A full-scene positioning method and system for railway detection equipment

Abstract

The application discloses a kind of full scene positioning method and system for railway detection equipment, method includes: S100, environment characteristic quantization: environment evaluation index is constructed, and environment index vector is formed;S200, sensor performance modeling: the dynamic correlation model matrix of sensor performance and each environment characteristic is established, and multiple positioning technologies are distributed based on weight;S300, dynamic fusion decision: based on real-time environment characteristic, the dynamic correlation model matrix of sensor performance and each environment characteristic, optimize sensor weight combination, realize the fusion of multiple positioning technologies.By constructing four-dimensional environment evaluation system including the sky visibility, illumination intensity, track geometry complexity, track reference information completeness, the dynamic correlation matrix of positioning technology-environment characteristic is combined with dynamically quantized environment characteristic vector, to realize the dynamic distribution of multiple positioning technology weight.Improve environmental adaptability to cope with the complexity and diversity of railway scene.

Description

Technical Field

[0001] This invention belongs to the field of railway inspection and positioning technology, and more specifically, relates to a full-scene positioning method and system for railway inspection equipment. Background Technology

[0002] Precise positioning of railway inspection equipment is a core requirement for ensuring the safe operation and maintenance of railways. During dynamic inspections, such as identifying track irregularities, catenary geometry defects, or power supply facility faults, the inspection data must be strictly correlated with specific locations (line number, row number, kilometer marker). Otherwise, it cannot guide on-site repairs. For example, if onboard inspection equipment cannot obtain the line name, even if abnormal data is detected, maintenance personnel cannot locate the fault, rendering the data unusable and severely impacting railway maintenance efficiency and traffic safety.

[0003] Currently, positioning is mainly achieved through the following methods: 1. Satellite positioning (GPS / BeiDou): Latitude and longitude are obtained through an onboard satellite module, and then matched with a pre-collected railway electronic map to convert the mileage. This map requires manual collection of coordinate points along the railway line to establish a correspondence. 2. Base station signal positioning: A communication module is embedded in the device, and the approximate position of the train is determined by matching the numbers of nearby base stations with a self-built mileage database. 3. Speed ​​sensor and mileage accumulation: Displacement is calculated using wheel axle encoders, and the position is estimated by combining it with a line database (such as pole number or span information). 4. Track circuit and transponder: A circuit system is constructed using the rails as conductors. The occupied section is determined by the shunt effect of the train wheelsets, or the positioning signal is triggered by a ground transponder.

[0004] While all of the above-mentioned existing technologies can achieve a certain positioning effect, satellite positioning suffers from signal interruption or increased error in tunnels, mountainous areas, or in severe weather; base station positioning is completely ineffective in areas without signal (Gobi Desert, tunnels), and the sparse base stations along railway lines lead to poor accuracy; speed sensors are prone to idling due to wheel slippage and wheel wear, resulting in cumulative mileage errors; track circuits and transponders require complex ground equipment, resulting in high construction and maintenance costs, and can only achieve segment-level positioning. Furthermore, the above positioning methods generally suffer from the following problems: 1. Limitations of single-modal perception: GNSS systems are prone to failure in obstructed environments such as tunnels and canyons; visual SLAM accuracy drops sharply in low light and rainy / foggy weather; inertial navigation systems experience error accumulation over time; 2. Defects in static fusion strategies: Traditional methods such as Kalman filtering and particle filtering use fixed sensor combination weights, failing to consider the impact of dynamic environmental changes on sensor positioning and attitude determination results. Preset weights cannot adapt to dynamic environmental changes, resulting in insufficient position and attitude accuracy after fusion; 3. Insufficient scene adaptability: Existing methods do not establish a mapping relationship between environmental features and positioning technology, and cannot automatically adjust algorithm strategies based on factors such as track geometry and line-of-sight conditions. Hard decision switching between different positioning modules causes pose jumps, failing to meet the requirement of seamless switching; 4. Poor system scalability: Current multi-sensor fusion algorithms are fixed in form, and adding new sensors requires reconstructing the fusion algorithm, resulting in a long development cycle. Therefore, a full-scene positioning method and system for railway inspection equipment is needed to improve environmental adaptability and cope with the complexity and diversity of railway scenarios. Summary of the Invention

