阅读说明：本技术 一种基于地貌参数的土层厚度估算方法 (Soil layer thickness estimation method based on landform parameters ) 是由 肖婷 田卫明 邓云开 于 2021-01-18 设计创作，主要内容包括：本发明公开了一种基于地貌参数的土层厚度估算方法,属于土壤地质勘测领域。方法包括：利用研究区栅格点剖面曲率分布情况获取研究区第一参数,利用研究区坡度分布情况及各子区域的临界坡度和内摩擦角计算各子区域的第二参数,利用研究区各样本点相对位置的投影情况拟合出第三参数函数。对样本点的第一参数、第二参数和第三参数和厚度值进行训练,通过待评估点的剖面曲率、所属子区域和相对位置分别计算出第一参数、第二参数和第三参数,输入训练好的模型,预测出待评估点厚度值。本发明集合了区域斜坡演化历史和现有地表形态信息,能一定程度上突破以往估算方法中精度与通用性不太兼容的问题。(The invention discloses a soil layer thickness estimation method based on landform parameters, and belongs to the field of soil geological survey. The method comprises the following steps: and obtaining a first parameter of the research area by utilizing the distribution condition of the section curvature of the grid points of the research area, calculating a second parameter of each sub-area by utilizing the distribution condition of the gradient of the research area and the critical gradient and the internal friction angle of each sub-area, and fitting a third parameter function by utilizing the projection condition of the relative position of each sample point of the research area. And training the first parameter, the second parameter, the third parameter and the thickness value of the sample point, respectively calculating the first parameter, the second parameter and the third parameter according to the section curvature, the sub-region and the relative position of the point to be evaluated, inputting the trained model, and predicting the thickness value of the point to be evaluated. The method integrates the regional slope evolution history and the existing surface morphology information, and can break through the problem that the precision and the universality are not compatible in the conventional estimation method to a certain extent.)

1. A soil layer thickness estimation method based on landform parameters is applied to stacked layer landforms in a three gorges reservoir area, and is characterized by comprising the following steps:

respectively calculating a first parameter C, a second parameter S and a third parameter P of each sample point in the research area:

A. acquiring section curvatures c of all grid points in a research area; setting a ratio such that a cross-sectional curvature C of a point within the ratio is between (-x, x), for each grid point a first parameter C: setting a first parameter C of the point with the section curvature C larger than x as 0, setting the first parameter C of the point with the section curvature smaller than-x as 1, and setting the first parameters of the other points as (x-C)/2 x; calculating a first parameter C of each sample point according to the section curvature C of each sample point;

B. dividing the research area into several sub-areas, respectively obtaining the slope s, critical slope theta and internal friction angle of each sub-areaFor each sub-area, according to its gradient s, critical gradient theta and internal friction angleSetting a second parameter S:

second parameter

Calculating a second parameter S of each sample point according to the gradient S of each sample point;

C. taking the thickness value of the sample point as a vertical coordinate, taking the relative position p of the sample point as a horizontal coordinate, projecting the relative position p of the sample point on a coordinate system, wherein the value of the relative position p of the point position is the ratio of the horizontal distance from the point to the vertex to the horizontal length of the slope; fitting a function which just can cover all sample points in the coordinate according to the projected sample points, wherein the abscissa of the function is a relative position P, and the ordinate is a third parameter P; calculating a third parameter P of each sample point according to the relative position P of each sample point;

and inputting the thickness value T, the first parameter C, the second parameter S and the third parameter P of each sample point in the research area into a pre-constructed machine learning model for training to obtain a prediction model.

2. The geomorphic parameter based soil layer thickness estimation method of claim 1 wherein the method C of calculating the third parameter is such that the sample point thickness values projected onto the coordinate system are normalized thickness values.

3. The geomorphic parameter based soil layer thickness estimation method of claim 1 wherein in method C of calculating the third parameter, the function that exactly covers all sample points in the coordinates is a polyline function.

4. The geomorphic parameter based soil layer thickness estimation method of claim 1 wherein the fraction is 90%.

5. A soil layer thickness estimation method based on geomorphic parameters as claimed in any of claims 1 to 4 wherein said machine learning model is a random forest machine learning model.

