A sonar depth real-time accuracy correction system and method

Abstract

The application relates to the technical field of wireless ranging, and discloses a sonar depth real-time precision correction system and method, which comprises the following steps: synchronously collecting depth values and hydrological information of each detection point in a target water area; predicting and completing the hydrological information of each detection point; dividing the water area corresponding to each detection point into different detection layers based on the hydrological information; predicting the depth measurement accuracy of each detection layer based on the hydrological information, and calculating the cumulative accuracy of each detection layer; constructing a dynamic correction boundary of the depth value based on the depth measurement accuracy and the cumulative accuracy; and compensating and correcting the depth value of the detection point whose depth value exceeds the dynamic correction boundary. The application dynamically determines the accuracy of the depth measurement data through the hydrological data, and improves the precision and compensation adjustment efficiency of the sonar depth measurement in a complex water area.

Description

Technical Field

[0001] This application relates to the technical field of wireless ranging, specifically to a real-time accuracy correction system and method for sonar depth sounding. Background Technology

[0002] Sonar bathymetry is a core tool for underwater topographic surveying. Its principle is to calculate the depth of the seabed by analyzing the propagation and reflection of sound waves in water. However, in practical applications, sound wave propagation is significantly affected by the aquatic environment, such as sediment concentration, particle size, and sound velocity profile, making it difficult to guarantee bathymetry accuracy. This is especially true in complex waters where hydrological conditions change dynamically, exacerbating the error problem in sonar bathymetry. Currently, sonar bathymetry accuracy correction techniques mainly rely on corrections based on sound velocity profiles and statistical corrections based on empirical models, both of which generally suffer from several technical limitations.

[0003] Due to equipment and environmental limitations, the vertical distribution of deep sediment concentration and particle size data is difficult to collect continuously, failing to reflect the true hydrological stratification characteristics. The effects of sound velocity, sediment concentration, and particle size on sound wave propagation are coupled, but existing correction models often treat this separately, leading to error accumulation. Current technologies often treat water bodies as homogeneous or simply stratified, such as dividing them by fixed depth intervals, without dynamic clustering based on hydrological characteristics, making it difficult to accurately characterize complex water body structures. Sonar sounding errors accumulate non-linearly with increasing water depth, but existing methods lack quantitative assessment of the multi-layer superposition effect, resulting in low reliability of deep sounding results. Traditional methods divide correction intervals using fixed depth thresholds, which cannot adapt to the dynamic accuracy attenuation patterns of different water bodies.

[0004] For example, Chinese Patent CN104569988B discloses a correction method for deep water measurement using echo sounding. This method corrects the water depth measurement value by correcting the sound velocity generated by temperature gradients in the water body and the delay between the receiver of the Global Navigation Satellite System and the data collected by the echo sounder. This method can accurately detect errors and discrepancies in the water depth measurement by the echo sounder, as well as system delay issues. It corrects the water depth measurement value by measuring water temperature and the measured sound velocity profile, and compares the correction model with the true values ​​of various correction standards to achieve an approximate true value or an error tolerance range.

[0005] For example, Chinese Patent CN101551453B discloses a method for correcting the error of an echo sounder at a work site and its depth measurement comparator. This method uses an echo intercepting metal plate, which is horizontally suspended from a rope with length markings, as a depth measurement comparator. It works in conjunction with the echo sounder and its transducer to correct the error of the echo sounder at the work site. The correction steps are as follows: First, install the echo sounder and its transducer according to the required water depth. Then, place the echo intercepting metal plate of the depth measurement comparator into the water at several depths. Determine the accurate value of each depth through the length markings on the rope of the echo intercepting metal plate. At each depth, the echo intercepting metal plate intercepts the sound waves emitted by the transducer. By adjusting the sound velocity of the echo sounder, the depth indicated by its recorder or display panel is made consistent with and stable with the depth of the corresponding echo intercepting metal plate, thus completing the comparison and correction of the echo sounder.

[0006] The existing technologies mentioned above all suffer from the problem described in this background: the lack of quantitative assessment of the multi-layer superposition effect leads to low reliability of deep depth sounding results.

[0007] The information disclosed in this background section is intended only to enhance the understanding of the overall background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0008] The technical problem to be solved by this application is to overcome the defects of the prior art and provide a real-time accuracy correction system and method for sonar depth sounding, which dynamically determines the accuracy of depth sounding data through hydrological data and improves the accuracy and compensation adjustment efficiency of sonar depth sounding in complex waters.

[0009] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0010] On the one hand, this application provides a method for real-time accuracy correction of sonar depth sounding, including the following steps:

[0011] Simultaneously collect depth values ​​and hydrological information at each detection point in the target water area;

[0012] Predict and complete the hydrological information for each monitoring point;

[0013] Based on the hydrological information, the water area corresponding to each detection point is divided into different detection layers;

[0014] The depth measurement accuracy of each detection layer is predicted based on hydrological information, and the cumulative accuracy of each detection layer is calculated.

[0015] A dynamic correction boundary for the depth value is constructed based on the depth measurement accuracy and cumulative accuracy.

[0016] For detection points whose depth values ​​exceed the dynamic correction boundary, depth value compensation correction is performed.

