An intelligent analysis and decision-making system for surveying and mapping geographic information data

Abstract

The present invention discloses an intelligent analysis and decision-making system for surveying and mapping geographic information data, belonging to the technical field of intelligent analysis and decision-making systems. The obtained data is cleaned and preprocessed through a confidence judgment optimization module, the data is standardized, the standardized data is subjected to data enhancement processing, the enhanced data is subjected to multi-source data fusion through a multi-source data fusion module, the models are fused, the logistic regression and support vector machine models are fused, the fused model is evaluated and monitored, and the optimized confidence level is output to the comprehensive evaluation module, thereby realizing the ability to accurately output the confidence values of point, line, surface, and volume data. This provides a quantitative basis for judging the reliability of the data. High-confidence data indicates that the model has high accuracy and credibility in classifying or positioning it, and reduces the technical problem that the data island phenomenon is prominent, resulting in relatively poor accuracy of the analysis structure.

Description

Technical Field

[0001] The present invention relates to an intelligent analysis and decision-making system, in particular to an intelligent analysis and decision-making system for surveying and mapping geographic information data, and belongs to the technical field of intelligent analysis and decision-making systems. Background Art

[0002] In the prior art, when collecting data, unmanned aerial vehicle images, remote sensing, and LiDAR are respectively used to collect geographic information data and transmit it to the data collection module. Although the types and methods of collected data are relatively diverse, the phenomenon of data islands is prominent, resulting in relatively poor accuracy of the analysis results.

[0003] One of the specific problems is that in the process of defining general feature classes of points, lines, planes, and solids, and expanding attribute fields such as feature class, confidence level, and data source, the confidence level is used to judge the relevant reliability or accuracy of points, lines, planes, and solids. However, because the data used to calculate the confidence level itself has biases, noise, or is incomplete, even if the confidence level calculation method is correct, the obtained confidence level cannot accurately reflect the true situation of points, lines, planes, and solids.

[0004] The confidence levels calculated by different machine learning models or algorithms are different, and the models themselves have problems of overfitting and underfitting, which affect the reliability of the confidence level.

[0005] In a complex geographical environment or a scenario with multi-factor interactions, a single confidence level indicator cannot comprehensively and accurately reflect all situations of points, lines, planes, and solids. Therefore, an intelligent analysis and decision-making system for surveying and mapping geographic information data is designed to solve the above problems. Summary of the Invention

[0006] The main object of the present invention is to provide an intelligent analysis and decision-making system for surveying and mapping geographic information data.

[0007] The object of the present invention can be achieved by adopting the following technical solutions:

[0008] An intelligent analysis and decision-making system for surveying and mapping geographic information data includes a data collection module for collecting data on geographic information to be surveyed and mapped;

[0009] A data preprocessing module for cleaning the collected data;

[0010] A feature extraction module for extracting points, lines, planes, solids, and expanding attribute fields such as feature class, confidence level, and data source from the data processed by the data preprocessing module;

[0011] A model training module for training various machine learning models with the extracted feature data;

[0012] It also includes a confidence judgment optimization module, which is used to judge and optimize the confidence of the point, line, surface, and volume data extracted by the feature extraction module and the data trained by the model training module, and output the optimized confidence value of each data object;

[0013] The confidence judgment optimization module includes a multi-source data fusion module;

[0014] The multi-source data fusion module fuses the coordinate transformation formula and the affine transformation formula, and processes the data extracted by the feature extraction module through the fused formula and then inputs it into the model training module;

[0015] The comprehensive evaluation module comprehensively evaluates and analyzes the data by combining the confidence judgment results, domain knowledge and rules, and multi-index comprehensive evaluation.

[0016] Preferably, the judgment method of the confidence judgment optimization module specifically includes the following steps:

[0017] S11: Clean and preprocess the obtained data;

[0018] S12: Standardize the data;

[0019] S13: Perform data augmentation on the standardized data;

[0020] S14: Perform multi-source data fusion on the augmented data in S13 through the multi-source data fusion module;

[0021] S15: Fuse the models, and fuse the logistic regression and support vector machine models;

[0022] S16: Evaluate and monitor the fused model;

[0023] S17: Output the optimized confidence to the comprehensive evaluation module.

[0024] Preferably, in S11, mainly clean the outlier detection, and specifically use the 3 Principle to process the point, line, surface, and volume data as follows. The formula is used to standardize the data: ;

[0025] Among them, is the value after standardization;

[0026] is the original value;

[0027] is the mean value;

[0028] is the standard deviation.

