CN111274738B - Elevation abnormal value calculation method - Google Patents
Elevation abnormal value calculation method Download PDFInfo
- Publication number
- CN111274738B CN111274738B CN202010117015.1A CN202010117015A CN111274738B CN 111274738 B CN111274738 B CN 111274738B CN 202010117015 A CN202010117015 A CN 202010117015A CN 111274738 B CN111274738 B CN 111274738B
- Authority
- CN
- China
- Prior art keywords
- elevation
- abnormal
- fitting
- residual error
- calculating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C25/00—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C5/00—Measuring height; Measuring distances transverse to line of sight; Levelling between separated points; Surveyors' levels
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Manufacturing & Machinery (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention discloses a method for calculating an elevation abnormal value, which comprises the following steps: 1: acquiring elevation abnormal values of a plurality of GPS level joint measuring points in a target area, performing pre-fitting by adopting a quadric surface method, taking an error in three times as a limit difference, and removing gross errors; s2: fitting the remaining N elevation abnormal values by adopting a quadratic polynomial; 3: calculating N elevation anomaly calculation values by utilizing a quadratic polynomial model, and calculating a fitting residual error; 4: constructing a wavelet neural network, wherein the input of the network is a plane coordinate and an elevation abnormal calculation value, and the output is a residual error; training the wavelet neural network by adopting the plane coordinates, the elevation anomaly calculation values and the residual errors of the N GPS level joint measurement points; 5: calculating the abnormal calculation value xi of the elevation of the land to be measured t (ii) a Coordinates and xi of the place to be surveyed t Inputting the data into the trained wavelet neural network to obtain the residual error of the land to be measured, and further calculating the elevation abnormal value of the land to be measured. The method has small error of the elevation abnormal value obtained in a large-area measuring area, and can calculate to obtain a more accurate elevation abnormal value.
Description
Technical Field
The invention relates to the field of regional quasi-geoid refinement, in particular to a calculation method based on an elevation abnormal value.
Background
In recent years, satellite positioning technology has been widely used due to its advantages of high efficiency, high precision, and strong real-time performance. The elevation abnormal value is the difference between the geodetic height of the GPS at the same point and the normal height of the precise level. And the elevation abnormal value at a certain point is obtained, and the conversion from the geodetic height to the normal height of the GNSS can be realized.
According to the specification of the 4.4 th item of the national standard engineering measurement Specification (GB 50026-2007), when the GPS fitting elevation is calculated, a plane fitting method can be adopted to obtain an elevation abnormal value for a small measurement area with flat terrain, and a curved surface fitting method can be adopted to obtain the elevation abnormal value for a large-area measurement area with large terrain fluctuation. At present, a quadratic curve is usually adopted to carry out elevation outlier fitting on a large-area measuring area, and the error obtained by the method is usually large.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide an elevation abnormal value calculation method which is small in error of an elevation abnormal value obtained in a large-area measuring area and capable of calculating to obtain a more accurate elevation abnormal value.
The technical scheme is as follows: the invention adopts the following technical scheme:
an elevation outlier calculation method comprising:
s1: acquiring elevation abnormal values of a plurality of GPS level connection points in a target area, pre-fitting the acquired data by adopting a quadric surface method, taking an error in three times as a limit difference, and removing gross errors;
s2: eliminating the height abnormal values of the N GPS level combined measuring points after the gross errors are eliminated from S1Fitting by using a quadratic polynomial:
ξ(x,y)=a 0 +a 1 x+a 2 y+a 3 x 2 +a 4 xy+a 5 y 2
where ξ (x, y) is the elevation anomaly at planar coordinates (x, y); a is 0 ,a 1 ,…,a 5 Is a fitting coefficient; n =1,2, \ 8230;, N;
s3: calculating the elevation abnormity calculation values of the N GPS level joint measurement points by utilizing the quadratic polynomial model obtained in the step S2Calculating the fitting residual:
s4: constructing a wavelet neural network, wherein the input of the wavelet neural network is a planeCoordinate (X, Y) and elevation anomaly calculated value xi 1 Outputting a residual error delta xi;
connecting the plane coordinates (x) of N GPS level measuring points n ,y n ) Elevation anomaly calculation valueAs input, Δ ξ n Taking the normalized value as output, training the wavelet neural network to obtain an elevation anomaly residual error model;
s5: for the plane coordinate in the target area as (x) t ,y t ) According to the quadratic polynomial model fitted in S2, calculating the abnormal elevation calculation value xi t (ii) a Will (x) t ,y t ) And xi t Inputting the normalized residual error into an S4 trained elevation anomaly residual error model to obtain a normalized residual error of a to-be-measured area, and performing inverse normalization to obtain a residual error delta xi t And then the elevation abnormal value of the to-be-measured land is as follows:
in the step S2, a least square method is adopted to determine a fitting coefficient a 0 ,a 1 ,…,a 5 。
The wavelet neural network established in the step S4 comprises an input layer, a hidden layer and an output layer which are sequentially connected;
the input layer comprises 3 input units, the hidden layer comprises 15 hidden units, and the output layer comprises 1 output unit;
the activation function of the hidden layer is a wavelet basis function, the weight of the hidden layer is a translation factor of the wavelet basis function, and the threshold of the hidden layer is a scaling factor of the wavelet basis function.