[0005] To address the aforementioned shortcomings or improvement needs of existing technologies, this invention provides a full-scene positioning method and system for railway inspection equipment. Addressing the problems of poor scene adaptability and rigid algorithms inherent in existing single technologies, it proposes an intelligent fusion framework based on dynamic quantification of environmental features and dynamic weighting of positioning technologies. This method constructs a four-dimensional environmental evaluation system encompassing visibility, illumination intensity, track geometric complexity, and the completeness of track reference information. It combines dynamically quantified environmental feature vectors with the correlation matrix between positioning technologies and environmental features to achieve dynamic weight allocation for various positioning technologies. This enhances environmental adaptability to cope with the complexity and diversity of railway scenarios.

[0006] To achieve the above objectives, according to a first aspect of the present invention, a full-scene positioning method for railway inspection equipment is provided, specifically including the following steps:

[0007] S100. Environmental characteristic quantification: Constructing environmental assessment indicators and forming an environmental indicator vector;

[0008] S200, Sensor Performance Modeling: Establish a dynamic correlation model matrix between sensor performance and various environmental characteristics, and assign weights to multiple positioning technologies;

[0009] S300, Dynamic Fusion Decision: Based on the dynamic correlation model matrix of real-time environmental characteristics, sensor performance and various environmental characteristics, optimize the combination of sensor weights to achieve the fusion of multiple positioning technologies.

[0010] Furthermore, in step S100, the environmental evaluation indicators include: visibility of the sky, lighting conditions, geometric complexity, and completeness of benchmark data.

[0011] The environmental indicator vector Q n×1 for:

[0012] Q n×1 =[SGCD] T

[0013] Where S represents the state of visibility to the sky, 0≤S≤1.

[0014] G is the light intensity index, where 0 ≤ G ≤ 1.

[0015] C represents the geometric complexity of the orbital path, where 0 ≤ C ≤ 1.

[0016] D represents the completeness of the baseline data, where 0 ≤ D ≤ 1.

[0017] Furthermore, when determining the GNSS visibility status S, it is necessary to perform multi-band GNSS signal fusion processing, signal masking physical model enhancement processing, and finally signal model integration.

[0018] When performing multi-band GNSS signal fusion processing, the influence of satellite elevation angle needs to be considered. Satellites with low elevation angles have poor signal quality, which can be addressed by using an elevation angle compensation factor w. θ The elevation angle compensation factor w is subjected to weight reduction processing. θ for:

[0019]

[0020] Where, θ k Let be the elevation angle of the k-th satellite;

[0021] For the k-th satellite, its effective signal-to-noise ratio is adjusted by both the frequency band weight and the elevation angle compensation factor, specifically as follows:

[0022] SNR k，eff =w band,k ×SNR k ×w θ

[0023] Among them, SNR k，eff The effective signal-to-noise ratio of the k-th satellite is

[0024] w band，k The weight of the frequency band where the k-th satellite is located.

[0025] SNR k Let be the original signal-to-noise ratio of the k-th satellite;

[0026] By fusing the effective signal-to-noise ratios of all visible satellites and dynamically weighting them based on the carrier noise density ratio, the comprehensive effective signal-to-noise ratio S is obtained. SNR Specifically:

[0027]

[0028] Among them, w k Let be the signal-to-noise ratio weight of the k-th satellite, and

[0029] C / N0 k denoted as the carrier noise density ratio of the k-th satellite.

[0030] Furthermore, when performing signal masking physical model enhancement processing, it is necessary to dynamically adjust the critical masking angle θ0 according to the environment;

[0031] The initial sky-penetrating state S is determined based on the critical obscuration angle θ0 and the transition region angle Δθ. base Specifically:

[0032]

[0033] Where θ is the minimum elevation angle threshold for visible satellites.

[0034] The synthesis of the signal model needs to be combined with the comprehensive effective signal-to-noise ratio S. SNR and the initial sky-penetrating state S base Specifically:

[0035] S0=α·S base +(1-α)·S SNR

[0036] Among them, S0 represents the overall visibility status of the SkyNet system.

