient is: Among them, w i is the i-th correlation coefficient; Corr(·) represents the correlation coefficient calculation function; P is the high-resolution panchromatic satellite remote sensing image data; MS i is the i-th multispectral band; Performing weighted summation on the multispectral bands using the correlation coefficient as a weight to obtain a simulated panchromatic image; performing difference calculation on the high-resolution panchromatic satellite remote sensing image data and the simulated panchromatic image to obtain original difference data, and performing triple standard deviation threshold filtering and interpolation processing on the original difference data to obtain optimized difference data; performing difference calculation on the multispectral band and a projection component of the multispectral band on the simulated panchromatic image to obtain a high-frequency detail component; The optimized difference data is proportionally replaced with the high-frequency detail component to obtain a high-resolution multispectral image; the expression of the high-resolution multispectral image is: MS′ i =Proj(MS i ,P′)+k·D; where MS′ i is the i-th optimized band data of the high-resolution multispectral image; Proj(·) is the projection formula; P′ is the simulated panchromatic image; k is the scaling factor; and D is the optimized difference data.

5. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 4, characterized in that: The high-resolution panchromatic satellite remote sensing image data, the multispectral satellite remote sensing image data and the DEM data are integrated to obtain a high-resolution multispectral ground three-dimensional model, further comprising: Converting the DEM data into a digital terrain model using GIS tools; The high-resolution multispectral image is mapped onto the surface of the digital terrain model to obtain the high-resolution multispectral ground three-dimensional model.

6. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 5, characterized in that: The drilling data and the basic geological data are integrated to obtain an original underground model, including: Constructing a grid model according to the regional grid and the depth data of the geological profile, and setting the resolution of the grid model; A spatial coordinate system is obtained by taking the upper left corner of the grid model as the origin, the two sides where the origin is located as the X-axis and the Y-axis, and the vertical depth as the Z-axis; According to the spatial coordinate system, the drilling data and the basic geological data are injected into the grid model according to a coordinate scaling ratio to obtain the original underground model.

7. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 6, characterized in that: Performing space filling processing on the original underground model to obtain a three-dimensional model of the underground geological body, including: The original underground model is divided into a plurality of cuboids; the attribute data of the vertical edges of the cuboids correspond to the drilling data; the attribute data of the vertical sides of the cuboids correspond to the basic geological data; Mapping the cuboid into a unit cube; Fitting the drilling data according to the unit cube to obtain an edge segment function; Fitting the basic geological data according to the unit cube to obtain a boundary surface function; A control equation is constructed based on the unit cube and the boundary surface function; the expression of the control equation is: Where F(u=0,v,w)=F front (v,w); F(u=1,v,w)=F back (v,w); F(u,v=0,w)=F left (u,w); F(u,v=1,w)=F right (u, w); F is the property function; u, v, w are the horizontal axis, vertical axis and vertical axis of the unit cube respectively; F front (v,w),F back (v,w),F left (u,w),F right (u, w) correspond to the boundary surface functions of the front side, back side, left side, and right side of the unit cube respectively; Using the edge segment function as a starting point, the control equation is discretized and iteratively solved using a finite difference method to obtain an original solution result; Performing tensor product spline fitting and reverse mapping on the original solution to obtain an optimized solution; When the local first-order derivative difference of the optimization solution on both sides of the vertical direction of the side of the cuboid is greater than a preset abnormal difference threshold, a secondary solution is performed on the two cuboids corresponding to the local first-order derivative difference.

8. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 7, characterized in that: The high-resolution multispectral ground three-dimensional model and the underground geological body three-dimensional model are registered and fused to obtain a regional full-space model, including: Using a coordinate transformation algorithm, the spatial coordinate system of the underground geological body three-dimensional model is converted into a global coordinate system of the high-resolution multispectral ground three-dimensional model; Extracting positioning data of the grid intersections of the three-dimensional model of the underground geological body; Matching the position data of the grid intersection points in the underground geological body three-dimensional model and the position data of the positioning data in the high-resolution multispectral ground three-dimensional model to obtain a mapping pair; A rigid transformation model is constructed based on the mapping pair, and the rigid transformation model is used to match the underground geological body three-dimensional model to the high-resolution multispectral ground three-dimensional model to obtain the regional full-space model; the expression of the rigid transformation model is: in, The original position data of the three-dimensional model of the underground geological body; is the mapped position data; R is the rotation matrix; T is the translation vector.

9. The rock and soil data analysis method based on multi-source satellite data and geological data according to claim 8, characterized in that: Performing regional segmentation on the regional full-space model to obtain a three-dimensional segmentation mask, and fusing the regional full-space model and the three-dimensional segmentation mask to obtain data to be analyzed, including: Determine each node of the drilling position in the regional full-space model as an initial seed point; the initial seed point includes rock and soil attribute data; Calculate the attribute difference data between the initial seed point and the adjacent points using the Euclidean distance formula to obtain the attribute distance value; Adding adjacent points whose attribute distance values ​​are less than the heterogeneous point threshold to the same area point set of the initial seed point; Determine the newly added adjacent point in the same-region point set as the new initial seed point, and return to step "calculating the attribute difference data between the initial seed point and the adjacent point using the Euclidean distance formula to obtain the attribute distance value" to perform iterative growth to obtain the propagated point set in the same region; Eliminate the point set in the same region whose number of points is less than the minimum value of the region volume; The points corresponding to the removed point set in the same region are labeled and numbered, and the points in the regional full-space model excluding the point set in the same region are assigned a value of 0 to obtain the three-dimensional segmentation mask.

10. The method for analyzing rock and soil data based on multi-source satellite data and geological data according to claim 9, characterized in that: The training process of the rock and soil data analysis model includes: Selecting a three-dimensional convolutional neural network and adding a plurality of classification output heads to the three-dimensional convolutional neural network; Setting target outputs and assigning similar target outputs to the same classification output head; the target outputs include: rock and soil boundary lines, rock and soil types, rock fragmentation degree, rock weathering degree, rock overburden thickness, rock overburden type, soil deposition type, and soil underlying bedrock type; Construct sub-loss functions based on Dice loss, cross entropy loss, and mean square error loss, and perform weighted fusion on all sub-loss functions to obtain the total loss function; Pre-acquiring several groups of the data to be analyzed and actual measurement results corresponding to the data to be analyzed; Using the pre-acquired data to be analyzed as input to the three-dimensional convolutional neural network, and iteratively training the three-dimensional convolutional neural network using the total loss function according to the actual measurement results; When the proportion of the classification output heads whose loss values ​​are less than the preset loss threshold exceeds the preset freezing ratio, the network layer parameters of the three-dimensional convolutional neural network are frozen, and the classification output heads whose loss values ​​are greater than the preset loss threshold are iteratively trained using the sub-loss function to obtain the trained rock and soil data analysis model.

Images