[0017] As a preferred embodiment of the real-time accuracy correction method for sonar depth sounding described in this application, the hydrological information includes sediment concentration, sediment particle size, and sound velocity.

[0018] For any detection point in the target water area, the depth value of the bottom of the water is detected by sonar, and sampling points are set at the same location, starting from the water surface and in a direction perpendicular to the horizontal plane downwards;

[0019] Vertical distribution data of hydrological information were collected at each sampling point; among them, the vertical distribution data of sound velocity included the sound velocity at all sampling points from the water surface to the bottom; the vertical distribution data of sediment concentration included at least continuous data from the water surface. Sediment concentration at each sampling point; vertical distribution data of sediment particle size including at least continuous values ​​starting from the water surface. The particle size of sediment at each sampling point; , All are positive integers.

[0020] As a preferred embodiment of the real-time accuracy correction method for sonar depth sounding described in this application, the step of predicting and completing the hydrological information for each detection point specifically includes predicting and completing the underwater sediment concentration and sediment particle size; the method is as follows:

[0021] S10: For any detection point, mark the sediment concentration or sediment particle size as the target hydrological information;

[0022] S20: Organize the vertical distribution data of the target hydrological information into a target sequence according to the sampling points from shallow to deep.

[0023] S30: Extract sequence segments from the target sequence and input them into the trained prediction model; the prediction model outputs predicted values ​​of the target hydrological information;

[0024] S40: Verify and correct the predicted values ​​of the target hydrological information based on the sound velocity at the corresponding sampling points;

[0025] S50: Add the predicted values ​​as target hydrological information for the corresponding sampling points to the target sequence; that is, fill the predicted values ​​at the end of the target sequence according to the order of the corresponding sampling points from shallow to dark.

[0026] S60: Repeat S30-S50 until the target sequence contains the target hydrological information of all sampling points.

[0027] As a preferred embodiment of the sonar bathymetry real-time accuracy correction method described in this application, the method for verifying and correcting the predicted value of any target hydrological information is as follows:

[0028] S401: Obtain the sound velocity at the sampling point corresponding to the target hydrological information, and use it as a reference sound velocity;

[0029] S402: Based on a pre-built hydrological information mapping table, query the reference value range of the target hydrological information corresponding to the reference sound velocity;

[0030] S403: If the predicted value of the target hydrological information falls within the corresponding reference value range, no adjustment is made; otherwise, execute S404.

[0031] S404: Correct the predicted values ​​of the target hydrological information based on the corresponding reference value range.

[0032] As a preferred embodiment of the sonar depth sounding real-time accuracy correction method described in this application, the hydrological information mapping table contains a reference value range for each type of target hydrological information corresponding to each value of sound velocity; the method for constructing the hydrological information mapping table is as follows:

[0033] Choose any value of sound speed as the target sound speed; determine the reference value range of any target hydrological information corresponding to the target sound speed, specifically including:

[0034] Acquire historical data; each historical data point contains a set of sediment concentration, sediment particle size, and sound velocity.

[0035] Calculate the absolute value of the difference between the sound speed and the target sound speed in each historical data point, and use it as the sound speed difference for each historical data point; set a sound speed difference threshold, and extract historical data with a sound speed difference less than the sound speed difference threshold as reference data;

[0036] Calculate the Gaussian kernel weight for each reference data based on the sound velocity difference; calculate the weighted mean and weighted standard deviation of the target hydrological information in all reference data based on the Gaussian kernel weight;

[0037] A reference value range for the target hydrological information is set based on the weighted mean and weighted standard deviation.

[0038] The reference intervals for each type of target hydrological information corresponding to each value of sound speed are determined sequentially to obtain the hydrological information mapping table.

[0039] As a preferred embodiment of the real-time accuracy correction method for sonar depth sounding described in this application, the method for dividing the water area corresponding to any detection point into different detection layers is as follows:

[0040] Each type of hydrological information at each sampling point is normalized and encoded into a feature vector;

[0041] Based on the feature vector of hydrological information of each sampling point, hierarchical clustering is used to divide the sampling points into different clusters;

[0042] Based on the partitioning of clusters, the water area corresponding to the detection point is divided into different detection layers; each cluster corresponds to a detection layer; the layer height of each detection layer is recorded.

[0043] As a preferred embodiment of the sonar depth sounding real-time accuracy correction method described in this application, the prediction of the depth sounding accuracy of each detection layer specifically includes:

[0044] Calculate the mean of each hydrological information item from all sampling points in each detection layer, and use it as the hydrological information item for each detection layer.

[0045] Encode each piece of hydrological information from each detection layer into a feature vector for that layer.

[0046] Prediction points are uniformly set in the vertical direction in each detection layer; prediction points are also set at the bottom surface of any detection layer; the distance from each prediction point to the top surface of the corresponding detection layer is recorded as the layer depth of each prediction point;

[0047] The feature vector of the detection layer and the layer depth of each prediction point are input into the trained error prediction model; the error prediction model calculates and outputs the depth measurement accuracy of each detection layer; the depth measurement accuracy of any detection layer includes the significant accuracy of each prediction point; the significant accuracy of any prediction point represents the probability that the depth measurement error of the prediction point is less than the preset depth measurement error threshold when sonar depth measurement is performed starting from the top surface of the corresponding detection layer.

