A feature-based localization method, apparatus, and equipment in a ship environment

By selecting features with high importance and stability in the ship environment, a positioning model is constructed, which solves the problem of low positioning accuracy in the ship environment and achieves high accuracy and stable position prediction, adapting to complex ship environments.

CN120751479BActive Publication Date: 2026-11-17ZHEJIANG OCEAN UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202510800607.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-06-16
Publication Date
2026-11-17
Estimated Expiration
2045-06-16

AI Technical Summary

Technical Problem

In a marine environment, traditional wireless positioning methods suffer from low positioning accuracy due to factors such as electromagnetic reflection, refraction, and obstruction. Furthermore, dynamic disturbances increase the instability of the signal environment, making it difficult to guarantee high-precision positioning performance.

Method used

By acquiring multiple sets of location-related data, a gradient decision tree model is constructed to evaluate the importance and stability of features, select the target feature set for location, and train the location model based on these features to predict the location.

Benefits of technology

It improves the accuracy of positioning prediction, enhances the model's generalization ability in different cabins and environments, reduces prediction fluctuations, adapts to the needs of resource-constrained edge computing, and improves the real-time performance of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120751479B_ABST
    Figure CN120751479B_ABST
Patent Text Reader

Abstract

This application provides a feature-based positioning method, apparatus, and device for ship environments. The method includes: acquiring multiple sets of positioning-related data; constructing a first training sample set using IQ data sequences from each set of positioning-related data as input features and location as a label; training a gradient decision tree model using the first training sample set, and determining the importance score of each feature in the input features based on the training results after training; evaluating the overall stability score of each feature using the trained gradient decision tree model; determining a comprehensive score for each feature based on its importance score and overall stability score, and selecting a target feature set from the input features based on the comprehensive score; constructing a second training sample set based on the target feature set and the training sample set, and training the positioning model using the second training sample set; and finally, predicting the location corresponding to the data to be predicted using the positioning model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of wireless positioning technology, and in particular to a positioning method, apparatus and equipment based on feature screening in a ship environment. Background Technology

[0002] With the continuous development of the Internet of Things and wireless communication technologies, indoor positioning technology is being increasingly widely used in fields such as industrial manufacturing, smart logistics, and ship management. In the unique environment of ships, indoor positioning can be used to achieve functions such as personnel positioning, safety management, cargo tracking, and intelligent inspection, and has significant practical value.

[0003] However, compared to conventional building environments, the interior space of a ship exhibits significant structural differences and electromagnetic characteristics. Firstly, the hull structure generally utilizes large areas of metallic materials, resulting in high electromagnetic reflectivity. Secondly, the enclosed interior space, densely packed compartments, and complex layout cause strong reflections, refractions, and blockages of wireless signals during propagation, leading to severe multipath effects and frequency drift. Furthermore, dynamic disturbances such as deck vibrations and slight changes in the relative positions of walls during ship navigation further exacerbate the instability of the signal environment.

[0004] Given the aforementioned complex context, traditional positioning methods based on received signal strength, time delay, or phase are unable to guarantee high-precision positioning performance. Summary of the Invention

[0005] In view of this, this application provides a feature-based positioning method, apparatus, and device for accurate positioning in a ship environment.

[0006] Specifically, this application is implemented through the following technical solution:

[0007] The first aspect of this application provides a feature-based screening method in a ship environment, the method comprising:

[0008] Multiple sets of location-related data are acquired; wherein each set of location-related data includes an IQ data sequence acquired by the receiver during the process of receiving a Constant Tone Extended (CTE) signal transmitted from the transmitter, and the location of the transmitter;

[0009] Using the IQ data sequence in each set of location-related data as input features and the location in that set of location-related data as a label, a first training sample set is constructed using the multiple sets of location-related data.

[0010] The gradient decision tree model is trained using the first training sample set, and after training is completed, the importance score of each feature in the input features is determined based on the training results.

[0011] For each feature, the predictive ability of the feature under independent input conditions is evaluated using a trained gradient decision tree model to obtain the overall stability score of the feature; wherein, the overall stability score is used to characterize the degree to which the feature maintains its predictive accuracy at different positions;

[0012] Based on the importance score and overall stability score of each feature, a comprehensive score for each feature is determined, and based on the comprehensive score of each feature, a target feature set for localization is selected from the multiple features contained in the input feature.

[0013] Based on the target feature set and the first training sample set, a second training sample set is constructed for training the localization model, and the localization model is trained using the second training sample set.

[0014] After obtaining the data to be predicted, feature values ​​corresponding to the target feature set are extracted from the data to be predicted, and the extracted feature values ​​are input into the positioning model so that the positioning model can predict the location corresponding to the data to be predicted.

[0015] A second aspect of this application provides a feature-based positioning device in a ship environment, the device comprising an acquisition module, a construction module, a determination module, and a prediction module, wherein...