[0037] α is the sky visibility coefficient.

[0038] Furthermore, when determining the sky-penetrating state S, it is necessary to obtain the final sky-penetrating state S through normalization processing.

[0039] The aforementioned sky-penetrating state S is:

[0040]

[0041] Where ∈ is the control coefficient.

[0042] S min The minimum value of S0,

[0043] S max This is the maximum value of S0.

[0044] Furthermore, in determining the illumination intensity index G, multispectral sensing fusion and satellite signal attenuation compensation are required to obtain the initial illumination state G. cam Tunnel distance attenuation coefficient A tunnel and satellite availability characterization V sat ;

[0045] After multi-source fusion processing, the basic illumination index G0 is obtained, specifically:

[0046] G0=γ·V sat ·A tunnel +(1-γ)·G cam

[0047] Where γ is the light intensity coefficient;

[0048] When determining the illuminance index G, a normalization process is also required to obtain the final illuminance index G, which is specifically:

[0049]

[0050] Among them, G min The minimum value of G0,

[0051] G max It is the maximum value of G0.

[0052] Furthermore, in determining the geometric complexity C of the orbit, it is necessary to first divide the orbit into a grid to obtain its basic information entropy H. base Curvature characteristics C curve and elevation change Δh std ;

[0053] Then the basic information entropy H of the orbit base Curvature characteristics C curve and elevation change Δh std The fusion yields the comprehensive orbital geometric complexity C0, which is:

[0054] C0 = v1·H base +v2·C curve +v3·△h std

[0055] Where v1 is the weight of the basic information entropy,

[0056] v2 represents the weight of the curvature feature.

[0057] v3 represents the weight of the elevation change feature;

[0058] When determining the orbital geometric complexity C, a normalization process is also required to obtain the final orbital geometric complexity C, which is specifically:

[0059]

[0060] Among them, C min The minimum value of C0,

[0061] C max This is the maximum value of C0.

[0062] Furthermore, when determining the baseline data completeness D, the preliminary data completeness D0 is first calculated using the baseline data provided by the track map, specifically as follows:

[0063]

[0064] Where Δx is the deviation between the current position of the detection device and the digital track map in the x-direction.

[0065] Δy represents the deviation between the current position of the detection device and the digital track map in the y-direction.

[0066] D threshold The maximum allowable matching deviation threshold is adjusted according to the track detection accuracy requirements.

[0067] Then, based on the initial data completeness D0, and taking the maximum value D of D0 throughout the entire positioning process. max and minimum value D min Normalizing D0 yields the baseline data completeness D, specifically:

[0068]

[0069] Furthermore, in step S200, it is necessary to establish a correlation matrix R between positioning technology and environmental indicators. m×n m is the number of sensor types, and n is the number of environmental indicators;

[0070] In step S300, specifically: based on the correlation matrix R between positioning technology and environmental indicators... m×n and environmental indicator vector Q n×1 Generate a weight vector b that adaptively fuses multiple localization methods m×1 Specifically:

[0071] b m×1 =R m×n ×Q m×1

[0072] Then, the multi-source localization results are fused to obtain the final three-dimensional localization coordinates P of the detection device.s Specifically:

[0073]

[0074] Where, λ k The weights for adaptive fusion of multiple positioning methods corresponding to the k-th positioning method.

[0075] P k (s) is the coordinate value in the s direction calculated by the k-th positioning method, where s is x, y, or z.

[0076] According to a second aspect of the present invention, a full-scene positioning system for railway inspection equipment is provided, comprising:

[0077] Feature quantification module: used for environmental feature quantification, constructing environmental assessment indicators, and forming environmental indicator vectors;

[0078] Performance Modeling Module: Sensor performance modeling, establishing a dynamic correlation model matrix between sensor performance and various environmental characteristics, and assigning weights to multiple positioning technologies;

[0079] Dynamic Fusion Module: Dynamic fusion decision-making, based on the dynamic correlation model matrix of real-time environmental characteristics, sensor performance and various environmental characteristics, optimizes the combination of sensor weights to achieve the fusion of multiple positioning technologies.