[0048] The cumulative accuracy of each detection layer is calculated based on the depth measurement accuracy, specifically including:

[0049] The significant accuracy of the prediction point corresponding to the bottom surface of each detection layer is marked as the minimum accuracy of the detection layer; if the top surface of the detection layer is the water surface of the target water area, the cumulative accuracy is the minimum accuracy of the detection layer; otherwise, the cumulative accuracy is the product of the minimum accuracy of the detection layer and the minimum accuracy of each preceding detection layer; the preceding detection layer is any detection layer above the detection layer.

[0050] As a preferred embodiment of the sonar depth sounding real-time accuracy correction method described in this application, the method includes: constructing a dynamic correction boundary for the depth value based on the depth sounding accuracy and cumulative accuracy, specifically including:

[0051] If the cumulative accuracy of any detection layer is less than the preset accuracy threshold, and there is no detection layer above it or the cumulative accuracy of the adjacent detection layer above it is greater than or equal to the preset accuracy threshold, then the corresponding detection layer is marked as a critical layer.

[0052] Determine the dynamic correction boundary in the water area corresponding to the detection point in the critical layer; specifically including:

[0053] The cumulative accuracy of each prediction point in the critical layer is calculated sequentially from shallow to deep. If the top surface of the critical layer is the water surface of the target water area, the cumulative accuracy of any prediction point is the significant accuracy of the prediction point. Otherwise, the cumulative accuracy of any prediction point is the product of the significant accuracy of the prediction point and the cumulative accuracy of the adjacent detection layer above the critical layer.

[0054] The depth corresponding to the prediction point whose cumulative accuracy is closest to the accuracy threshold is marked as the dynamic correction boundary.

[0055] As a preferred embodiment of the sonar depth sounding real-time accuracy correction method described in this application, wherein: a depth value exceeding the dynamic correction boundary indicates that the depth value of the detection point is greater than the depth corresponding to the dynamic correction boundary in the water area; the depth value compensation correction specifically includes:

[0056] Mark the detection point whose depth value is less than or equal to the depth corresponding to the dynamic correction boundary as a reference point;

[0057] For any detection point whose depth value needs to be compensated and corrected, the reference depth of the detection point is obtained by interpolation using the depth value of the reference point.

[0058] The depth value of the detection point is filtered and fused with the corresponding reference depth to obtain the depth value after compensation and correction of the detection point.

[0059] Secondly, this application provides a real-time accuracy correction system for sonar bathymetry, including a data acquisition module, a hydrological prediction module, a clustering module, a bathymetry prediction module, a boundary module, and a compensation correction module; wherein:

[0060] The data acquisition module is used to collect the depth value and hydrological information of each detection point in the target water area;

[0061] The hydrological prediction module is used to predict and complete the hydrological information of each detection point;

[0062] The clustering module divides the water area at each detection point into different detection layers based on hydrological information;

[0063] The depth prediction module is used to predict the depth accuracy of each detection layer and calculate the cumulative accuracy of each detection layer.

[0064] The boundary module constructs a dynamic correction boundary for the depth value based on the depth measurement accuracy and cumulative accuracy.

[0065] The compensation and correction module is used to compensate and correct the depth values ​​of detection points whose depth values ​​exceed the dynamic correction boundary.

[0066] Compared with the prior art, the beneficial effects achieved by this application are as follows:

[0067] This application achieves full-depth completion of sediment concentration and particle size by using a sequence prediction model and a sound velocity-assisted verification mechanism, expanding the coverage of hydrological data and laying a data foundation for subsequent accuracy correction. By constructing a hydrological information mapping table, the correlation between sound velocity and sediment parameters is quantified, enabling the correction process to comprehensively consider the influence of multiple factors, thereby achieving more accurate error compensation under complex hydrological conditions. This application proposes an adaptive clustering method based on hydrological features, using a hierarchical clustering algorithm to dynamically divide the detection layers, more accurately reflecting the actual structural characteristics of the water body. A Bayesian neural network is combined to predict the depth measurement accuracy of each detection layer, and a cumulative accuracy quantification of the multi-layer superposition effect is introduced, making the error assessment more scientific and reasonable, and improving the reliability of depth measurement in deep water areas. Attached Figure Description

[0068] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0069] Figure 1 A flowchart of a real-time accuracy correction method for sonar depth sounding provided in this application;

[0070] Figure 2 A schematic diagram of a real-time accuracy correction system for sonar depth sounding provided in this application;

[0071] Figure 3 This is a schematic diagram of a dynamically corrected boundary provided in this application. Detailed Implementation

[0072] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0073] Example 1

[0074] This embodiment describes a method for real-time accuracy correction of sonar depth sounding, referring to... Figure 1 The method includes the following steps:

[0075] Simultaneously collect depth values ​​and hydrological information at each detection point in the target water area;

[0076] The hydrological information includes sediment concentration, sediment particle size, and sound velocity.

[0077] For any detection point in the target water area, the depth value of the bottom is detected by sonar, and sampling points are set at the same location, starting from the water surface and in a direction perpendicular to the horizontal plane downwards; preferably, sampling points are set at fixed sampling intervals.