[0029] Preferably, in S13, the following formula is used to enhance the data. For a point in the image , the coordinate calculation formula for the rotated point is: ;

[0030] where is the center coordinate of the image.

[0031] Preferably, in S14, multi-source data fusion is performed by integrating and deriving the seven-parameter Bursa model and the affine transformation formula;

[0032] First, perform the transformation of the seven-parameter Bursa model to obtain the spatial coordinates , and then project it onto a plane to obtain the plane coordinates , and then perform an affine transformation on the plane coordinates to obtain the final coordinates .

[0033] Preferably, for the transformation of the seven-parameter Bursa model: ;

[0034] Convert the spatial coordinates to plane coordinates through projection;

[0035] Specifically, the projection function is used:

[0036] ;

[0037] ;

[0038] Finally, perform an affine transformation: ;

[0039] The obtained fused formula is: .

[0040] Preferably, before the model fusion in S15, model selection and optimization processing are also included;

[0041] The gradient descent update weight formula of logistic regression is used: ;

[0042] where is the th weight;

[0043] α is the learning rate;

[0044] J(w) is the loss function, .

[0045] Preferably, the S15 model fusion specifically includes the assumption that there are a logistic regression model LR and a support vector machine model SVM. For the input x, their prediction probabilities are respectively and . The formula for the fused prediction probability is: ;

[0046] Preferably, S16 specifically includes evaluating and monitoring the fused model using the following formula: Accuracy ;

[0047] where TP is the true positive;

[0048] TN is the true negative;

[0049] FP is the false positive;

[0050] FN is the false negative.

[0051] Preferably, S17 specifically includes establishing a comprehensive evaluation level and introducing domain knowledge and rules;

[0052] According to domain knowledge, the river line feature should be continuous and have a consistent flow direction. If a line feature has a breakpoint or inconsistent flow direction, its confidence level is reduced;

[0053] Let the original confidence level be . If there is a breakpoint, the confidence level is adjusted to ;

[0054] If the flow direction is inconsistent, the confidence level is adjusted to ;

[0055] Adopt multi-index comprehensive evaluation, combining the classification of polygon features with boundary roughness and classification confidence;

[0056] Let the classification confidence level of the polygon feature be , the boundary roughness is R in the range of 0 - 1, and the larger the value, the rougher the boundary. The comprehensive evaluation confidence level , where is the weight .

[0057] The beneficial technical effects of the present invention:

[0058] An intelligent analysis and decision-making system for surveying and mapping geographic information data provided by the present invention uses a confidence judgment optimization module to clean and preprocess the acquired data, perform standardization processing on the data, perform data enhancement processing on the standardized data, perform multi-source data fusion on the enhanced data through a multi-source data fusion module, fuse the models, fuse the logistic regression and support vector machine models, evaluate and monitor the fused model, and output the optimized confidence level to the comprehensive evaluation module, thereby realizing the construction of machine learning models such as logistic regression. By learning and training on the extracted rich features, it can accurately output the confidence values of point, line, surface, and volume data. This provides a quantitative basis for judging the reliability of the data. High-confidence data indicates that the model has high accuracy and credibility in classifying or positioning it, and reduces the technical problem that the prominent data island phenomenon leads to poor accuracy of the analysis structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 FIG. is a system diagram of a preferred embodiment of an intelligent analysis and decision-making system for surveying and mapping geographic information data according to the present invention;

[0060] Figure 2 FIG. is a flowchart of a confidence judgment optimization module of a preferred embodiment of an intelligent analysis and decision-making system for surveying and mapping geographic information data according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] To make the technical solutions of the present invention clearer and more definite to those skilled in the art, the present invention will be further described in detail below in conjunction with the embodiments and the drawings, but the embodiments of the present invention are not limited thereto.

[0062] When performing intelligent analysis and decision-making on surveying and mapping geographic information data, it is necessary to first collect the geographic information data to be surveyed;

[0063] In the prior art, when collecting data, the geographic information data is collected by using unmanned aerial vehicle images, remote sensing, and LiDAR respectively and transmitted to the data collection module. Although the types and methods of collecting data are relatively diverse, the data island phenomenon is prominent, resulting in poor accuracy of the analysis structure. The present invention reduces the problem of data islands by fusing the data of unmanned aerial vehicle images, remote sensing, and LiDAR;

[0064] The following operations are performed through the data preprocessing module:

[0065] Establish a basic layer to unify the spatial and temporal benchmarks;

[0066] Use the WGS84 geographic coordinate system as the benchmark and support dynamic conversion to the UTM projection;

[0067] Unify the recording of data acquisition time and support the overlay analysis of multi-temporal data.