[0016] The acquisition module is used to acquire multiple sets of positioning-related data; wherein, each set of positioning-related data includes an IQ data sequence acquired by the receiver during the process of receiving a constant tone spread CTE signal transmitted from the transmitter, and the location of the transmitter;

[0017] The construction module is used to construct a first training sample set using the IQ data sequence in each set of location-related data as input features and the position in the set of location-related data as a label;

[0018] The determining module is used to train a gradient decision tree model using the first training sample set, and after training is completed, to determine the importance score of each feature in the input features based on the training results;

[0019] The determining module is used to evaluate the predictive ability of each feature under independent input conditions using a trained gradient decision tree model, and obtain the overall stability score of the feature; wherein, the overall stability score is used to characterize the degree to which the feature maintains its predictive accuracy at different positions.

[0020] The determining module is used to determine the comprehensive score of each feature based on the importance score and the overall stability score of each feature, and to select the target feature set for localization from the multiple features contained in the input feature based on the comprehensive score of each feature.

[0021] The construction module is used to construct a second training sample set for training the localization model based on the target feature set and the first training sample set, and to train the localization model with the second training sample set.

[0022] The prediction module is used to extract feature values ​​corresponding to the target feature set from the data to be predicted after obtaining the data to be predicted, and input the extracted feature values ​​into the positioning model so that the positioning model can predict the location corresponding to the data to be predicted.

[0023] A third aspect of this application provides a feature-based screening device for a marine environment, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods provided in the first aspect of this application.

[0024] The feature-based localization method, apparatus, and equipment provided in this application for ship environments combine feature importance scores and overall stability scores to eliminate irrelevant or unstable features while retaining key features with high discriminative power for localization tasks, effectively improving the accuracy of predicted location. Furthermore, when determining the overall stability score of a feature, the consistency of feature performance across different locations is comprehensively considered, which can improve the model's generalization ability in different compartments and environments and reduce prediction fluctuations. Moreover, feature selection reduces the input dimension, lowers the training and inference burden of the model, improves the real-time performance of the system, and adapts to the resource-constrained edge computing deployment requirements in ship environments. Attached Figure Description

[0025] Figure 1 A flowchart of an embodiment of the feature-based localization method in a ship environment provided in this application;

[0026] Figure 2 A schematic diagram of the structure of a feature-based positioning device in a ship environment provided in this application;

[0027] Figure 3 This is a hardware structure diagram of a feature-based positioning device in a ship environment, which is the location of the feature-based positioning device in the ship environment described in this application. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0029] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0030] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0031] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0032] Figure 1 This is a flowchart of an embodiment of the feature-based localization method for ship environments provided in this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0033] S101. Acquire multiple sets of positioning-related data; wherein, each set of positioning-related data includes the IQ data sequence acquired by the receiver during the process of receiving the constant tone spread CTE signal transmitted from the transmitter, and the location of the transmitter.

[0034] Specifically, the transmitters emitting the Constant Tone Extended CTE signal can be Bluetooth 5.1-based tags. These tags are installed inside the ship's cabins and periodically transmit the Constant Tone Extended CTE signal externally. It should be noted that the position coordinates of each transmitter are pre-set and known. Furthermore, the Constant Tone Extended CTE signal is obtained by appending a CTE segment to the end of the signal transmitted by the Bluetooth 5.1-based tag. The appended CTE segment is a continuous carrier signal with a constant frequency and constant modulation. Thus, the resulting Constant Tone Extended signal contains multiple sampling points, supporting high-frequency sampling and IQ data acquisition at the receiving end.

[0035] It should be noted that the IQ data sequence contains IQ samples collected at multiple sampling points, and each IQ sample includes an I component and a Q component; where the I component is the in-phase component and the Q component is the quadrature component.

[0036] Furthermore, the receiver is a receiving device that supports Bluetooth 5.1 and is equipped with a multi-antenna array, used to capture the CTE signal transmitted by the transmitter and extract the IQ data sequence from it.

[0037] It should be noted that the IQ data sequence can be represented as:

[0038] ;

[0039] in, Let be the IQ data sequence; n is the logarithm of the IQ data sequence; i k Extend the I component at the k-th sampling point of the CTE signal for each constant tone; q k Extend the Q component of the CTE signal at the kth sampling point for each constant tone.

[0040] Optionally, in one possible implementation, the specific steps of acquiring multiple sets of location-related data include:

[0041] Step 1: Generate a set of reference point locations based on a pre-defined reference point placement strategy.

[0042] In this step, the pre-set reference point placement strategy is set according to the actual structure and layout complexity of the ship's compartments, as well as the desired positioning accuracy. This strategy can be a regular grid-based placement strategy or a non-uniform dense placement strategy. It should be noted that, in one embodiment, regular grid-based reference points can be arranged at fixed intervals within the compartments to form a uniform regular grid-based placement strategy, suitable for open or regular compartment areas. In another embodiment, high-density reference points can be placed in key areas of the ship, while low-density reference points can be placed in ordinary areas, forming a non-uniform dense placement strategy to optimize resource allocation and adapt to complex ship layouts. For example, in one possible implementation, the spacing between reference points can be set to less than or equal to 0.5m to form a uniform regular grid-based placement strategy.