[0078] Vertical distribution data of hydrological information were collected at each sampling point; among them, the vertical distribution data of sound velocity included the sound velocity at all sampling points from the water surface to the bottom; the vertical distribution data of sediment concentration included at least continuous data from the water surface. Sediment concentration at each sampling point; vertical distribution data of sediment particle size including at least continuous values ​​starting from the water surface. The particle size of sediment at each sampling point; , All are positive integers.

[0079] Optionally, underwater sound velocity can be collected using a sound velocity profiler. A sound velocity profiler can collect sound velocity profile data at different depths underwater, and the sampling points can cover the entire range of sonar depth sounding. Changes in sound velocity with depth cause deflections in the sound wave propagation path, resulting in errors in sonar measurements. This is a core reference factor for correcting the accuracy of sonar depth sounding.

[0080] Optionally, underwater sediment concentration is collected using an optical turbidimeter, and underwater sediment particle size is collected using a laser particle size analyzer. Higher sediment concentration leads to stronger scattering and absorption losses during sound wave propagation; high sediment concentration results in severe attenuation of the sound wave echo signal, weakening the primary reflected echo or causing false echoes, thus reducing the accuracy of sonar depth sounding. Sediment particle size determines the scattering intensity of the sound wave; different particle size distributions alter the sound wave energy propagation path and reflection characteristics, interfering with echo signal identification.

[0081] Hydrological information at each monitoring point is predicted and supplemented; specifically, this includes predicting and supplementing underwater sediment concentration and sediment particle size; the method is as follows:

[0082] S10: For any detection point, mark the sediment concentration or sediment particle size as the target hydrological information;

[0083] S20: Organize the vertical distribution data of the target hydrological information into a target sequence according to the sampling points from shallow to deep.

[0084] S30: Extract sequence segments from the target sequence and input them into the trained prediction model; the prediction model outputs predicted values ​​of the target hydrological information;

[0085] The step of extracting a sequence segment from the target sequence specifically includes: extracting the last segment from the target sequence. The target hydrological information is used as a sequence segment input to the prediction model; It is a positive integer; the sequence segment corresponds to the current deepest... Sampling points for known target hydrological information;

[0086] The prediction model has sequence prediction capabilities, and its output includes continuous... The predicted value of the target hydrological information; the continuous The sampling points and sequence segments corresponding to the target hydrological information The sampling points corresponding to the target hydrological information are adjacent; Let be a positive integer; for example, number the sampling points in ascending order of depth. If the current target sequence includes target hydrological information from sampling points 1 to 30, let... If the value is 5, then the predicted values ​​of the five consecutive target hydrological information outputs by the prediction model correspond to sampling points 31 to 35. Optionally, an RNN (Recurrent Neural Network) model can be trained as the prediction model. Further, prediction models are trained separately for sediment concentration and sediment particle size. After training the prediction model for one of the two parameters (sediment concentration or sediment particle size), the model structure is saved, and the saved model structure is trained and its parameters are fine-tuned using historical data for the other parameter, thereby avoiding excessive computational consumption from training two models separately.

[0087] S40: Verify and correct the predicted values ​​of the target hydrological information based on the sound velocity at the corresponding sampling points;

[0088] The method for verifying and correcting the predicted value of any target hydrological information is as follows:

[0089] S401: Obtain the sound velocity at the sampling point corresponding to the target hydrological information, and use it as a reference sound velocity;

[0090] S402: Based on a pre-built hydrological information mapping table, query the reference value range of the target hydrological information corresponding to the reference sound velocity;

[0091] S403: If the predicted value of the target hydrological information falls within the corresponding reference value range, no adjustment is made; otherwise, execute S404.

[0092] S404: Correct the predicted values ​​of the target hydrological information based on the corresponding reference value range.

[0093] The preferred method for correcting the predicted value of the target hydrological information in this embodiment is as follows: Extract the value corresponding to the midpoint of the reference value interval as the reference value of the target hydrological information; subtract the corresponding reference value from the predicted value of the target hydrological information to obtain the correction term for the target hydrological information; subtract the corresponding correction term from the predicted value of the target hydrological information to obtain the corrected predicted value; optionally, an adjustment coefficient greater than 0 and less than 1 is set. Subtract the predicted value of the target hydrological information from A correction term is added to obtain the corrected predicted value to prevent overcorrection.

[0094] The hydrological information mapping table contains a reference value range for each type of target hydrological information corresponding to each value of sound velocity; the method for constructing the hydrological information mapping table is as follows.

[0095] Choose any value of sound speed as the target sound speed; determine the reference value range of any target hydrological information corresponding to the target sound speed, specifically including:

[0096] Acquire historical data; each historical data point contains a set of sediment concentration, sediment particle size, and sound velocity.

[0097] Calculate the absolute value of the difference between the sound speed and the target sound speed in each historical data point, and use it as the sound speed difference for each historical data point; set a sound speed difference threshold, and extract historical data with a sound speed difference less than the sound speed difference threshold as reference data;

[0098] Calculate the Gaussian kernel weight for each reference data based on the sound speed difference;

[0099] The weighted mean and weighted standard deviation of the target hydrological information in all reference data are calculated based on Gaussian kernel weights.