6. The geomorphic parameter based soil layer thickness estimation method of claim 1 wherein the geomorphic parameter based soil layer thickness estimation method is implemented by a soil layer thickness estimation device comprising:

the first computing unit is provided with a geographic information system and a first configuration module, and the first computing unit acquires the section curvature c of all grid points in a research area through the geographic information system; the first configuration module is used for configuring the occupation ratio; the first calculating unit calculates x and-x according to the proportion configured by the first configuration unit, and constructs a calculation formula of a first parameter C:

first parameter

A second calculation unit configured with a region division module that divides the region of interest into a plurality of sub-regions, and a second configuration module for configuring a gradient s, a critical gradient θ, and an internal friction angle of each sub-regionThe second calculating unit configures a second parameter S of each sub-area:

second parameter

The third calculation unit is provided with a coordinate projection module and a function fitting module, wherein the coordinate projection module is used for projecting the thickness value of the sample point as a vertical coordinate and the relative position p of the sample point as a horizontal coordinate onto a coordinate system; the function fitting module is used for fitting a function which just can cover all sample points in the coordinate, the abscissa of the function is a relative position P, and the ordinate of the function is a third parameter P;

the machine learning model is a pre-constructed learning model and is used for learning and modeling the thickness value T, the first parameter C, the second parameter S and the third parameter P of each sample point;

inputting the section curvature C of the point location to be estimated into a first calculating unit to obtain a first parameter C, inputting the sub-region to which the point location to be estimated belongs into a second calculating unit to obtain a second parameter S, inputting the relative position P of the point location to be estimated into a third calculating unit to obtain a third parameter P, and inputting the first parameter C, the second parameter S and the third parameter P into a trained model, so that the corresponding thickness T can be predicted.

Technical Field

The invention relates to the field of soil geological survey, in particular to a soil layer thickness estimation method based on landform parameters.

Background

Soil layer thickness is an important parameter in many geological environment studies, such as slope stability, slope hydrology, seismic local effects, topographic evolution, soil water distribution, heat flux distribution, soil conservation, and the like. When the research area range is small, the thickness of the soil layers of a plurality of sample points is accurately measured by a direct or indirect measuring method, and the precision requirement of the research area can be met. When the research area is in a regional scale, due to the reasons of time, labor, equipment cost and the like, large-scale on-site accurate measurement cannot be completed, so that a scientific estimation method is required to obtain a soil layer thickness map of the whole area to be used as an input parameter of a complex environment model. At present, methods applied to thickness estimation on a regional scale are mainly of two types: firstly, the soil layer thickness is directly approximate to the surface form parameters (elevation, gradient or curvature), and the method has a too rough result and can only be applied to areas with low requirements on the soil layer thickness precision; secondly, a relation model of the soil layer thickness and environmental index factors is established, the environmental index factors are the elevation, the gradient, the curvature, the stratum lithology, the specific water collecting area, the terrain humidity index, the runoff strength index and the like which are directly extracted from the digital elevation, and the soil layer thickness estimation distribution diagram under the method has the problems of poor numerical spatial continuity and large error.

Disclosure of Invention

The invention aims to: aiming at the existing problems, the soil layer thickness estimation method based on the landform parameters is provided to improve the accuracy of soil layer thickness estimation of the landform area of the accumulation layer in the three gorges reservoir area.

The technical scheme adopted by the invention is as follows:

a soil layer thickness estimation method based on landform parameters is applied to three gorges reservoir region accumulation layer landforms, and the method comprises the following steps:

respectively calculating a first parameter C, a second parameter S and a third parameter P of each sample point in the research area:

A. acquiring section curvatures c of all grid points in a research area; setting a ratio such that a cross-sectional curvature C of a point within the ratio is between (-x, x), for each grid point a first parameter C: setting a first parameter C of the point with the section curvature C larger than x as 0, setting the first parameter C of the point with the section curvature smaller than-x as 1, and setting the first parameters of the other points as (x-C)/2 x; calculating a first parameter C of each sample point according to the section curvature C of each sample point;