[0043] Simultaneously, for any reference point set in any deployment strategy, record the position coordinates (x, y, y) of each reference point. k y k Using this as a unique identifier, we obtain a set of all preset reference point locations.

[0044] Step 2: For each reference point in the set of reference point locations, place a transmitter at that reference point and place a receiver at a specific location on the hull.

[0045] In this step, for each reference point in the set of reference point locations obtained in step 1, a transmitter is placed at that reference point location to transmit a constant tone extended CTE signal, while the receiver is placed at a fixed position on the hull to capture the CTE signal and extract the IQ data sequence therein.

[0046] Step 3: For each of the preset multiple sampling scenarios, perform multiple rounds of sampling at each reference point position, and in each round of sampling, collect the IQ data sequence obtained by the receiver during the process of receiving the CTE signal from the transmitter; wherein, the preset multiple sampling scenarios are different, and the preset multiple sampling scenarios cover scenarios where any of the following conditions exist, as well as scenarios where none of the following conditions exist: personnel movement, equipment activation, and electromagnetic interference; the multiple rounds of sampling cover different time periods, and the hull is in different attitudes during the different time periods.

[0047] In this step, multiple sampling scenarios are preset for various situations during ship navigation. Furthermore, for multiple combinations of sampling scenarios, the position of each reference point is sampled multiple times, covering multiple time periods during ship operation to encompass different ship attitudes (possible ship attitudes include tilting, rolling, and other attitude changes caused by waves, engine vibrations, etc.) within these different time periods. IQ data sequences are also collected by the receiver during the reception of CTE signals from the transmitter. For example, in one embodiment, under the preset "equipment on + electromagnetic interference" sampling scenario, 10 rounds of sampling are performed at reference point P1 on the stern deck.

[0048] It should be noted that, based on the CTE signals collected in this step, a three-dimensional IQ data sequence covering dynamic scene interference, ship spatial position, and ship attitude changes can be extracted to simulate the fluctuation of the IQ data sequence during actual ship operation and improve the robustness of the collected data.

[0049] Step 4: Bind each IQ data sequence collected at each reference point location to that reference point location to obtain a set of positioning-related data.

[0050] In this step, for each reference point location, a binding relationship is established between all IQ data sequences transmitted from that reference point location and that reference point location, thereby obtaining a set of positioning-related data. For example, the complete IQ data sequence received by the receiver from a certain reference point device is... The actual spatial location label of the reference point device is In this way, a set of location-related data was obtained. .

[0051] It should be noted that by repeating the above steps, multiple sets of location-related data can be obtained.

[0052] Furthermore, in one possible implementation, after acquiring multiple sets of location-related data, the method further includes:

[0053] (1) Use statistical analysis methods to detect whether there are missing values ​​in each group of location-related data, and fill or remove the data with missing values.

[0054] In this step, for each set of location-related data, we check whether there are any null values ​​or incomplete records. For example, we can identify the missing data in each set by calculating the proportion of non-null values ​​or visualizing the distribution of missing values ​​through statistical analysis methods. Furthermore, for the location-related data with detected missing values, we select an appropriate imputation method based on the proportion of missing values ​​and the data characteristics, such as mean imputation, median imputation, interpolation, or model-based prediction imputation. If the proportion of missing values ​​is too high or imputation may introduce significant bias, we choose to remove the corresponding data records.

[0055] (2) Identify whether there are abnormal jump features in the IQ data sequence of each set of location-related data by using statistical thresholding or machine learning-based anomaly detection algorithms, and remove or correct the detected abnormal samples.

[0056] In this step, after handling missing values, it is necessary to further identify whether there are abnormal jump characteristics in the IQ data sequence of each set of localization-related data through statistical thresholding or anomaly detection algorithms based on machine learning. Among them, statistical thresholding usually sets upper and lower limits for specific indicators, and marks data points that exceed the threshold as anomalies. For example, assuming there is a set of IQ data sequences representing signal strength with values ​​of [50, 52, 51, 80, 53, 49, 100, 51], and the threshold range is [51, 101], then data point 50 is not within the threshold range, and this data point is marked as anomaly. For example, machine learning-based anomaly detection algorithms learn the normal distribution pattern of data and identify abrupt changes that deviate from the normal pattern. For instance, suppose there is a set of IQ data sequences representing signal strength with values ​​of [50, 52, 51, 80, 53, 49, 100, 51]. Since most values ​​are around 50, and 100 significantly deviates from the clustering area of ​​other data points, 100 is identified as an abrupt change that deviates from the normal pattern.

[0057] Furthermore, abnormal samples detected can be removed to avoid interference with subsequent analysis, or corrected through smoothing, interpolation, or other methods. This preserves the continuity of location-related data, improves its quality, and enhances the accuracy of subsequent analysis based on the data.