and S4, training the wavelet neural network by adopting a gradient descent method of error back propagation to obtain an elevation anomaly residual error model.
Has the advantages that: the elevation abnormal value calculation method disclosed by the invention adopts the wavelet neural network to establish a high-precision elevation abnormal residual error model, and can accurately calculate the residual error of the elevation abnormal value for the position points in the large-area measurement area, thereby obtaining more accurate elevation abnormal value and realizing accurate conversion between the geodetic height and the normal height of the GPS.
Drawings
FIG. 1 is a flow chart of a method for calculating an elevation anomaly according to the present disclosure;
FIG. 2 is a distribution diagram of leveling co-measurements in an embodiment;
FIG. 3 is a diagram showing error distributions of fitting results of two methods in the example.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention are described below with reference to the accompanying drawings.
As shown in fig. 1, the present invention discloses an elevation abnormal value calculation method, including:
s1: acquiring elevation abnormal values of a plurality of GPS level connection points in a target area, pre-fitting the acquired data by adopting a quadric surface method, taking an error in three times as a limit difference, and removing gross errors;
in this embodiment, taking a Jiangsu area as an example, 171 level-C GPS network level joint measurement points in Jiangsu province are adopted, and three leveling measurements are adopted to obtain an elevation abnormal value of each level joint measurement point. 171 point positions are uniformly distributed on 100000km in Jiangsu province 2 The average density of leveling joint points is about: every 1000km 2 The number of level connection points is equal to 1.7. Jiangsu province belongs to plain areas, and the maximum and minimum values of the elevation abnormality are respectively as follows: xi shape max =12.5377m,ξ min =-8.2549m。
Of the 171 data, 131 data were selected as modeling sample data, and the other 40 data were selected as verification samples, and the distribution is as shown in fig. 2. And (3) carrying out quadratic surface method pre-fitting on the elevation abnormal data of all 131 modeling samples to obtain a curve fitting model with the error of +/-43.3 cm, and taking the error of 3 times as a tolerance, wherein the tolerance is +/-129.9 cm. According to the result, fitting errors of 4 points are all larger than a limit difference, so that the 4 points are considered as gross errors, and are removed;
s2: eliminating the height abnormal values of the N GPS level combined measuring points after the gross errors are eliminated from S1Fitting by using a quadratic polynomial:
ξ(x,y)=a 0 +a 1 x+a 2 y+a 3 x 2 +a 4 xy+a 5 y 2
where ξ (x, y) is the elevation anomaly at planar coordinates (x, y); a is 0 ,a 1 ,…,a 5 Is a fitting coefficient; n =1,2, \ 8230;, N;
and performing quadric surface fitting on the remaining N =127 elevation abnormal data. And the error in the overall fitting is +/-24.6 cm, and taking the error in the triple fitting as a limit difference, all point positions meet the limit difference requirement, so that all gross errors are considered to be eliminated.
The present embodiment determines the fitting coefficient a by using the least square method 0 ,a 1 ,…,a 5 。
S3: calculating the elevation abnormity calculation values of the N GPS level joint measurement points by utilizing the quadratic polynomial model obtained in the step S2Calculating the fitting residual:
s4: constructing a wavelet neural network, wherein the input of the wavelet neural network is plane coordinates (X, Y) and an elevation anomaly calculation value xi 1 Outputting a residual error delta xi;
the wavelet neural network has two structures: the method is a loose structure, namely, wavelet analysis is carried out on data firstly, and a result is input into a neural network to be used as an input layer; secondly, a compact structure is adopted, the wavelet basis function replaces an activation function in a neural network hidden layer structure, the translation and expansion factors of the wavelet basis function replace the weight and the threshold of a hidden layer, and structural realization is more advancedAnd (4) tight fusion. The invention adopts a compact wavelet neural network, which comprises an input layer, a hidden layer and an output layer which are connected in sequence; the input layer comprises 3 input units, the hidden layer comprises 15 hidden units, and the output layer comprises 1 output unit; the wavelet basis function is Morlet wavelet, and the expression is as follows:
connecting the plane coordinates (x) of N GPS level measuring points n ,y n ) And elevation anomaly calculation valueAs input, Δ ξ n Taking the normalized value as output, training the wavelet neural network to obtain an elevation anomaly residual error model;
the normalization in this embodiment adopts the following calculation formula:
wherein R is an initial value before normalization, R max ,R min Respectively, the maximum value and the minimum value of the initial data, T is the normalized value, T max ,T min Respectively the maximum and minimum values after the normalized transformation. In this example, T max The value of T is 0.9 min The value is 0.1.