B. dividing the research area into several sub-areas, respectively obtaining the slope s, critical slope theta and internal friction angle of each sub-areaFor each sub-area, according to its gradient s, critical gradient theta and internal friction angleSetting a second parameter S:

calculating a second parameter S of each sample point according to the gradient S of each sample point;

C. taking the thickness value of the sample point as a vertical coordinate, taking the relative position p of the sample point as a horizontal coordinate, projecting the relative position p of the sample point on a coordinate system, wherein the value of the relative position p of the point position is the ratio of the horizontal distance from the point to the vertex to the horizontal length of the slope; fitting a function which just can cover all sample points in the coordinate according to the projected sample points, wherein the abscissa of the function is a relative position P, and the ordinate is a third parameter P; calculating a third parameter P of each sample point according to the relative position P of each sample point;

and inputting the thickness value T, the first parameter C, the second parameter S and the third parameter P of each sample point in the research area into a pre-constructed machine learning model for training to obtain a prediction model.

Further, in the method C of calculating the third parameter, the thickness value of the sample point projected onto the coordinate system is a normalized thickness value.

Further, in the method C for calculating the third parameter, the function that just can cover all the sample points in the coordinates is a polygonal line function.

Further, the proportion is 90%.

Further, the machine learning model is a random forest machine learning model.

Further, the soil thickness estimation method based on geomorphic parameters is implemented by a soil thickness estimation device, which includes:

the first computing unit is provided with a geographic information system and a first configuration module, and the first computing unit acquires the section curvature c of all grid points in a research area through the geographic information system; the first configuration module is used for configuring the occupation ratio; the first calculating unit calculates x and-x according to the proportion configured by the first configuration unit, and constructs a calculation formula of a first parameter C:

a second calculation unit configured with a region division module that divides the region of interest into a plurality of sub-regions, and a second configuration module for configuring a gradient s, a critical gradient θ, and an internal friction angle of each sub-regionThe second calculating unit configures a second parameter S of each sub-area:

the third calculation unit is provided with a coordinate projection module and a function fitting module, wherein the coordinate projection module is used for projecting the thickness value of the sample point as a vertical coordinate and the relative position p of the sample point as a horizontal coordinate onto a coordinate system; the function fitting module is used for fitting a function which just can cover all sample points in the coordinate, the abscissa of the function is a relative position P, and the ordinate of the function is a third parameter P;

the machine learning model is a pre-constructed learning model and is used for learning and modeling the thickness value T, the first parameter C, the second parameter S and the third parameter P of each sample point;

inputting the section curvature C of the point location to be estimated into a first calculating unit to obtain a first parameter C, inputting the sub-region to which the point location to be estimated belongs into a second calculating unit to obtain a second parameter S, inputting the relative position P of the point location to be estimated into a third calculating unit to obtain a third parameter P, and inputting the first parameter C, the second parameter S and the third parameter P into a trained model, so that the corresponding thickness T can be predicted.

In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:

1. the method utilizes the natural law of geological landforms, performs parameter representation on the characteristics of the geological landforms, combines the relation between geological parameters and thickness, and realizes accurate evaluation on the thickness of the area by means of machine learning. In addition, the method and the device are completely estimated according to the parameter characteristics of the geological landform of the research area, so that the method and the device have strong universality.

2. The method combines the regional slope evolution history and the existing surface morphology information, so that the estimation of regional thickness estimation can break through the precision of the conventional estimation method.

Drawings

The invention will now be described, by way of example, with reference to the accompanying drawings, in which:

FIG. 1 is a schematic diagram of sample point projection and function fitting.

Detailed Description

All of the features disclosed in this specification, or all of the steps in any method or process so disclosed, may be combined in any combination, except combinations of features and/or steps that are mutually exclusive.

Any feature disclosed in this specification (including any accompanying claims, abstract) may be replaced by alternative features serving equivalent or similar purposes, unless expressly stated otherwise. That is, unless expressly stated otherwise, each feature is only an example of a generic series of equivalent or similar features.

Example one

The embodiment discloses a soil layer thickness estimation method based on landform parameters, which is applied to the landform of a stacked layer in a three gorges reservoir area, and comprises the following steps:

A. calculating a first parameter C: the parameter C represents the contribution of the section curvature C to the thickness of the soil layer.