[0058] The feature-based positioning method for ship environments provided in this embodiment employs a regular gridding or non-uniform dense point deployment strategy. Bluetooth tags are deployed at preset reference points to transmit constant tone extended CTE signals, which are then captured by a receiver equipped with a multi-antenna array. IQ data sequences and angle-of-arrival (AOA) information are extracted. Subsequently, through multi-scenario, multi-round sampling covering dynamic scenarios such as personnel movement, equipment activation, electromagnetic interference, and different ship attitudes, multiple sets of positioning-related data containing IQ data sequences and reference point positions are obtained. Statistical analysis and anomaly detection methods are used to impute missing values ​​and correct abnormal transitions in the data. This effectively constructs a highly robust positioning dataset that can effectively simulate complex environmental interference during ship operation, improving the quality and accuracy of the positioning dataset.

[0059] S102. Using the IQ data sequence in each set of location-related data as input features and the location in that set of location-related data as a label, construct a first training sample set using the multiple sets of location-related data.

[0060] In this step, we assume that the IQ data sequence is in the i-th group of location-related data. As input features, this set locates the positions of reference points in the relevant data. As a label, the IQ data sequence in the i-th group of location-related data can be represented by the following formula:

[0061] ;

[0062] in, , where is the IQ feature vector of the i-th sample.

[0063] Furthermore, the location of the reference point in the i-th set of location-related data. This can be represented by the following formula:

[0064] ;

[0065] in, To locate the reference point in the relevant data for the i-th group, These are the location labels for C reference points.

[0066] In this way, by using multiple sets of location-related data, the first training sample set can be constructed:

[0067] ;

[0068] in, is the first training sample set; m is the number of location-related data.

[0069] S103. Train the gradient decision tree model using the first training sample set, and after training, determine the importance score of each feature in the input features based on the training results.

[0070] In this step, using the first training sample set obtained in the previous steps, the gradient decision tree model is trained using the LightGradient Boosting Machine (LightGBM) algorithm to learn from the input feature vector. Go to space tag The mapping relationship.

[0071] Specifically, the gradient boosting tree algorithm constructs a sequence containing T decision trees. The model is gradually optimized using a gradient boosting approach; specifically, for the first decision tree in the gradient decision tree model, the model optimizes the IQ feature vector from the input at a given time. To the predicted location The prediction result is obtained by mapping the tree, the error of the prediction in this round is fitted by the tree, and the prediction result of the next decision tree is guided by the error of the fitted tree, so as to gradually reduce the error and realize the training of the gradient decision tree model.

[0072] Furthermore, after training the gradient decision tree model, the importance score of each feature in the input features can be determined based on the training results using the following formula:

[0073] ;

[0074] in, Score the importance of feature j; For the j-th input feature; denoted as , where is the information gain when the j-th input feature is used as a split node in the t-th tree; r is a preset gain weight coefficient, the specific value of which can be set according to actual needs, and is not limited in this application; T is the total number of decision trees in the gradient decision tree model.

[0075] It should be noted that the calculation of information gain can be based on loss functions such as the Gini index, cross-entropy, or squared error; this application does not limit its calculation. Furthermore, the higher the importance score of an input feature, the greater its contribution to improving classification accuracy. It should be further clarified that the importance score in this application is used to characterize the contribution of the input feature to improving classification accuracy; specifically, it is a quantitative score obtained by accumulating the information gain brought by the input feature when used as a splitting node in the gradient decision tree model. The higher the importance score, the greater the contribution of the input feature to improving classification accuracy.

[0076] S104. For each feature, the predictive ability of the feature under independent input conditions is evaluated using the trained gradient decision tree model to obtain the overall stability score of the feature; wherein, the overall stability score is used to characterize the degree to which the feature maintains the stability of prediction accuracy at different positions.

[0077] It should be noted that the overall stability score measures the accuracy of the feature when it participates in localization prediction independently at different locations. The higher the overall stability score of the feature, the more stable and accurate the prediction is under multiple location and multiple sample conditions, and the stronger its versatility and robustness. The lower the overall stability score of the feature, the more the prediction effect of the feature depends on a specific location, the poor the stability, and the limited scope of application.

[0078] Specifically, in one possible implementation, the evaluation of the predictive ability of the feature under independent input conditions using a trained gradient decision tree model to obtain an overall stability score for the feature includes:

[0079] Step 1: For each feature, set the values ​​of all other features in each training sample in the first training sample set to 0 to generate a single-feature input dataset corresponding to that feature.

[0080] In this step, for each feature to be evaluated, all samples in the training sample set are iterated through, and all other feature values ​​in the current sample are forced to zero, thereby generating a dedicated dataset that retains only the original data of that feature. For example, for the training sample set... One of the training samples, where, for The first feature in the dataset, the single-feature input data corresponding to this feature can be represented as: .