[0068] Establish a data layer and adopt a structured description of multi-source data;

[0069] Through the feature extraction module, define general feature classes of points, lines, surfaces, and volumes, and extend the attribute fields of feature categories, confidence levels, and data sources;

[0070] The model training module is used to train various machine learning models with the extracted feature data.

[0071] In the process of defining general feature classes of points, lines, surfaces, and volumes and extending the attribute fields of feature categories, confidence levels, and data sources, the confidence level is used to judge the relevant reliability or accuracy of points, lines, surfaces, and volumes. However, because the data used to calculate the confidence level itself has deviations, noise, or is incomplete, even if the confidence level calculation method is correct, the obtained confidence level cannot accurately reflect the true situation of points, lines, surfaces, and volumes;

[0072] The confidence levels calculated by different machine learning models or algorithms are different, and the models themselves have overfitting and underfitting problems, which affect the reliability of the confidence level;

[0073] In a complex geographical environment or a scenario with multi-factor interactions, a single confidence level indicator cannot comprehensively and accurately reflect all situations of points, lines, surfaces, and volumes;

[0074] Therefore, in the present invention, the method for solving the above problems through the confidence level judgment optimization module is specifically as follows:

[0075] First, clean and preprocess the acquired data;

[0076] Specifically, it includes outlier detection;

[0077] The coordinates of point data use 3 Principle formula:

[0078] For a set of coordinate data of points assumed to be one-dimensional coordinate x, first calculate the mean value: , standard deviation ;

[0079] The outlier judgment condition is: or .

[0080] Example 1: Assume there is coordinate data of points x = [1, 2, 3, 4, 100], calculate the mean value .

[0081] Calculate the standard deviation .

[0082] Outlier judgment: 1 < 22 - 3×39.01 does not hold, < 22 - 3×39.01 does not hold, 3 < 22 - 3×39.01 does not hold, 4 < 22 - 3×39.01 does not hold, 100 > 22 + 3×39.01 does not hold. Here, 100 is obviously an outlier.

[0083] The coordinates of the line data use 3 Principle formula:

[0084] For a certain coordinate component (x - coordinate or y - coordinate) in the line data, assume it contains n data points, which are respectively Then the mean value of this coordinate component The calculation formula is to first calculate the mean value , standard deviation ;

[0085] The outlier judgment condition is: or .

[0086] The coordinates of the surface data use 3 Principle formula: Extract the x - coordinates and y - coordinates of all points from the surface data, and record them as and where n is the number of points in the surface data;

[0087] Calculate the mean value:

[0088] The mean value of the x - coordinate : ;

[0089] The mean value of the y - coordinate : ;

[0090] Calculate the standard deviation:

[0091] The standard deviation of the x - coordinate ;

[0092] The standard deviation of the y - coordinate ;

[0093] Outlier judgment: For the x - coordinate, if or , then the corresponding point is an outlier;

[0094] For the y - coordinate, if or , then the corresponding point is an outlier.

[0095] The coordinates of the volume data use 3 Principle formula:

[0096] Extract the x - coordinates, y - coordinates, and z - coordinates of all points from the volume data, denoted as , , where n is the number of points in the surface data;

[0097] Calculate the mean values:

[0098] The mean value of the x - coordinates ;

[0099] The mean value of the y - coordinates ;

[0100] The mean value of the z - coordinates ;

[0101] Calculate the standard deviations: ;

[0102] The standard deviation of the y - coordinates ;

[0103] The standard deviation of the z - coordinates ;

[0104] Outlier judgment:

[0105] For the x - coordinates, if or , then the corresponding point is an outlier;

[0106] For the y - coordinates, if or , then the corresponding point is an outlier;

[0107] For the z - coordinates, if or , then the corresponding point is an outlier.

[0108] Standardize the data and use the following formula: , where is the value after standardization, is the original value, is the mean value, is the standard deviation.

[0109] Example 2:

[0110] For the above x = [1, 2, 3, 4, 100];

[0111] (The ) has been calculated;

[0112] .

[0113] Perform data augmentation on the standardized data; The coordinates of the rotated point are calculated by the formula: ;

[0114] where is the central coordinate of the image.