[0100] The reference value range for the target hydrological information is set based on the weighted mean and weighted standard deviation.

[0101] In this embodiment, a Gaussian kernel weight is preferably used as the weight for each reference data point. The Gaussian kernel weight is calculated based on the Gaussian kernel function; the smaller the sound velocity difference, the larger the Gaussian kernel weight. When calculating the mean and standard deviation of the target hydrological information in all reference data, the Gaussian kernel weight is multiplied by the target hydrological information to replace the original value in the reference data. This ensures that when determining the reference value range of the target hydrological information, more attention is paid to reference data that is closer to the target sound velocity. Optionally, the reference value range is set to the weighted mean plus or minus three times the weighted standard deviation.

[0102] The reference intervals for each type of target hydrological information corresponding to each value of sound speed are determined sequentially to obtain the hydrological information mapping table.

[0103] The speed of sound underwater is affected by sediment concentration and particle size. For example, lower sound speeds often correspond to underwater environments with high sediment concentration and fine sediment particle size; higher sound speeds often correspond to underwater environments with low sediment concentration and coarse sediment particle size. The reference range for each target hydrological information recorded in the hydrological information mapping table can verify the rationality of its predicted values ​​and perform numerical correction, ensuring prediction accuracy.

[0104] S50: Add the predicted values ​​as target hydrological information for the corresponding sampling points to the target sequence; that is, fill the predicted values ​​at the end of the target sequence according to the order of the corresponding sampling points from shallow to dark.

[0105] In this embodiment, the predicted values ​​output by the prediction model in each round are added to the target sequence and then truncated into the sequence segment in the next round of iteration as the input value for the next round of prediction. This application verifies the reasonableness of the predicted value by using the sound velocity value at the corresponding depth for each predicted value. If it is unreasonable, the predicted value is corrected according to the reference value range corresponding to the reference sound velocity, thereby preventing the accumulation and spread of prediction errors in iterative prediction. Even when the measured values ​​of sediment concentration and sediment particle size are few, it can maintain high-precision prediction across the entire depth range.

[0106] S60: Repeat S30-S50 until the target sequence contains the target hydrological information of all sampling points.

[0107] This application utilizes observable sediment particle size and concentration in shallow layers to predict the evolution trend with depth, and verifies and adjusts it by measuring the sound velocity variation trend across the entire depth range. This allows the predicted evolution trend to be reasonably extended to unobserved depths, thus completing the prediction and supplementation of hydrological information. It solves the current problem of limited sampling range for sediment concentration and particle size distribution due to environmental and equipment limitations, and can obtain hydrological information across the entire range of sonar sounding, providing a reference data basis for the accuracy correction of sounding data.

[0108] Based on the hydrological information, the water area corresponding to each detection point is divided into different detection layers;

[0109] The method for dividing the water area corresponding to any detection point into different detection layers is as follows:

[0110] Each type of hydrological information at each sampling point is normalized and encoded into a feature vector;

[0111] Based on the feature vector of hydrological information at each sampling point, hierarchical clustering is used to divide the sampling points into different clusters;

[0112] Based on the partitioning of clusters, the water area corresponding to the detection point is divided into different detection layers; each cluster corresponds to a detection layer; the layer height of each detection layer is recorded.

[0113] In this embodiment, for any given detection point, all sampling points are located on the same vertical line; therefore, in the water area corresponding to the detection point, different detection layers are continuous in the vertical direction but do not intersect in the horizontal direction. Optionally, hierarchical clustering based on a distance matrix can be used, combined with post-processing techniques such as layer thickness limitation and window smoothing, to divide the water area into different detection layers to adapt to the complexity of the water structure.

[0114] Predict the depth measurement accuracy for each detection layer and calculate the cumulative accuracy for each detection layer;

[0115] The predicted depth measurement accuracy for each detection layer specifically includes:

[0116] Calculate the mean of each hydrological information item from all sampling points in each detection layer, and use it as the hydrological information item for each detection layer.

[0117] Encode each piece of hydrological information from each detection layer into a feature vector for that layer.

[0118] Prediction points are uniformly set in the vertical direction in each detection layer; prediction points are also set at the bottom surface of any detection layer; the distance from each prediction point to the top surface of the corresponding detection layer is recorded as the layer depth of each prediction point;

[0119] The feature vector of the detection layer and the layer depth of each prediction point are input into the trained error prediction model; the error prediction model calculates and outputs the depth measurement accuracy of each detection layer; the depth measurement accuracy of any detection layer includes the significant accuracy of each prediction point; the significant accuracy of any prediction point represents the probability that the depth measurement error at the prediction point is less than the preset depth measurement error threshold when sonar depth measurement is performed starting from the top surface of the corresponding detection layer.