[0081] Step 2: For each location of all transmitters contained in the multiple sets of positioning-related data, use the trained gradient decision tree model to predict the single feature input data corresponding to each feature, and obtain the prediction result of the feature at that location.

[0082] Specifically, referring to the examples above, for instance, regarding Find the single-feature input data corresponding to all training samples at that position from all training samples, and input this single-feature input data into the trained gradient decision tree model. For example, ... The input is fed into the gradient decision tree model to obtain the prediction result of the feature at that location.

[0083] Step 3: Based on the prediction result of the feature at that location, determine the local stability score of the feature at that location; wherein, the local stability score of the feature at that location is equal to the ratio of the number of samples effectively predicted by the feature at that location to the total number of samples corresponding to the feature at that location.

[0084] Specifically, if the prediction result of the feature at that location matches its actual location, then it is a valid prediction sample; otherwise, it is an invalid prediction sample.

[0085] Specifically, the local stability score of this feature at this location can be calculated using the following formula:

[0086] ;

[0087] in, Features In position Local stability score; Features In position The number of samples that are effectively predicted; Features In position The corresponding total number of samples.

[0088] Step 4: Determine the overall stability score of the feature based on its local stability scores at various locations.

[0089] In a specific implementation, in one embodiment, a weighted value can be calculated based on the local stability score of the feature at each location and the weights pre-set for each location, and the weighted value can be determined as the overall stability score of the feature.

[0090] Optionally, in one possible implementation, the specific implementation process of this step may include:

[0091] 1. Determine the weight of each position;

[0092] Specifically, in one possible implementation, the specific implementation process of this step may include: for each position, determining the weight of the position based on the number of samples corresponding to that position in the first training sample set and the total number of samples contained in the first training sample set; or, determining the weight of each position based on the number of positions contained in the first training sample set; wherein, the weight of each position is the same.

[0093] Specifically, in this implementation, the weight of a position can be determined based on the number of samples corresponding to that position:

[0094] ;

[0095] in, For position The weights; This represents the number of samples corresponding to this position in the first training sample set. This represents the total number of samples contained in the first training sample set.

[0096] Furthermore, if there are M transmitter locations, a fixed weight can be set so that the weight of each location is 1 / M.

[0097] It should be noted that both of the above weight determination methods are related to the number of samples, and the two methods are applicable to different scenarios. The first weight determination method based on the number of samples is suitable for scenarios where the sample distribution in the first training sample set is uneven. The second weight determination method based on equal weights is suitable for scenarios where the sample distribution in the first training sample set is uniform.

[0098] Furthermore, in another possible implementation, determining the weights of each position further includes:

[0099] (1) Determine the original weight of each location based on the importance level set by the user for each location and the correspondence between the preset importance level and the original weight.

[0100] Specifically, the user sets an importance level for each location, for example, cabin = emergency level, deck = important level, rest area = normal level; further, the correspondence between importance level and original weight is preset, for example, in one embodiment, the preset weight of emergency level is 0.7, the preset weight of important level is 0.5, and the preset weight of normal level is 0.3, thus obtaining the original weight of each location.

[0101] (2) Normalize the original weights of each position to obtain the weights of each position.

[0102] Specifically, the original weights at each position can be normalized using the following formula to obtain the weights at each position:

[0103] ;

[0104] in, For position The weights; For position The original weights; This is the sum of the original weights for all positions.

[0105] It should be noted that the specific method for determining the weights can be selected according to actual needs, and this application does not limit it.

[0106] In this step, the original weights at each position are normalized. This normalization process standardizes the original weights into a standard, easy-to-understand, and easy-to-use probability distribution format. This significantly improves the interpretability of the weights, the convenience of downstream applications, and mathematical consistency. It also reflects the relative importance differences between the original weights.

[0107] 2. The local stability scores of the feature at each location are weighted based on the weights at each location to obtain the overall stability score of the feature.

[0108] Specifically, in this step, the overall stability score of this feature can be obtained using the following formula:

[0109] ;

[0110] in, Features Overall stability score; For position The weights; Features In position Local stability score at the location.

[0111] S105. Determine the comprehensive score of each feature based on the importance score and overall stability score of each feature, and based on the comprehensive score of each feature, select the target feature set for localization from the multiple features contained in the input features.

[0112] In this step, before determining the overall score for each feature, it is also necessary to standardize the overall stability score and importance score of the feature (e.g., by standardizing based on Min-Max normalization).

[0113] Specifically, the importance score and overall stability score can be standardized using the Min-Max normalization process according to the following formula:

[0114] ;

[0115] in, For the first The standardized score of each feature ranges from [0, 1]; X is the score to be standardized, i.e., the importance score or the overall stability score. X is the minimum value among all features; X is the maximum value among all features.