Acquiring the section curvature c of all grid points in a research area through a GIS (geographic information system); setting a ratio such that a cross-sectional curvature c of a point within the ratio is between (-x, x); setting a first parameter C for each grid point:

setting a first parameter C of a point with the section curvature C larger than x as 0; setting a first parameter C of a point with the section curvature smaller than-x as 1; the first parameter for the remaining points is (x-c)/2 x.

Obviously, the value of x varies from one study area to another, which also results in different first parameters C for each grid point in different study areas.

B. A step of calculating a second parameter S:

dividing the research area into several sub-areas, respectively obtaining the slope s, critical slope theta and internal friction angle of each sub-areaCritical slope theta and internal friction angleDepending on the characteristics of the rock mass parameters in the area under investigation. For each sub-area, according to its gradient s, critical gradient theta and internal friction angleSetting a second parameter S:

when the gradient is larger than the critical gradient theta, the corrosion and the slippage of the surface substances are considered to be thorough, and only the bedrock surface is left; when the gradient is less than the internal friction angleAnd in time, the influence of the gradient on the thickness of the soil layer is considered to be ignored.

C. Calculating a third parameter P: the parameter P is used for expressing the relation between the relative position P of a point on the slope section and the thickness of the soil layer.

The relative position p of a point is the ratio of the horizontal distance from the point to the apex to the horizontal length of the ramp. Taking the thickness value of the sample point as a vertical coordinate, and projecting the relative position p of the sample point as a horizontal coordinate onto a coordinate system; and fitting a function which just can cover all the sample points in the coordinate according to the projected sample points, wherein the abscissa of the function is the relative position P, the ordinate is the third parameter P, and the third parameter P of any point in the research area can be obtained through the function.

Inputting the thickness value T, the first parameter C, the second parameter S and the third parameter P of each sample point in the research area into a pre-constructed machine learning model for training, wherein the trained model can be used for predicting the thickness value of the whole research area. The three parameters C, S, P of the point location to be estimated are input into the trained model, and the corresponding thickness T can be predicted. The three parameters C, S, P are calculated as follows: calculating the section curvature C of the point to be evaluated by using a first parameter C calculation method, calculating the sub-region of the point to be evaluated by using a second parameter S calculation method, and calculating the relative position P of the point to be evaluated by using a third parameter P calculation method.

The scheme of the embodiment of the invention can be expanded to be used in other geological landforms, and only the parameters are required to be adaptively adjusted according to the landform characteristics of the researched area.

Example two

The embodiment discloses a soil layer thickness estimation method based on landform parameters, which is applied to the landform of a stacked layer in a three gorges reservoir area. The thickness value T of the sample point in the study region has a nonlinear function relationship (T ═ f (C, S, P)) with the three parameters C, S, P, and this embodiment trains and models the sample point through a random forest machine learning model, and then predicts the thickness value of the whole study region by using the trained model. The calculation method of the three parameters is as follows:

1. method for calculating first parameter C

The first parameter C represents the contribution of the section curvature C to the thickness of the soil layer. First, the relationship (positive, negative or other simple multi-segment linear relationship) between the section curvature and the soil thickness in the study area is determined through field investigation and geological knowledge. Taking the landform of the accumulation layer in the three gorges reservoir area as an example, the slope section curvature c and the soil layer thickness T are in a negative correlation relationship, the section curvatures of all grid points in the research area are extracted through the GIS, and 90% of the section curvature c is between (-x, x), then: the point with the section curvature larger than x has the parameter C of 0; a point with a section curvature smaller than-x, the first parameter C of which is 1; the first parameter C value of the rest points is (x-C)/2 x.