[0080] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0081] 1. The railway inspection equipment full-scene positioning method of the present invention addresses the problems of poor scene adaptability and algorithm rigidity in existing single technologies by proposing an intelligent fusion framework based on dynamic quantification of environmental features and dynamic weighting of positioning technologies. This method constructs a four-dimensional environmental evaluation system including visibility, illumination intensity, track geometric complexity, and track reference information completeness. It combines dynamically quantified environmental feature vectors and the correlation matrix between positioning technologies and environmental features to achieve dynamic weight allocation of multiple positioning technologies. This enhances environmental adaptability to cope with the complexity and diversity of railway scenarios.

[0082] 2. The railway inspection equipment full-scene positioning method of the present invention comprises three core layers: an environmental feature perception layer, a sensor performance evaluation layer, and a dynamic fusion decision layer. It supports the expansion and deletion of various positioning technologies and environmental evaluation indicators. Compared with fixed fusion schemes, this method has significant advantages in terms of positioning continuity, environmental adaptability, and scalability.

[0083] 3. Compared with the fixed fusion strategy of traditional positioning methods, the railway inspection equipment full-scene positioning method of the present invention can adjust the weight of different positioning technologies according to the actual environmental characteristics. The positioning technology with greater advantages contributes the most to the final positioning result, which is conducive to improving the accuracy of the positioning result.

[0084] 4. The railway inspection equipment full-scene positioning method of the present invention adjusts the correlation coefficient matrix of positioning technology and environmental indicators according to the actual situation. In actual navigation, if there is no corresponding technology, the corresponding row can be directly deleted from the matrix. If a corresponding technology is added, only the corresponding row needs to be added. If there is a new indicator, only the corresponding column needs to be added. Similarly, the corresponding column can be deleted as needed, which greatly increases the scalability of the system. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating a full-scene positioning method for railway inspection equipment according to an embodiment of the present invention;

[0086] Figure 2 This is a data diagram illustrating the correlation matrix of positioning technology and environmental indicators for a full-scene positioning method for railway inspection equipment according to an embodiment of the present invention. Detailed Implementation

[0087] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0088] Example 1

[0089] like Figure 1 As shown, this embodiment of the invention provides a full-scene positioning method for railway inspection equipment, specifically including the following steps:

[0090] S100. Environmental characteristic quantification: Constructing environmental assessment indicators and forming an environmental indicator vector;

[0091] S200, Sensor Performance Modeling: Establish a dynamic correlation model matrix between sensor performance and various environmental characteristics, and assign weights to multiple positioning technologies;

[0092] S300, Dynamic Fusion Decision: Based on the dynamic correlation model matrix of real-time environmental characteristics, sensor performance and various environmental characteristics, optimize the combination of sensor weights to achieve the fusion of multiple positioning technologies.

[0093] In step S100, the environmental evaluation indicators include: visibility to the sky, illumination conditions, geometric complexity, and completeness of baseline data. The environmental indicator vector Q... n×1 for:

[0094] Q n×1 =[SGCD] T

[0095] Where S represents the state of visibility to the sky, 0≤S≤1.

[0096] G is the light intensity index, where 0 ≤ G ≤ 1.

[0097] C represents the geometric complexity of the orbital path, where 0 ≤ C ≤ 1.

[0098] D represents the completeness of the baseline data, where 0 ≤ D ≤ 1.

[0099] In determining the GNSS visibility state S, it is necessary to perform multi-band GNSS signal fusion processing, signal masking physical model enhancement processing, and finally signal model integration.

[0100] When performing multi-band GNSS signal fusion processing, the influence of satellite elevation angle needs to be considered. Satellites with low elevation angles have poor signal quality, which can be addressed by using an elevation angle compensation factor w. θ The elevation angle compensation factor w is subjected to weight reduction processing. θ for:

[0101]

[0102] Where, θ k Let be the elevation angle of the k-th satellite.

[0103] For the k-th satellite, its effective signal-to-noise ratio is adjusted by both the frequency band weight and the elevation angle compensation factor, specifically as follows:

[0104] SNR k，eff =w band,k ×SNR k ×w θ

[0105] Among them, SNR k，eff The effective signal-to-noise ratio of the k-th satellite is

[0106] w band，k The weight of the frequency band where the k-th satellite is located.