[0115] Perform multi-source data fusion on the augmented data;

[0116] The present invention adopts a method of integrating and deriving the seven-parameter Bursa model and the affine transformation formula using a multi-source data fusion module:

[0117] Let's first observe:

[0118] The formula of the seven-parameter Bursa model is: ;

[0119] where is the transformed coordinate, is the original coordinate, m is the scale change parameter, is the rotation parameter, is the translation parameter.

[0120] The affine transformation formula is: ;

[0121] where is the original coordinate, is the transformed coordinate, and a, b, c, d, e, f are the affine transformation parameters.

[0122] Introduce the idea of affine transformation into coordinate transformation, especially in the process of transforming from geodetic coordinates to plane coordinates. On the basis of the seven-parameter Bursa model, further consider the affine transformation in the plane to adapt to more local deformation situations.

[0123] Assume that the transformation of the seven-parameter Bursa model is first performed to obtain the spatial coordinate , and then it is projected onto the plane to obtain the plane coordinate , and then an affine transformation is performed on the plane coordinate to obtain the final coordinate .

[0124] The transformation of the seven-parameter Bursa model: ;

[0125] The spatial coordinate is converted into the plane coordinate by projection. Taking the Gauss projection as an example, its formula is relatively complex and is simplified here as the projection function , ;

[0126] Finally, perform an affine transformation: ;

[0127] Combining the above steps, we obtain the fused formula: ;

[0128] The integrated formula not only considers the conversion between geodetic coordinate systems but also can further adjust the plane coordinates through affine transformation to adapt to local terrain changes or projection deformations, thereby improving the accuracy of coordinate conversion.

[0129] In practical applications, appropriate parameters and projection methods are selected according to specific situations, and in some cases, the calculation process can be simplified. For example, if only simple translation and scaling are required, the affine transformation parameters can be adjusted without performing complex seven-parameter calculations.

[0130] Next, select and optimize the model;

[0131] The present invention adopts the gradient descent update weight formula of logistic regression: , where is the j-th weight, α is the learning rate, and J(w) is the loss function (logarithmic loss function .

[0132] Example 3: Assume there is only one feature x, y = [1,0] x = [1,2], the initial weight w = [0], and the learning rate α = 0.1.

[0133] Calculate .

[0134] The loss function .

[0135] Calculate the gradient ,

[0136] .

[0137] Update the weight w = 0 - 0.1 × 0.25 = -0.025.

[0138] Then fuse the models. The formula for fusing the logistic regression and support vector machine models is as follows:

[0139] Suppose there are a logistic regression model LR and a support vector machine model SVM. For the input x, their predicted probabilities are respectively and , and the fused predicted probability .

[0140] Example 4: Assume that the predicted probability of the logistic regression model for input \(x\) is \(p_{LR}(x) = 0.6\), and the predicted probability of the support vector machine model for input \(x\) is \(p_{SVM}(x) = 0.4\). The predicted probability after fusion .

[0141] The present invention uses a comprehensive evaluation module to evaluate and monitor the fused model, and specifically uses the following formula:

[0142] Accuracy rate , where \(TP\) is the true positive, \(TN\) is the true negative, \(FP\) is the false positive, and \(FN\) is the false negative.

[0143] Example 5: Assume that \(TP = 80\), \(TN = 70\), \(FP = 10\), and \(FN = 20\), then .

[0144] Establish a comprehensive evaluation level and introduce domain knowledge and rules;

[0145] Assume that according to domain knowledge, river line features should be continuous and have the same flow direction. If a line feature has a break or inconsistent flow direction, its confidence level is reduced.

[0146] Let the original confidence level be , if there is a break, the confidence level is adjusted to ;

[0147] If the flow direction is inconsistent, the confidence level is adjusted to .

[0148] Example 6: Assume that the original confidence level of a river line feature , and it is detected that there is an inconsistent flow direction. The adjusted confidence level \(C = 0.54\).

[0149] Further adopt multi-index comprehensive evaluation, combining the classification of polygon features with boundary roughness and classification confidence level;

[0150] Let the classification confidence level of the polygon feature be , the boundary roughness be \(R\) (range 0 - 1, the larger the value, the rougher the boundary), and the comprehensive evaluation confidence level where is the weight .

[0151] Example 7: Assume that the classification confidence level is , the boundary roughness \(R = 0.2\), and the weight = 0.6, \(C = 0.74\).