2. Method for calculating second parameter S

The second parameter S represents the degree of contribution of the gradient S to the thickness of the soil layer. The method can be approximately judged according to the natural law: when the gradient is larger than the critical gradient theta, the corrosion and the slippage of the surface substances are considered to be thorough, and only the bedrock surface is left; when the gradient is less than the internal friction angleAnd in time, the influence of the gradient on the thickness of the soil layer is considered to be ignored. Critical slope theta and internal friction angleDepending on the characteristics of the rock mass parameters in the area under investigation. The present embodiment divides the study area into several sub-areas, and the second parameter S in the same sub-area is constant. The value and the gradient S, the critical gradient theta and the internal friction angle of the second parameter SThe three are related, and the formula is as follows:

3. method for calculating third parameter P

The third parameter P is used for representing the relation between the relative position P of the point on the slope section and the thickness of the soil layer. On a slope, the relative position p of the top of the slope is 0, the relative position p of the bottom of the slope is 1, and the value of the relative position p at any position is the ratio of the horizontal distance from the position to the top to the horizontal length of the slope. And (3) establishing a two-dimensional coordinate graph, wherein the abscissa is the relative position p, the ordinate is the thickness value of the sample point after the thickness normalization, and all the sample points are projected in the graph. A polyline/simple function curve (i.e. a function) is used, which exactly covers all sample points in the graph, as shown in fig. 1, with the relative position P on the abscissa and the third parameter P on the ordinate (as shown in the following figure). By means of this polyline/curve function (P ═ f (P)), the value of the third parameter P at an arbitrary position in the region of interest can be obtained.

EXAMPLE III

The embodiment discloses a soil layer thickness estimation system based on landform parameters, which comprises a first calculation unit, a second calculation unit, a third calculation unit and a machine learning model.

The first computing unit is provided with a geographic information system and a first configuration module, and the first computing unit acquires the section curvature c of all grid points in a research area through the geographic information system; the first configuration module is used for configuring the occupation ratio; the first calculating unit calculates x and-x according to the proportion configured by the first configuration unit, and constructs a calculation formula of a first parameter C:

and inputting the section curvature C of the point to be evaluated into a first calculating unit to obtain a corresponding first parameter C.

A second calculation unit configured with a region division module that divides the region of interest into a plurality of sub-regions, and a second configuration module for configuring a gradient s, a critical gradient θ, and an internal friction angle of each sub-regionThe second calculating unit configures a second parameter S of each sub-area:

and matching the second parameter S of the point to be evaluated through the sub-area to which the point to be evaluated belongs and the second parameter S of each sub-area calculated by the second calculating unit.

The third calculation unit is provided with a coordinate projection module and a function fitting module, wherein the coordinate projection module is used for projecting the thickness value of the sample point as a vertical coordinate and the relative position p of the sample point as a horizontal coordinate onto a coordinate system; the function fitting module is used for fitting a function which just can cover all sample points in the coordinate, the abscissa of the function is the relative position P, and the ordinate is the third parameter P. For the function fitting module, generally, as long as the fitted function can be expressed by a function and meets the requirement of covering the sample point, the difference of the third parameter P calculated by the fitting result is not too large, the final training result is not greatly affected, the situations such as jumping and the like do not occur, and the function fitting module meets the natural law.

And corresponding the relative position P of the point to be evaluated to the function fitted by the third calculating unit, so as to obtain a corresponding third parameter P.

And the machine learning model is a pre-constructed learning model (such as a random forest machine learning model or other machine learning models, the learning model is not limited in the invention), and the machine learning model is used for learning the thickness value T, the first parameter C, the second parameter S and the third parameter P of each sample point to construct a prediction model. It will be appreciated by those skilled in the art that for the trained model, the first parameter C, the second parameter S and the third parameter P are input, and the thickness value T is output.

The embodiment also discloses a method for realizing estimation of soil layer thickness based on landform parameters by using the soil layer thickness estimation device, which comprises the following steps:

inputting the section curvature C of the point location to be estimated into a first calculating unit to obtain a first parameter C, inputting the sub-region to which the point location to be estimated belongs into a second calculating unit to obtain a second parameter S, inputting the relative position P of the point location to be estimated into a third calculating unit to obtain a third parameter P, and inputting the first parameter C, the second parameter S and the third parameter P into a trained model, so that the corresponding thickness T can be predicted.

The invention is not limited to the foregoing embodiments. The invention extends to any novel feature or any novel combination of features disclosed in this specification and any novel method or process steps or any novel combination of features disclosed.