[0107] SNR k is the original signal-to-noise ratio of the k-th satellite.

[0108] By fusing the effective signal-to-noise ratios of all visible satellites and dynamically weighting them based on the carrier noise density ratio, the comprehensive effective signal-to-noise ratio S is obtained. SNRSpecifically:

[0109]

[0110] Among them, w k Let be the signal-to-noise ratio weight of the k-th satellite, and

[0111] C / N0 k Let be the carrier noise density ratio of the k-th satellite. Satellites with a noise density lower than 30 dB-Hz are considered invalid signals and their weights are set to zero.

[0112] When performing signal masking physical model enhancement processing, the critical masking angle θ0 needs to be dynamically adjusted according to the environment. The adjustment method is as follows: set the initial value of the critical masking angle θ0; if the number of effective satellites has been less than 5 for the past minute, then θ0 is increased by 5°; if it is in an urban canyon environment, then θ0 is decreased by 3°. The transition area angle Δθ is dynamically determined according to the scene type, specifically: tunnel entrance or exit area: Δθ = 15°; normal open environment: Δθ = 8°; dense tree area: Δθ = 25°.

[0113] The initial sky-penetrating state S is determined based on the critical obscuration angle θ0 and the transition region angle Δθ. base Specifically:

[0114]

[0115] Where θ is the minimum elevation angle threshold for visible satellites.

[0116] The synthesis of the signal model needs to be combined with the comprehensive effective signal-to-noise ratio S. SNR and the initial sky-penetrating state S base Specifically:

[0117] S0=α·S base +(1-α)·S SNR

[0118] Among them, S0 represents the overall visibility status of the SkyNet system.

[0119] α is the sky visibility coefficient.

[0120] When determining the line-of-sight status S, normalization processing is also required to obtain the final line-of-sight status S. Specifically, the statistical window length is dynamically adjusted according to the scene change rate. The S0 sequence is collected within the window, the 5% and 95% quantiles of the sequence are calculated, outliers less than the 5% or greater than the 95% quantile are removed, and the minimum value S is taken from the remaining valid data. min and maximum value S max Then, normalization is performed to obtain the sky-viewing state S. The sky-viewing state S is:

[0121]

[0122] Where ∈ is the control coefficient, ∈=1×10 -6 .

[0123] When determining the light intensity index G, multispectral sensing fusion, satellite signal attenuation compensation, and multi-source fusion processing are required.

[0124] When performing multispectral sensing fusion, it is necessary to quantize the visible light channels to obtain the illumination complexity and extract the thermal radiation intensity through infrared channel feature extraction. The illumination complexity G is... vis for:

[0125]

[0126] Where, p ij The probability distribution of brightness.

[0127] Determine the luminance distribution probability p ij First, the image is divided into a 16×16 grid, and the average brightness L of each grid is calculated. ij The average brightness L of each grid ij Determine the luminance distribution probability p ij Specifically:

[0128]

[0129] The thermal radiation intensity G IR for:

[0130]

[0131] Among them, T obj This is the highest temperature in the orbital region.

[0132] T env The ambient average temperature.

[0133] Based on the illumination complexity G vis and thermal radiation intensity G IR The initial lighting state G is obtained by fusion. cam Specifically:

[0134] G cam =β·G vis +(1-β)·G IR

[0135] Where β is the initial illumination intensity coefficient.

[0136] The initial illumination intensity coefficient β is:

[0137]

[0138] When performing satellite signal attenuation compensation, it is necessary to determine the tunnel distance attenuation coefficient and satellite availability characterization.

[0139] The tunnel distance attenuation coefficient A tunnel for:

[0140] A tunnel =e -λd

[0141] Where λ is the attenuation coefficient.

[0142] d represents the distance from the tunnel entrance.

[0143] The satellite availability characterization V sat for:

[0144]

[0145] Where, N vis For the number of effective satellites,

[0146] N max This represents the theoretical maximum number of visible satellites.

[0147] After the multi-source fusion processing, the basic illumination index G0 is obtained, specifically:

[0148] G0=γ·V sat ·A tunnel +(1-γ)·G cam

[0149] Where γ is the light intensity coefficient.