[0152] Through the above specific formulas, calculation processes and implementation solutions, the limitations in confidence judgment are solved to a certain extent, and the quality of data related to points, lines, surfaces and volumes and the accuracy of confidence judgment are improved.

[0153] As described above, only further embodiments of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the scope disclosed by the present invention, according to the technical solution and concept of the present invention, makes equivalent substitutions or changes, all belong to the protection scope of the present invention.

Claims

1. An intelligent analysis and decision-making system for surveying and mapping geographic information data, including a data acquisition module for acquiring data on the geographic information to be surveyed and mapped; A data preprocessing module for cleaning the acquired data; A feature extraction module for extracting points, lines, surfaces, volumes, and extended attribute fields such as feature categories, confidence levels, and data sources from the data processed by the data preprocessing module; A model training module for training various machine learning models with the extracted feature data; It is characterized in that: It further includes a confidence level judgment and optimization module for judging and optimizing the confidence levels of the point, line, surface, and volume data extracted by the feature extraction module and the data trained by the model training module, and outputting the optimized confidence level values of each data object; The confidence level judgment and optimization module includes a multi-source data fusion module; The multi-source data fusion module fuses the coordinate transformation formula and the affine transformation formula, and processes the data extracted by the feature extraction module through the fused formula and inputs it into the model training module; A comprehensive evaluation module for comprehensively evaluating and analyzing the data by combining the confidence level judgment results, domain knowledge and rules, and multiple indicators; The judgment method of the confidence level judgment and optimization module specifically includes the following steps: S11: Clean and preprocess the acquired data; S12: Standardize the data; S13: Perform data enhancement processing on the standardized data; S14: Perform multi-source data fusion on the enhanced data in S13 through the multi-source data fusion module; S15: Fuse the models, and fuse the logistic regression and support vector machine models; S16: Evaluate and monitor the fused model; S17: Output the optimized confidence level to the comprehensive evaluation module; In S11, outlier detection is mainly used for cleaning. Specifically, the following formula is used to process point, line, surface, and volume data to standardize the data: ; where is the value after standardization; ​ is the original value; is the mean value; is the standard deviation; In S13, the following formula is used to perform data enhancement processing on the data. For the points in the image , the coordinate calculation formula for the rotated point is as follows: ； Among them, is the central coordinate of the image; In S14, the seven-parameter Bursa model and the affine transformation formula are integrated and deduced to perform multi-source data fusion in this way; First, perform the conversion of the seven-parameter Bursa model to obtain the spatial coordinates , and then project it onto a plane to obtain the planar coordinates . Next, perform an affine transformation on the planar coordinates to obtain the final coordinates ; S17 specifically includes establishing a comprehensive evaluation level and introducing domain knowledge and rules; According to domain knowledge, river line features should be continuous and have the same flow direction. If a line feature has breakpoints or inconsistent flow directions, its confidence level is reduced; Let the original confidence level be , if there is a breakpoint, the confidence level is adjusted to ; If the flow directions are inconsistent, the confidence level is adjusted to ; Adopt multi-index comprehensive evaluation, combining polygon feature classification with boundary roughness and classification confidence level; Let the classification confidence of the planar feature be , the boundary roughness be R in the range of 0 - 1, and the larger the value, the rougher the boundary. The comprehensive evaluation confidence , where is the weight .

2. The intelligent analysis and decision-making system for surveying and mapping geographic information data according to claim 1, wherein: Conversion of the seven-parameter Bursa model: ; Convert the spatial coordinates to planar coordinates through projection ; Specifically adopt a projection function: ； ； Finally, perform an affine transformation: ; The obtained fused formula is: 。 3. An intelligent analysis and decision-making system for surveying and mapping geographic information data according to claim 2, characterized in that: Before the model fusion in S15, it also includes the selection and optimization processing of the model; The weight update formula of gradient descent using logistic regression: ; Among them, is the th weight; is the learning rate; is the loss function, 。 4. An intelligent analysis and decision-making system for surveying and mapping geographic information data according to claim 2, characterized in that: The S15 model fusion specifically includes the assumption that there are a Logistic Regression model LR and a Support Vector Machine model SVM. For the input x, their prediction probabilities are respectively and . The prediction probability formula after fusion is: 。 5. The intelligent analysis and decision-making system for surveying and mapping geographic information data according to claim 2, characterized in that: S16 specifically includes evaluating and monitoring the fused model, and specifically using the following formula: Accuracy ; Where TP is the true positive example; TN is the true negative example; FP is the false positive example; FN is the false negative example.

Images