[0116] It should be noted that the above formula is a general formula for Min-Max normalization. When standardizing the importance score and the overall stability score separately, the variables will also be adjusted accordingly; for example, according to the above formula, if the i-th... When the importance scores of the features are standardized, the resulting feature is the first one. The importance score of each feature after standardization is: , The minimum value among all feature importance scores; The maximum value among all feature importance scores; if for the th When the overall stability score of the feature is standardized, the result is the first feature. The overall stability score after standardization of each feature is: , The minimum value among all features' overall stability scores; The maximum value among all features in the overall stability score.

[0117] The comprehensive score for each feature is determined based on its importance score and overall stability score, including:

[0118] ;

[0119] in, For the first A comprehensive score for each feature; For the first The importance score of each feature after standardization; The weighting coefficients for importance scores; For the first The overall stability score after standardization of each feature; The weighting coefficients for the overall stability score.

[0120] It should be noted that the formula for determining the comprehensive score of each feature above uses a weighted linear combination method to calculate the comprehensive score. This calculation method can dynamically adjust the weights of the importance score and the overall stability score according to actual needs, which is convenient for interpretation and optimization.

[0121] Furthermore, in another embodiment, the comprehensive score of each feature can also be determined according to the following formula based on the importance score of each feature and the overall stability score:

[0122] ;

[0123] in, For the first A comprehensive score for each feature; For the first The importance score of each feature after standardization; The weighting coefficients for importance scores; For the first The overall stability score after standardization of each feature; The weighting coefficients for the overall stability score.

[0124] It should be noted that determining the comprehensive score for each feature by weighting the standardized importance scores and overall stability scores using power exponents with weight coefficients, and then summing the two power terms, is more suitable for scenarios requiring both importance and overall stability scores, and can provide higher environmental robustness. For example, in one embodiment, =0.7, =0.3; In another embodiment, =0.4, =0.6.

[0125] Furthermore, based on the comprehensive score of each feature, a target feature set for localization is selected from the multiple features contained in the input features.

[0126] Specifically, a preset comprehensive score threshold is used. If the comprehensive score of a feature in the input features is higher than the threshold, it is selected and added to the target feature set. Conversely, if the comprehensive score of a feature is lower than the threshold, it is removed and not selected for the target feature set. For example, in one embodiment, the comprehensive score of each feature is determined by the aforementioned linear weighted summation method. At this time, the comprehensive score of each feature is in the range of [0,1], and the comprehensive score threshold is 0.8. Then, features with a comprehensive score higher than 0.8 are selected and added to the target feature set for localization.

[0127] In addition, this step also sets an upper limit on the number of features contained in the target feature set. This limits the feature dimensions of the target feature set used for localization, thereby controlling the model complexity of the final trained localization model. For example, in one embodiment, The number is 20. In practice, if the number of target features selected using the above method is greater than 20, then the top 20 target features are selected in descending order of comprehensive score.

[0128] It should be noted that by evaluating each feature at each location point separately, calculating the local stability score of the feature at different spatial locations, and further integrating them into an overall stability score, the consistency of the feature's performance in variable environments can be effectively characterized. This avoids the problem of only considering the global average prediction performance while ignoring spatial differences. The selected features have more stable prediction capabilities across scenarios, thereby improving the model's generalization performance in new locations or dynamic environments. Furthermore, it is understandable that some features may perform well in some locations (e.g., due to a coincidence of a specific multipath pattern resulting in local high discriminative power), but perform extremely poorly in other regions. If only importance or global error is evaluated, such "locally advantageous" features may be incorrectly selected. The method provided in this embodiment evaluates the proportion of effective prediction samples location-by-location, which can eliminate features that lack stability in the overall space, thereby improving the reliability of feature selection. The final target feature set is selected by dual evaluation of "importance score" and "overall stability score," effectively reducing the negative impact of redundant features and unstable inputs on model training. This not only improves the prediction accuracy of the final positioning model, but also enhances the model's adaptability to new samples, boundary regions, and channel fluctuations, meeting the dual requirements of accuracy and stability in actual deployment.

[0129] S106. Based on the target feature set and the first training sample set, construct a second training sample set for training the localization model, and train the localization model using the second training sample set.

[0130] In this step, based on the selected target feature set for localization, features other than the input features in the target feature set are removed from the training sample set. The input features contained in the target feature set are retained. For example, if the training sample set contains 40 features and the target feature set contains only 12 features, then the remaining 28 features in the training sample set are removed, excluding the 12 features in the target feature set. The retained features are used as high-value features to construct a second training sample set, and the localization model is trained using the second training sample set. This not only effectively reduces the localization error of the localization model but also avoids redundant information caused by unnecessary low-value features.

[0131] S107. After obtaining the data to be predicted, extract the feature values ​​corresponding to the target feature set from the data to be predicted, and input the extracted feature values ​​into the positioning model so that the positioning model can predict the location corresponding to the data to be predicted.

[0132] Specifically, upon receiving the Bluetooth IQ data to be predicted, the system first dynamically extracts the feature values ​​of the features corresponding to the target feature set and removes other unnecessary redundant features. Then, the extracted optimized feature vector is input into the pre-trained localization model, which predicts the location corresponding to the data to be predicted.