The inverse normalization is calculated as follows:
and training the wavelet neural network by adopting a gradient descent method of error back propagation to obtain an elevation abnormal residual error model.
S5: for the plane coordinate in the target area as (x) t ,y t ) According to the quadratic polynomial model of S2 fitting, calculating the elevation anomaly meterCalculated value xi t (ii) a Will (x) t ,y t ) And xi t Inputting the normalized data into an S4 trained elevation anomaly residual error model to obtain a normalized residual error of a place to be measured, and performing inverse normalization to obtain a residual error delta xi t And then the elevation abnormal value of the to-be-measured land is as follows:
in order to compare the effects of the method disclosed by the invention, the elevation abnormal values of the 40 verification sample data are calculated by respectively adopting a quadric surface fitting method and the method disclosed by the invention and are compared with the elevation abnormal values measured at the 40 verification samples. The results of the two methods are shown in table 1.
As can be seen from Table 1, when elevation anomaly fitting modeling is performed in a large-area (provincial area), fitting is performed by a traditional quadric surface fitting method, and the obtained precision is +/-23.1 cm. The accuracy of the elevation abnormal value calculation method provided by the invention is +/-8.6 cm, and is improved by 62.8% compared with the accuracy of the traditional quadric surface model.
Fig. 3 shows the error distribution of the fitting results of the two methods in this embodiment. As can be seen from FIG. 3, the elevation abnormal value calculation method proposed by the present invention has a larger improvement at each inspection point compared to the quadric surface method specified in the engineering survey Specification
TABLE 1 comparison of fitting residuals
Claims (5)
1. An elevation anomaly calculation method, comprising:
s1: acquiring elevation abnormal values of a plurality of GPS level connection points in a target area, pre-fitting the acquired data by adopting a quadric surface method, taking an error in three times as a limit difference, and removing gross errors;
s2: eliminating the height abnormal values of the N GPS level combined measuring points after the gross errors are eliminated from S1Fitting by using a quadratic polynomial:
ξ(x,y)=a 0 +a 1 x+a 2 y+a 3 x 2 +a 4 xy+a 5 y 2
where ξ (x, y) is the elevation anomaly at planar coordinates (x, y); a is 0 ,a 1 ,...,a 5 Is a fitting coefficient; n =1,2, \ 8230;, N;
s3: calculating the abnormal elevation calculation values of the N GPS level joint measuring points by using the quadratic polynomial model obtained in the step S2Calculating the fitting residual:
s4: constructing a wavelet neural network, wherein the input of the wavelet neural network is plane coordinates (X, Y) and an elevation anomaly calculation value xi 1 Outputting a residual error delta xi;
connecting the plane coordinates (x) of N GPS level measuring points n ,y n ) Elevation anomaly calculation valueAs an input, Δ ξ n Taking the normalized value as output, training the wavelet neural network to obtain an elevation anomaly residual error model;
s5: for the plane coordinate in the target area as (x) t ,y t ) According to the quadratic polynomial model fitted in S2, calculating the abnormal elevation calculation value xi t (ii) a Will (x) t ,y t ) And xi t Inputting the normalized residual error into an S4 trained elevation anomaly residual error model to obtain a normalized residual error of a to-be-measured area, and performing inverse normalization to obtain a residual error delta xi t And if the height abnormal value of the land to be measured is as follows:
2. the elevation abnormal value calculation method according to claim 1, wherein the fitting coefficient a is determined by a least square method in the step S2 0 ,a 1 ,...,a 5 。
3. The elevation outlier calculating method of claim 1, wherein the wavelet neural network established in step S4 comprises an input layer, a hidden layer and an output layer which are connected in sequence;
the input layer comprises 3 input cells, the hidden layer comprises 15 hidden cells, and the output layer comprises 1 output cell;
the activation function of the hidden layer is a wavelet basis function, the weight of the hidden layer is a translation factor of the wavelet basis function, and the threshold of the hidden layer is a scaling factor of the wavelet basis function.