[0120] In this embodiment, the error prediction model calculates the accuracy probability of the depth value at each prediction point in the detection layer, i.e., at each depth, when performing depth sounding starting from the detection layer with the input feature vector. This means that the error prediction model does not need to process the different hydrological characteristics of multiple detection layers, but only needs to calculate the accuracy probability of sonar depth sounding at different depths of a single detection layer, simplifying the model complexity and the computing power required for model training, deployment, and operation, and improving the prediction accuracy of depth sounding for each detection layer. When training the error prediction model, it is only necessary to select sonar depth sounding data and corresponding actual depth data of a water body with relatively stable and uniform underwater hydrological characteristics. In this embodiment, a BNN (Bayesian Neural Network) model is preferred as the error prediction model; the BNN model can learn the mapping relationship between sediment concentration, sediment particle size, sound velocity profile, and sonar depth sounding accuracy, and outputs the sonar depth sounding accuracy at each depth, i.e., the probability that the depth sounding error is less than a preset depth sounding error threshold, through an output layer configured with a Gaussian mixture model.

[0121] The cumulative accuracy of each detection layer is calculated based on the depth measurement accuracy, specifically including:

[0122] The significant accuracy of the prediction point corresponding to the bottom surface of each detection layer is marked as the minimum accuracy of the detection layer; if the top surface of the detection layer is the water surface of the target water area, the cumulative accuracy is the minimum accuracy of the detection layer; otherwise, the cumulative accuracy is the product of the minimum accuracy of the detection layer and the minimum accuracy of each preceding detection layer; the preceding detection layer is any detection layer above the detection layer.

[0123] In this embodiment, the error prediction model assumes that sonar depth sounding is performed starting from the top surface of each detection surface, and the depth sounding accuracy is calculated. Underwater, the significant accuracy of depth sounding gradually decreases with increasing water depth. For the detection layer at the water surface, the significant accuracy at the top surface is 1. For other detection layers, since there are other detection layers above, the significant accuracy at the top surface is not 1, but rather the cumulative accuracy of the previous detection layer. That is, the starting point for the significant accuracy to decrease in the detection layer is not 1, but rather the cumulative accuracy of the previous detection layer. This embodiment integrates the accuracy of the independent detection layers output by the error prediction model into the actual hydrological state of multiple superimposed detection layers by multiplying the minimum accuracy of each detection layer with the minimum accuracy of each preceding detection layer.

[0124] A dynamic correction boundary for the depth value is constructed based on the depth measurement accuracy and cumulative accuracy; specifically including:

[0125] If the cumulative accuracy of any detection layer is less than the preset accuracy threshold, and there is no detection layer above it or the cumulative accuracy of the adjacent detection layer above it is greater than or equal to the preset accuracy threshold, then the corresponding detection layer is marked as a critical layer.

[0126] Determine the dynamic correction boundary in the water area corresponding to the detection point in the critical layer; specifically including:

[0127] The cumulative accuracy of each prediction point in the critical layer is calculated sequentially from shallow to deep. If the top surface of the critical layer is the water surface of the target water area, the cumulative accuracy of any prediction point is the significant accuracy of the prediction point. Otherwise, the cumulative accuracy of any prediction point is the product of the significant accuracy of the prediction point and the cumulative accuracy of the adjacent detection layer above the critical layer.

[0128] The depth corresponding to the prediction point whose cumulative accuracy is closest to the accuracy threshold is marked as the dynamic correction boundary.

[0129] Figure 3 A schematic diagram of dynamically corrected boundaries is provided. For example... Figure 3 As shown, the water area corresponding to detection point A has three detection layers: A1, A2, and A3. The dashed line in the water area corresponding to detection point A represents the dynamic correction boundary for its depth value. The water area corresponding to detection point B also has three detection layers: B1, B2, and B3. The dashed line in the water area corresponding to detection point B represents the dynamic correction boundary for its depth value. Due to the different hydrological information of the water areas corresponding to detection points A and B, the division of detection layers and the depth corresponding to the dynamic correction boundary differ between the two.

[0130] For detection points whose depth values ​​exceed the dynamic correction boundary, depth value compensation correction is performed.

[0131] A depth value exceeding the dynamic correction boundary indicates that the depth of the detection point is greater than the depth corresponding to the dynamic correction boundary in the water area. In this embodiment, the dynamic correction boundary is used to delineate the vertical interval requiring compensation correction. In the water area below each detection point, the cumulative accuracy gradually decreases as the sonar detection depth increases. If the cumulative accuracy is less than a preset accuracy threshold, the error in the depth value measured by sonar exceeds the allowable range and cannot be ignored. The cumulative accuracy corresponding to the dynamic correction boundary is equal to the accuracy threshold. When the depth value of a certain detection point is below the dynamic correction boundary in the vertical direction, its cumulative accuracy is too low and compensation correction is required.

[0132] The depth value compensation correction specifically includes:

[0133] Mark the detection point whose depth value is less than or equal to the depth corresponding to the dynamic correction boundary as a reference point;

[0134] For any detection point whose depth value needs to be compensated and corrected, the reference depth of the detection point is obtained by interpolation using the depth value of the reference point; alternatively, the reference depth of the detection point is obtained by spline interpolation using the depth value of the reference point near the detection point.

[0135] The depth value of the detection point is filtered and fused with the corresponding reference depth to obtain the depth value after compensation and correction of the detection point.

[0136] Optionally, this embodiment achieves the filtering fusion of depth value and corresponding reference depth through weighted summation, wherein the sum of the weights of depth value and reference depth is 1, and the greater the difference between depth value and the depth corresponding to dynamic correction boundary, the greater the weight value of reference depth.