[0133] The feature-based localization method for ship environments provided in this embodiment combines feature importance scores and overall stability scores to eliminate irrelevant or unstable features while retaining key features with high discriminative power for the localization task, effectively improving the accuracy of predicted location. Furthermore, when determining the overall stability score of a feature, the consistency of feature performance across different locations is comprehensively considered, which can improve the model's generalization ability in different cabins and environments and reduce prediction fluctuations. Moreover, feature selection reduces the input dimension, lowers the training and inference burden of the model, improves the real-time performance of the system, and adapts to the resource-constrained edge computing deployment requirements in ship environments.

[0134] Corresponding to the aforementioned embodiment of a feature-based positioning method in a ship environment, this application also provides an embodiment of a feature-based positioning device in a ship environment.

[0135] An embodiment of a feature-based positioning device in a ship environment, as described in this application, can be applied to feature-based positioning equipment in a ship environment. The device embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of the feature-based positioning equipment in its ship environment reading the corresponding computer program instructions from non-volatile memory into memory and executing them. From a hardware perspective, such as... Figure 3 The diagram shown is a hardware structure diagram of a feature-based positioning device in a ship environment, where the feature-based positioning device of this application is located. (Except for...) Figure 3 In addition to the processor, memory, network interface, and non-volatile memory shown, the positioning device in the embodiment, which is used in a marine environment, may also include other hardware depending on the actual function of the feature-based positioning device in that marine environment, which will not be described in detail here.

[0136] Figure 2 This is a schematic diagram of the first embodiment of the feature-based positioning device for a ship environment provided in this application. Please refer to... Figure 2The apparatus provided in this embodiment includes an acquisition module 210, a construction module 220, a determination module 230, and a prediction module 240; wherein,

[0137] The acquisition module 210 is used to acquire multiple sets of positioning-related data; wherein, each set of positioning-related data includes an IQ data sequence acquired by the receiver during the process of receiving a constant tone spread CTE signal transmitted from the transmitter, and the location of the transmitter;

[0138] The construction module 220 is used to construct a first training sample set using the IQ data sequence in each set of location-related data as input features and the position in the set of location-related data as a label;

[0139] The determining module 230 is used to train a gradient decision tree model using the first training sample set, and after training is completed, to determine the importance score of each feature in the input features based on the training results.

[0140] The determining module 230 is used to evaluate the predictive ability of each feature under independent input conditions using a trained gradient decision tree model to obtain an overall stability score for the feature; wherein, the overall stability score is used to characterize the degree to which the feature maintains its predictive accuracy at different positions.

[0141] The determining module 230 is used to determine the comprehensive score of each feature based on the importance score and the overall stability score of each feature, and to select the target feature set for localization from the multiple features contained in the input feature based on the comprehensive score of each feature.

[0142] The construction module 220 is used to construct a second training sample set for training the localization model based on the target feature set and the first training sample set, and to train the localization model with the second training sample set.

[0143] The prediction module 240 is used to extract feature values ​​corresponding to the target feature set from the data to be predicted after acquiring the data to be predicted, and input the extracted feature values ​​into the positioning model so that the positioning model can predict the location corresponding to the data to be predicted.

[0144] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0145] Please continue to refer to Figure 3This application also provides a feature-based positioning device in a ship environment, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods provided in the first aspect of this application.

[0146] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods provided in this application.

[0147] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0148] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0149] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A feature-based localization method, characterized in that, The positioning method is applied to a shipboard environment; the method includes: Multiple sets of location-related data are acquired; wherein each set of location-related data includes an IQ data sequence acquired by the receiver during the process of receiving a Constant Tone Extended (CTE) signal transmitted from the transmitter, and the location of the transmitter; Using the IQ data sequence in each set of location-related data as input features and the location in that set of location-related data as a label, a first training sample set is constructed using the multiple sets of location-related data. A gradient decision tree model is trained using the first training sample set. After training, the importance score of each feature in the input features is determined based on the training results. The importance score is used to characterize the degree of contribution of the feature to improving the classification accuracy. The higher the importance score, the greater the effect of the feature on improving the classification accuracy. For each feature, the predictive ability of the feature under independent input conditions is evaluated using a trained gradient decision tree model to obtain the overall stability score of the feature; wherein, the overall stability score is used to characterize the degree to which the feature maintains its predictive accuracy at different positions; Based on the importance score and overall stability score of each feature, a comprehensive score for each feature is determined, and based on the comprehensive score of each feature, a target feature set for localization is selected from the multiple features contained in the input feature. Based on the target feature set and the first training sample set, a second training sample set is constructed for training the localization model, and the localization model is trained using the second training sample set. After obtaining the data to be predicted, feature values ​​corresponding to the target feature set are extracted from the data to be predicted, and the extracted feature values ​​are input into the positioning model so that the positioning model can predict the location corresponding to the data to be predicted.