[0150] When determining the illuminance index G, normalization is also required to obtain the final illuminance index G. Specifically, the statistical window length is dynamically adjusted according to the scene change rate, a G0 sequence is collected within the window, the 5th and 95th quantiles of the sequence are calculated, outliers less than the 5th or greater than the 95th quantile are removed, and the minimum value G is taken from the remaining valid data. min and maximum value G max Then, normalization is performed to obtain the light intensity index G. The light intensity index G is specifically:

[0151]

[0152] When determining the orbital geometric complexity C, it is necessary to first divide the orbit into grids to obtain its basic information entropy, curvature characteristics, and elevation changes. Then, the basic information entropy, curvature characteristics, and elevation changes of the orbit are integrated to obtain the comprehensive orbital geometric complexity.

[0153] The basic information entropy H basefor:

[0154]

[0155] Where, q i Let be the proportion of the i-th grid point, representing the proportion of each grid point in the total grid.

[0156] N is the total number of grid cells.

[0157] When determining the curvature characteristics of a track, a quadratic surface fitting is first performed on each neighborhood of the track, specifically as follows:

[0158] z = ax 2 +bxy+cy 2 +dx+ey+f

[0159] Where x, y, and z are the three-dimensional coordinates of the orbit.

[0160] a, b, c, d, and e are all coefficients.

[0161] f is a constant;

[0162] Based on the quadratic surface fitting equation, the curvature characteristic C of the track is obtained. curve Specifically:

[0163]

[0164] Where K is the number of neighborhood samples of the orbital.

[0165] The elevation change Δh of the orbit std for:

[0166]

[0167] Where M represents the number of elevation abrupt change points.

[0168] h j Let j be the elevation value of the j-th elevation change point on the track.

[0169] This represents the average elevation value of the points where elevation changes abruptly on the track.

[0170] The overall orbital geometric complexity C0 is:

[0171] C0 = v1·H base +v2·C curve +v3·△h std

[0172] Where v1 is the weight of the basic information entropy,

[0173] v2 represents the weight of the curvature feature.

[0174] v3 represents the weight of the elevation change feature.

[0175] When determining the track geometric complexity C, it is also necessary to extract the track's structural features and analyze the sleeper distribution, and then fuse them to obtain the track's structural features.

[0176] During the structural feature extraction, a standard rail section template T(x, y) needs to be established. The matching similarity of the rail sections is then determined based on this standard rail section template T(x, y). Specifically:

[0177]

[0178] Among them, P cloud Point cloud data for the track section;

[0179] And calculate orbital integrity I gauge Specifically:

[0180]

[0181] Where Δd is the deviation between the measured track gauge and the standard value.

[0182] The sleeper distribution analysis requires projecting a point cloud along the track direction and using Fourier transform to extract the dominant frequency f of the sleeper distribution. dominant The regularity R of sleeper distribution is obtained. sleeper ,

[0183]

[0184] Among them, f std This refers to the standard sleeper spacing.

[0185] The structural features of the track are as follows:

[0186] C struct =δ·S rail +η·I gauge +ζ·R sleeper

[0187] Where δ is the weight of the similarity of the track cross-section matching.

[0188] η is the weight for track integrity.

[0189] ζ is the weight of the regularity of sleeper distribution.

[0190] When determining the orbital geometric complexity C, normalization is also required to obtain the final orbital geometric complexity C. Specifically, the statistical window length is dynamically adjusted according to the scene change rate. Within the window, the C0 sequence is collected, the 5th and 95th quantiles of the sequence are calculated, outliers less than the 5th or greater than the 95th quantile are removed, and the minimum value C is taken from the remaining valid data. min and maximum value C max Then, normalization is performed to obtain the orbital geometric complexity C. Specifically, the orbital geometric complexity C is:

[0191]

[0192] When determining the baseline data completeness D, the preliminary data completeness D0 is first calculated using the baseline data provided by the track map, specifically as follows:

[0193]

[0194] Where Δx is the deviation between the current position of the detection device and the digital track map in the x-direction.