[0137] Example 2

[0138] This embodiment is the second embodiment of this application; it is based on the same inventive concept as Embodiment 1, and refers to... Figure 2 This embodiment introduces a real-time accuracy correction system for sonar bathymetry, including a data acquisition module, a hydrological prediction module, a clustering module, a bathymetry prediction module, a boundary module, and a compensation correction module; wherein:

[0139] The data acquisition module is used to collect depth values ​​and hydrological information at each detection point in the target water area; the data acquisition module is equipped with necessary detection instruments and equipment such as sonar, sound velocity profiler, optical turbidimeter, and laser particle size analyzer.

[0140] The hydrological prediction module is used to predict and complete the hydrological information of each detection point; specifically, it includes predicting and completing the vertical distribution data of underwater sediment concentration and sediment particle size, so that each piece of hydrological information can cover the entire range of sonar depth sounding.

[0141] The clustering module divides the water area at each detection point into different detection layers based on hydrological information. The clustering module is equipped with a hierarchical clustering algorithm, which can divide the sampling points into different clusters according to the hydrological information of each sampling point, and each cluster corresponds to a detection layer.

[0142] The depth measurement prediction module is used to predict the depth measurement accuracy of each detection layer and calculate the cumulative accuracy of each detection layer. The depth measurement prediction module is equipped with a trained error prediction model, which can predict the significant accuracy of each prediction point based on the layer depth and hydrological information of each detection layer.

[0143] The boundary module constructs a dynamic correction boundary for the depth value based on the depth measurement accuracy and cumulative accuracy.

[0144] The compensation and correction module is used to compensate and correct the depth values ​​of detection points whose depth values ​​exceed the dynamic correction boundary. The module is equipped with an interpolation algorithm that interpolates the depth value to be compensated and corrected based on the depth value of the reference point, obtains the corresponding reference depth, and then filters and fuses the depth value with the reference depth to achieve depth value compensation and correction.

[0145] The specific functions of each module described above are explained in the relevant content of the real-time accuracy correction method for sonar depth sounding described in Example 1, and will not be repeated here.

[0146] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0147] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of protection of this application, and these forms are all within the protection scope of this application.

Claims

1. A method for real-time accuracy correction of sonar depth sounding, characterized in that: Includes the following steps: Simultaneously collect depth values ​​and hydrological information at each detection point in the target water area; The hydrological information includes sediment concentration, sediment particle size, and sound velocity. For any detection point in the target water area, the depth value of the bottom of the water is detected by sonar, and sampling points are set at the same location, starting from the water surface and in a direction perpendicular to the horizontal plane downwards; Vertical distribution data of hydrological information were collected at each sampling point; among them, the vertical distribution data of sound velocity included the sound velocity at all sampling points from the water surface to the bottom; the vertical distribution data of sediment concentration included at least continuous data from the water surface. Sediment concentration at each sampling point; vertical distribution data of sediment particle size including at least continuous values ​​starting from the water surface. The particle size of sediment at each sampling point; , All are positive integers; Hydrological information at each monitoring point is predicted and supplemented, specifically including the prediction and supplementation of underwater sediment concentration and sediment particle size; the method is as follows: S10: For any detection point, mark the sediment concentration or sediment particle size as the target hydrological information; S20: Organize the vertical distribution data of the target hydrological information into a target sequence according to the sampling points from shallow to deep. S30: Extract sequence segments from the target sequence and input them into the trained prediction model; the prediction model outputs predicted values ​​of the target hydrological information; S40: Verify and correct the predicted values ​​of the target hydrological information based on the sound velocity at the corresponding sampling points; The method for verifying and correcting the predicted value of any target hydrological information is as follows: S401: Obtain the sound velocity at the sampling point corresponding to the target hydrological information, and use it as a reference sound velocity; S402: Based on a pre-built hydrological information mapping table, query the reference value range of the target hydrological information corresponding to the reference sound velocity; S403: If the predicted value of the target hydrological information falls within the corresponding reference value range, no adjustment is made; otherwise, execute S404. S404: Correct the predicted values ​​of the target hydrological information based on the corresponding reference value range; S50: Add the predicted values ​​as target hydrological information for the corresponding sampling points to the target sequence; that is, fill the predicted values ​​at the end of the target sequence according to the order of the corresponding sampling points from shallow to dark. S60: Repeat S30-S50 until the target sequence contains the target hydrological information of all sampling points; Based on the hydrological information, the water area corresponding to each detection point is divided into different detection layers; The depth measurement accuracy of each detection layer is predicted based on hydrological information, and the cumulative accuracy of each detection layer is calculated. The predicted depth measurement accuracy for each detection layer specifically includes: Calculate the mean of each hydrological information item from all sampling points in each detection layer, and use it as the hydrological information item for each detection layer. Encode each piece of hydrological information from each detection layer into a feature vector for that layer. Prediction points are uniformly set in the vertical direction in each detection layer; prediction points are also set at the bottom surface of any detection layer; the distance from each prediction point to the top surface of the corresponding detection layer is recorded as the layer depth of each prediction point; The feature vector of the detection layer and the layer depth of each prediction point are input into the trained error prediction model; the error prediction model calculates and outputs the depth measurement accuracy of each detection layer; the depth measurement accuracy of any detection layer includes the significant accuracy of each prediction point; the significant accuracy of any prediction point represents the probability that the depth measurement error of the prediction point is less than the preset depth measurement error threshold when sonar depth measurement is performed starting from the top surface of the corresponding detection layer. The cumulative accuracy of each detection layer is calculated based on the depth measurement accuracy, specifically including: The significant accuracy of the prediction point corresponding to the bottom surface of each detection layer is marked as the minimum accuracy of the detection layer; if the top surface of the detection layer is the water surface of the target water area, the cumulative accuracy is the minimum accuracy of the detection layer; otherwise, the cumulative accuracy is the product of the minimum accuracy of the detection layer and the minimum accuracy of each preceding detection layer; the preceding detection layer is any detection layer above the detection layer. A dynamic correction boundary for the depth value is constructed based on the depth measurement accuracy and cumulative accuracy. For detection points whose depth values ​​exceed the dynamic correction boundary, depth value compensation correction is performed.