2. The method according to claim 1, characterized in that, For each feature, the predictive power of that feature under independent input conditions is evaluated using a trained gradient decision tree model to obtain an overall stability score for that feature, including: For each feature, the values ​​of all other features in each training sample in the first training sample set are set to 0 to generate a single-feature input dataset corresponding to that feature; For each location among all transmitter locations included in the multiple sets of location-related data, a trained gradient decision tree model is used to predict the single feature input data corresponding to each feature, thereby obtaining the prediction result of that feature at that location. Based on the prediction result of the feature at that location, determine the local stability score of the feature at that location; wherein, the local stability score of the feature at that location is equal to the ratio of the number of samples effectively predicted by the feature at that location to the total number of samples corresponding to the feature at that location; The overall stability score of the feature is determined based on its local stability scores at various locations.

3. The method according to claim 1, characterized in that, The comprehensive score for each feature is determined based on its importance score and overall stability score, including: ; in, For the first A comprehensive score for each feature; For the first The importance score of each feature after standardization; The weighting coefficients for importance scores; For the first The overall stability score after standardization of each feature; The weighting coefficients for the overall stability score.

4. The method according to claim 2, characterized in that, The process of determining the overall stability score of a feature based on its local stability scores at various locations includes: Determine the weight of each position; The local stability scores of the feature at each location are weighted according to the weights at each location to obtain the overall stability score of the feature.

5. The method according to claim 4, characterized in that, Determining the weights of each position includes: For each position, the weight of that position is determined based on the number of samples corresponding to that position in the first training sample set and the total number of samples contained in the first training sample set. or, The weight of each position is determined based on the number of positions contained in the first training sample set; wherein, the weight of each position is the same.

6. The method according to claim 4, characterized in that, Determining the weights of each position includes: Based on the importance level set by the user for each location, and the correspondence between the preset importance level and the original weight, the original weight of each location is determined. The original weights at each position are normalized to obtain the weights at each position.

7. The method according to claim 1, characterized in that, The acquisition of multiple sets of location-related data includes: A set of reference point locations is generated based on a pre-defined reference point placement strategy; For each reference point in the set of reference point locations, a transmitter is placed at that reference point, and a receiver is placed at a specific location on the hull. For each of the preset multiple sampling scenarios, multiple rounds of sampling are performed at each reference point position. In each round of sampling, the IQ data sequence acquired by the receiver during the reception of the CTE signal from the transmitter is collected. The preset multiple sampling scenarios are different and cover scenarios where any of the following conditions exist, as well as scenarios where none of the following conditions exist: personnel movement, equipment activation, and electromagnetic interference. The multiple rounds of sampling cover different time periods, and the hull is in different attitudes during the different time periods. Each IQ data sequence collected at each reference point is bound to that reference point location to obtain a set of positioning-related data.

8. The method according to claim 1 or 7, characterized in that, After acquiring multiple sets of location-related data, the method further includes: Statistical analysis methods were used to detect whether there were missing values ​​in each group of location-related data, and data with missing values ​​were filled or removed. The statistical threshold method or machine learning-based anomaly detection algorithm is used to identify whether there are abnormal jump features in the IQ data sequence of each set of location-related data, and the detected abnormal samples are removed or corrected.

9. A feature-based positioning device for a ship environment, characterized in that, The device includes an acquisition module, a construction module, a determination module, and a prediction module; wherein, The acquisition module is used to acquire multiple sets of positioning-related data; wherein, each set of positioning-related data includes an IQ data sequence acquired by the receiver during the process of receiving a constant tone spread CTE signal transmitted from the transmitter, and the location of the transmitter; The construction module is used to construct a first training sample set using the IQ data sequence in each set of location-related data as input features and the position in the set of location-related data as a label; The determining module is used to train a gradient decision tree model using the first training sample set, and after training, to determine the importance score of each feature in the input features based on the training results. The importance score is used to characterize the degree of contribution of the feature to improving the classification accuracy. The higher the importance score, the greater the effect of the feature on improving the classification accuracy. The determining module is used to evaluate the predictive ability of each feature under independent input conditions using a trained gradient decision tree model, and obtain the overall stability score of the feature; wherein, the overall stability score is used to characterize the degree to which the feature maintains its predictive accuracy at different positions. The determining module is used to determine the comprehensive score of each feature based on the importance score and the overall stability score of each feature, and to select the target feature set for localization from the multiple features contained in the input feature based on the comprehensive score of each feature. The construction module is used to construct a second training sample set for training the localization model based on the target feature set and the first training sample set, and to train the localization model with the second training sample set. The prediction module is used to extract feature values ​​corresponding to the target feature set from the data to be predicted after obtaining the data to be predicted, and input the extracted feature values ​​into the positioning model so that the positioning model can predict the location corresponding to the data to be predicted.

10. A feature-based positioning device for a ship environment, characterized in that, The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method according to any one of claims 1-8.

Citation Information

Patent Citations

  • Indoor positioning method and equipment

    CN118444245A

  • KR20250062665A