[0195] Δy represents the deviation between the current position of the detection device and the digital track map in the y-direction.

[0196] D threshold The maximum allowable matching deviation threshold is adjusted according to the track detection accuracy requirements.

[0197] Then, based on the initial data completeness D0, and taking the maximum value D of D0 throughout the entire positioning process. max and minimum value D min Normalizing D0 yields the baseline data completeness D, specifically:

[0198]

[0199] like Figure 2 As shown, step S200 specifically involves: establishing a correlation matrix R between positioning technology and environmental indicators. m×n , where m is the number of sensor types and n is the number of environmental indicators. Matrix R m×n element a in ij Let R represent the correlation coefficient between the positioning technology and environmental indicators in row i and column j, where i ≤ m and j ≤ n. In actual navigation, if no corresponding technology exists, it can be directly used in matrix R. m×n In this case, delete the corresponding rows. If a corresponding technology is added, simply add the corresponding rows. If a new indicator is added, simply add the corresponding column; similarly, you can delete corresponding columns as needed.

[0200] In step S300, specifically: based on the correlation matrix R between positioning technology and environmental indicators... m×nand environmental indicator vector Q n×1 Generate a weight vector b that adaptively fuses multiple localization methods m×1 Specifically:

[0201]

[0202] Then, the multi-source localization results are fused to obtain the final three-dimensional localization coordinates P of the detection device. s Specifically:

[0203]

[0204] Where, λ k The weights for adaptive fusion of multiple positioning methods corresponding to the k-th positioning method.

[0205] P k (s) is the coordinate value in the s direction calculated by the k-th positioning method, where s is x, y, or z.

[0206] Example 2

[0207] This invention provides a full-scene positioning system for railway inspection equipment, specifically including:

[0208] Feature quantification module: used for environmental feature quantification, constructing environmental assessment indicators, and forming environmental indicator vectors;

[0209] Performance Modeling Module: Sensor performance modeling, establishing a dynamic correlation model matrix between sensor performance and various environmental characteristics, and assigning weights to multiple positioning technologies;

[0210] Dynamic Fusion Module: Dynamic fusion decision-making, based on the dynamic correlation model matrix of real-time environmental characteristics, sensor performance and various environmental characteristics, optimizes the combination of sensor weights to achieve the fusion of multiple positioning technologies.

[0211] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A full-scene positioning method for a railway detection device, characterized in that, Specifically comprising the following steps: S100, environment feature quantification: constructing environment evaluation indexes to form an environment index vector; S200, sensor performance modeling: establishing a dynamic correlation model matrix of sensor performance and various environment features, and distributing various positioning technologies based on weights; S300, dynamic fusion decision: based on real-time environment features, sensor performance and the dynamic correlation model matrix of various environment features, optimizing sensor weight combination to realize the fusion of various positioning technologies; In step S100, the environment evaluation indexes include: sky visibility, illumination conditions, geometric complexity and reference data completeness; The environmental index vector is: ， wherein to the state of the sky view, , for the light intensity index, , for track geometry complexity, , to reference data completeness, ; In determining the state of the sky view Multi-band GNSS signal fusion processing, signal shielding physical model strengthening processing, and finally signal model synthesis need to be performed. When performing multi-frequency band GNSS signal fusion processing, the influence of satellite elevation angle needs to be considered. The signal quality of low elevation angle satellite is poor, and the elevation angle compensation factor is used for weight reduction processing, and the elevation angle compensation factor is: ， wherein, is the first elevation angle of the geostationary satellite; For the first For GEO satellites, the effective SNR is adjusted by both the frequency band weight and the elevation angle compensation factor, which is given by: ， wherein is the effective signal-to-noise ratio of the GEO satellite, is the effective signal-to-noise ratio of the GEO satellite, For the first weight of the frequency band in which the geostationary satellite is located, The first Original signal-to-noise ratio of the GEO satellite; The effective signal-to-noise ratio of all visible satellites is fused, and a comprehensive effective signal-to-noise ratio is obtained by dynamic weighting based on a carrier-to-noise density ratio , specifically: ， wherein is the signal-to-noise ratio weight for the first geostationary satellite, and , For the first carrier-to-noise density ratio of the GEO satellite; In determining the light intensity index , multispectral perception fusion, satellite signal attenuation compensation need to be carried out to obtain the initial light state , tunnel distance attenuation coefficient and satellite availability characterization ; Then, multi-source fusion processing is performed to obtain the basic illumination index , specifically: ， wherein is the light intensity coefficient; In determining the illumination intensity index , it is also necessary to obtain the final illumination intensity index by normalization processing , which is specifically: ， wherein is a minimum value, is a maximum value of the function f(x) for x in the interval [a, b]; and is a In determining the track geometry complexity , it is necessary to first obtain its basic information entropy , curvature characteristics and elevation mutation by meshing the track Then the basic information entropy of the orbit Curvature characteristics and elevation changes By merging, we obtain the comprehensive orbital geometric complexity. The comprehensive orbital geometric complexity for: ， wherein, a weight of the base information entropy, The weights of the curvature features, weight for elevation abruptness feature; In determining the track geometry complexity , it is also necessary to obtain the final track geometry complexity by normalization processing , which is specifically: ， wherein is the minimum value of is the maximum value of is the maximum value of When determining the reference data completeness , a preliminary data completeness is calculated by the reference data provided by the track map , specifically: ， wherein, to detect the deviation of the current position of the device from the digital track map in direction. To detect the deviation of the current position of the device from the digital track map in direction, is the maximum allowed matching deviation threshold, which is adjusted according to the track detection accuracy requirement; Then, based on the completeness of the preliminary data... and take the entire positioning process maximum value and minimum value ,right Normalization is performed to obtain the completeness of the baseline data. Specifically: 。 2. A full scene positioning method for railway inspection apparatus according to claim 1, characterized in that, When performing signal masking physical model reinforcement processing, it is necessary to dynamically adjust the critical masking angle according to the environment ; According to the critical obscuring angle and the transition region angle Determining an initial pair sky visibility state Specifically: ， wherein, is the minimum elevation threshold for visible satellites; The synthesis of the signal model needs to be combined with a synthesized effective signal-to-noise ratio and an initial pair-to-sky visibility state , in particular: ， wherein to synthesize the state of the sky view, The atmospheric transmission coefficient.

3. A full scene positioning method for railway inspection apparatus according to claim 2, characterized in that, determining the state of sky visibility It is also necessary to obtain the final state of sky visibility by normalization ; The pair of antennas Is: ， wherein is a regulatory coefficient, is a minimum value of is a minimum value of is a maximum value of is a maximum value of 4. A full scene positioning method for railway inspection apparatuses according to any one of claims 1-3, characterized in that, In step S200, a positioning technology-environment index correlation matrix needs to be established , is the number of sensor types, is the number of environment indexes; In step S300, specifically, the positioning technology-environment index correlation matrix is generated based on the positioning technology-environment index correlation matrix and the environment index vector to generate a weight vector of adaptive fusion of multiple positioning modes , specifically: ， Then the multi-source positioning result fusion is performed to obtain the three-dimensional positioning coordinates of the final detection device Specifically, ， wherein, is the first weight corresponding to the adaptive fusion of the plurality of positioning modes of the positioning method, The coordinate values of the direction calculated for the first positioning method, The coordinate values of the direction calculated for the second positioning method, The coordinate values of the direction calculated for the third positioning method, The coordinate values of the direction calculated for the fourth positioning method, The coordinate values of the direction calculated for the fifth positioning method, The coordinate values of the direction calculated for the sixth positioning method, The coordinate values of the 5. A full scene positioning system for a railway inspection apparatus for implementing a full scene positioning method for a railway inspection apparatus according to any one of claims 1 to 4, characterized in that, Comprise: The feature quantification module is used for environment feature quantification, constructing environment evaluation indexes to form an environment index vector; The performance modeling module is used for sensor performance modeling, establishing a dynamic correlation model matrix of sensor performance and various environment features, and distributing various positioning technologies based on weights; The dynamic fusion module is used for dynamic fusion decision, based on real-time environment features, sensor performance and the dynamic correlation model matrix of various environment features, optimizing sensor weight combination to realize the fusion of various positioning technologies.

Images