2. The sonar depth sounding real-time accuracy correction method as described in claim 1, characterized in that: The hydrological information mapping table contains a reference value range for each type of target hydrological information corresponding to each value of sound velocity; the method for constructing the hydrological information mapping table is as follows. Choose any value of sound speed as the target sound speed; determine the reference value range of any target hydrological information corresponding to the target sound speed, specifically including: Acquire historical data; each historical data point contains a set of sediment concentration, sediment particle size, and sound velocity. Calculate the absolute value of the difference between the sound speed and the target sound speed in each historical data point, and use it as the sound speed difference for each historical data point; set a sound speed difference threshold, and extract historical data with a sound speed difference less than the sound speed difference threshold as reference data; Calculate the Gaussian kernel weight for each reference data based on the sound velocity difference; calculate the weighted mean and weighted standard deviation of the target hydrological information in all reference data based on the Gaussian kernel weight; A reference value range for the target hydrological information is set based on the weighted mean and weighted standard deviation. The reference intervals for each type of target hydrological information corresponding to each value of sound speed are determined sequentially to obtain the hydrological information mapping table.

3. The real-time accuracy correction method for sonar depth sounding as described in claim 2, characterized in that: The method for dividing the water area corresponding to any detection point into different detection layers is as follows: Each type of hydrological information at each sampling point is normalized and encoded into a feature vector; Based on the feature vector of hydrological information of each sampling point, hierarchical clustering is used to divide the sampling points into different clusters; Based on the partitioning of clusters, the water area corresponding to the detection point is divided into different detection layers; each cluster corresponds to a detection layer; the layer height of each detection layer is recorded.

4. The real-time accuracy correction method for sonar depth sounding as described in claim 3, characterized in that: Based on the depth measurement accuracy and cumulative accuracy, a dynamic correction boundary for the depth value is constructed, specifically including: If the cumulative accuracy of any detection layer is less than the preset accuracy threshold, and there is no detection layer above it or the cumulative accuracy of the adjacent detection layer above it is greater than or equal to the preset accuracy threshold, then the corresponding detection layer is marked as a critical layer. Determine the dynamic correction boundary in the water area corresponding to the detection point in the critical layer; specifically including: The cumulative accuracy of each prediction point in the critical layer is calculated sequentially from shallow to deep. If the top surface of the critical layer is the water surface of the target water area, the cumulative accuracy of any prediction point is the significant accuracy of the prediction point. Otherwise, the cumulative accuracy of any prediction point is the product of the significant accuracy of the prediction point and the cumulative accuracy of the adjacent detection layer above the critical layer. The depth corresponding to the prediction point whose cumulative accuracy is closest to the accuracy threshold is marked as the dynamic correction boundary.

5. The real-time accuracy correction method for sonar depth sounding as described in claim 4, characterized in that: A depth value exceeding the dynamic correction boundary indicates that the depth value of the detection point is greater than the depth corresponding to the dynamic correction boundary in the water area; the depth value compensation correction specifically includes: Mark the detection point whose depth value is less than or equal to the depth corresponding to the dynamic correction boundary as a reference point; For any detection point whose depth value needs to be compensated and corrected, the reference depth of the detection point is obtained by interpolation using the depth value of the reference point. The depth value of the detection point is filtered and fused with the corresponding reference depth to obtain the depth value after compensation and correction of the detection point.

6. A real-time accuracy correction system for sonar depth sounding, used to implement the real-time accuracy correction method for sonar depth sounding as described in any one of claims 1-5, characterized in that: It includes a data acquisition module, a hydrological prediction module, a clustering module, a depth sounding prediction module, a boundary module, and a compensation and correction module; among which: The data acquisition module is used to collect the depth value and hydrological information of each detection point in the target water area; The hydrological prediction module is used to predict and complete the hydrological information of each detection point; The clustering module divides the water area at each detection point into different detection layers based on hydrological information; The depth prediction module is used to predict the depth accuracy of each detection layer and calculate the cumulative accuracy of each detection layer. The boundary module constructs a dynamic correction boundary for the depth value based on the depth measurement accuracy and cumulative accuracy. The compensation and correction module is used to compensate and correct the depth values ​​of detection points whose depth values ​​exceed the dynamic correction boundary.

Images