5. the elevation abnormal value calculation method according to claim 1, wherein in the step S4, the wavelet neural network is trained by a gradient descent method of error back propagation to obtain an elevation abnormal residual error model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010117015.1A CN111274738B (en) | 2020-02-25 | 2020-02-25 | Elevation abnormal value calculation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010117015.1A CN111274738B (en) | 2020-02-25 | 2020-02-25 | Elevation abnormal value calculation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111274738A CN111274738A (en) | 2020-06-12 |
CN111274738B true CN111274738B (en) | 2022-11-25 |
Family
ID=71002402
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010117015.1A Active CN111274738B (en) | 2020-02-25 | 2020-02-25 | Elevation abnormal value calculation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111274738B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112731453B (en) * | 2020-12-21 | 2022-03-01 | 自然资源部第一海洋研究所 | Vertical reference detection method for tide station by utilizing GNSS buoy |
CN112946691B (en) * | 2021-02-06 | 2024-03-12 | 长江水利委员会水文局长江上游水文水资源勘测局 | Band-shaped region coordinate conversion segmentation method considering elevation anomaly trend change |
CN113111573B (en) * | 2021-03-24 | 2022-09-23 | 桂林电子科技大学 | Landslide displacement prediction method based on GRU |
CN113358092B (en) * | 2021-06-10 | 2023-01-13 | 国家基础地理信息中心 | Big data numerical algorithm for determining vertical deviation of national elevation standard |
CN113532397B (en) * | 2021-07-07 | 2022-07-15 | 天津大学 | Regional elevation anomaly fitting method based on expansion algorithm |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101319893A (en) * | 2008-07-18 | 2008-12-10 | 东南大学 | Method for accurately determining region height anomaly |
-
2020
- 2020-02-25 CN CN202010117015.1A patent/CN111274738B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101319893A (en) * | 2008-07-18 | 2008-12-10 | 东南大学 | Method for accurately determining region height anomaly |
Non-Patent Citations (2)
Title |
---|
基于RBF神经网络的GPS高程异常拟合;席军峰;《测绘标准化》;20130915(第03期);全文 * |
多面函数拟合法转换GPS高程;金时华;《测绘与空间地理信息》;20051230(第06期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN111274738A (en) | 2020-06-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111274738B (en) | Elevation abnormal value calculation method | |
CN113532397B (en) | Regional elevation anomaly fitting method based on expansion algorithm | |
CN110516350A (en) | A kind of ERS point tolerance modification method based on anisotropy weighting | |
CN109521444B (en) | Self-adaptive least square fitting estimation algorithm for GPS horizontal velocity field of crustal movement | |
CN112146663B (en) | GNSS satellite precision orbit determination enhancing method and system based on bias satellite-borne accelerometer | |
CN101957193B (en) | Optimization method for sea island reef height transmission | |
CN110345867A (en) | A method of the space CMM Point Coordinates error is corrected based on inverse distance weighting | |
CN112131752B (en) | Super-collapse pollution rate tolerance estimation algorithm based on quasi-calibration | |
CN115238327B (en) | Three-dimensional space methane distribution pattern simulation method based on satellite monitoring | |
CN111765902A (en) | Laser point cloud precision evaluation method based on polygonal pyramid target | |
CN114993604B (en) | Wind tunnel balance static calibration and measurement method based on deep learning | |
CN113281748B (en) | Surface deformation monitoring method | |
CN113587921B (en) | Gravity gradient field and gravity anomaly field submersible vehicle fusion positioning method and system | |
CN111538943B (en) | Novel high-space-time resolution global ZTD vertical section grid model construction method | |
CN113639893B (en) | Near-earth weighted average temperature information acquisition method based on multiple meteorological factors | |
CN115033941A (en) | Method for calculating contact area of three-dimensional rough surface | |
CN113947628A (en) | Method for calculating silt charging amount based on river channel topographic maps with different scales | |
CN109960776B (en) | Improved method for hydraulic travel time and hydraulic signal attenuation inversion calculation | |
CN113899344B (en) | High-precision settlement monitoring system and method for long tunnel considering temperature effect | |
CN113916181B (en) | Data processing method of surface-internal integrated deformation monitoring device | |
CN111522071A (en) | Bulk deposition layer density measuring method based on bulk sample method | |
CN201876879U (en) | Processing system having improved accuracy of spatial data locations | |
Zaletnyik et al. | Combination of GPS/leveling and the gravimetric geoid by using the thin plate spline interpolation technique via finite element method | |
CN116070533B (en) | Neutron energy spectrum determination method | |
CN102222058A (en) | Total-least-square-based spatial data accuracy improvement method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |