A method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation

Abstract

The present invention relates to the field of marine surveying and mapping technology, and more specifically, to an underwater navigation adaptability improvement method based on curvature direction feature adaptation area aggregation, which is beneficial for eliminating isolated adaptation areas and disconnected adaptation blocks and filling internal voids of adaptation blocks, thereby improving directional adaptability and regional connectivity. The method comprises the following steps: obtaining a gravity anomaly reference map: gridding gravity anomaly data and dividing it using a sliding window method to obtain a gravity anomaly reference map; designing curvature direction features: combining the directional characteristics of maximum and minimum curvatures and the characteristics of Gaussian curvature for evaluating regional gravity anomaly changes to construct curvature direction features; and pre-calibrating adaptation area clustering and implementing an adaptation area reclosing mechanism to obtain a final calibrated adaptation area. Compared with the prior art, the selected area not only has stronger connectivity but also has good directional adaptability, thereby ensuring the positioning accuracy of an underwater vehicle in multiple headings.

Description

Technical Field

[0001] The present invention relates to the field of marine surveying and mapping technology, and more specifically to an underwater navigation adaptability improvement method based on curvature directional feature adaptation zone aggregation, which is beneficial for eliminating isolated adaptation zones, connecting broken adaptation blocks, and filling internal holes of adaptation blocks, thereby improving directional adaptability and regional connectivity. Background Art

[0002] The inertial navigation system (INS) is the core navigation system for underwater vehicles. It uses sensors to estimate the vehicle's true position, velocity, and attitude, achieving short-term, high-precision positioning. However, gyroscope errors accumulate over time, reducing navigation accuracy. Therefore, regular calibration and readjustment of INS system parameters are necessary. Gravity, with its advantages of concealment, stability, and susceptibility to external interference, can provide long-term, stable position information, reliably assisting in correcting INS errors. Therefore, it holds significant research value in the field of underwater assisted navigation.

[0003] Gravity-aided navigation systems use onboard gravimeters to measure gravity information in real time, match it to a gravity reference map, and then use a gravity matching algorithm to correct for accumulated INS errors. However, the positioning performance of underwater gravity-aided navigation systems varies significantly across different navigation areas. Selecting a suitable gravity field adaptation zone can effectively improve underwater vehicle positioning accuracy. Generally, the denser the gravity anomaly contours and the more dramatic the variations, the better the positioning accuracy of the gravity-aided navigation system. However, within the same adaptation zone, underwater submersible positioning performance varies significantly across different navigation directions, with performance in some directions even worse than in non-adaptive zones. Traditional gravity characteristic parameters calibrate the adaptation zone by evaluating the overall gravity variation in the region. Adaptive zones are then selected based on parameter values. However, the parameter construction process and threshold setting rely heavily on subjective judgment. For example, the gravity anomaly correlation coefficient is derived by combining the variation of gravity anomalies in longitude and latitude. The selected adaptation zone is only highly adaptable within a small range of navigation directions, such as those near longitude and latitude. In addition, selecting the adaptation area with strong connectivity can also ensure the path continuity of the underwater vehicle. To this end, it is necessary to construct a gravity matching area selection algorithm with direction adaptability and connectivity. Summary of the Invention

[0004] In response to the shortcomings and deficiencies in the prior art, the present invention proposes a method for improving underwater navigation adaptability based on the aggregation of curvature directional feature adaptation zones, which is beneficial for eliminating isolated adaptation zones, disconnected adaptation blocks, and filling internal voids in adaptation blocks, thereby improving directional adaptability and regional connectivity.

[0005] The present invention is achieved by the following measures:

[0006] A method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation, characterized by comprising the following steps:

[0007] Step 1: Obtain gravity anomaly reference map: grid the gravity anomaly data and divide it using the sliding window method to obtain the gravity anomaly reference map;

[0008] Step 2: Design curvature directional features: Combine the directional features of maximum and minimum curvatures with the Gaussian curvature's ability to assess regional gravity anomaly changes to construct curvature directional features.

[0009] Step 3: Cluster pre-calibration of the adaptation area: Calculate the gravity anomaly characteristics in each area, including: curvature direction characteristics, gravity anomaly standard deviation, roughness, difference entropy, correlation coefficient, and the integral of standard deviation and absolute roughness, and calibrate the adaptation area and non-adaptation area using the k-means clustering method;

[0010] Step 4: Adaptation zone reclosing mechanism: Morphological closing operations are used to enhance the connectivity between adaptation zones and fill holes in the adaptation zone. Isolated adaptation zones are then removed to obtain updated adaptation zones. The distance of the newly added adaptation zones from the cluster center is calculated, and threshold limits are applied to update the adaptation zone status again, enhancing the feasibility of navigation in the adaptation zone.

[0011] Step 5: Get the final calibrated adaptation area.

[0012] In step 1 of the present invention, the gravity anomaly reference map is a three-dimensional gravity anomaly surface of the seabed generated by latitude and longitude grid data and gravity anomaly values ​​at corresponding positions. The corresponding surface curvature can provide adaptation information of gravity anomaly changes and directions.

[0013] In step 2 of the present invention, the curvature of the three-dimensional surface is first defined: suppose the three-dimensional surface embedded in the Euclidean space is , Indicates the longitude grid coordinates corresponding to the parameter domain, indicating the longitude grid coordinates corresponding to the parameter domain The corresponding longitude grid coordinates are Indicates in the parameter domain The corresponding latitude grid coordinates are Indicates that at the grid point The gravity anomaly at the location, the surface exist and The tangent vectors in the directions are , , then the tangent plane can be defined as:

[0014] (1),

[0015] A point on the surface The normal vector is defined as:

[0016] (2),

[0017] Then the point on the surface and normal vector Determined planes can produce intersecting arcs , the degree of curvature is defined as the curvature, and the maximum principal curvature can be obtained by rotating the surface and the minimum principal curvature , then the Gaussian curvature is defined as:

[0018] (3),

[0019] In order to ensure the adaptability of the underwater area in the navigation direction, two conditions must be met at the same time: one is that the overall gravity anomaly in the area fluctuates violently, and the other is that the gravity changes in all directions are similar. However, there is no completely isotropic gravity distribution in the actual sea area, so it is only necessary to make the gravity changes in all directions approximately isotropic, which can be intuitively demonstrated as the gravity anomaly contour lines in the sea area tending to concentric circles. In order to reflect the near circularity of the gravity anomaly changes in the matching area, the maximum principal curvature and the minimum principal curvature are projected on the horizontal plane. and Construct the directional adaptation feature (Curvature directional feature, CD), whose expression is as follows:

[0020] (4),

[0021] in, M and N Respectively represent the number of longitude and latitude that divide the window, is the standard deviation of the Gaussian curvature in the region, which is used to measure the overall gravity anomaly changes in the region. and They are and The angle between the direction and the horizontal plane, It is used to evaluate the adaptability of each heading in the area. The closer it is to 1, the more consistent the direction adaptability of the area is.

[0022] The expressions of the five gravity anomaly characteristic parameters in step 3 of the present invention are as follows:

[0023] (1) Gravity anomaly standard deviation ( ):

[0024] (5),

[0025] in, is the average gravity anomaly in the matching domain;

[0026] (2) Roughness ( ):

[0027] (6),

[0028] (7),

[0029] (8),

[0030] (3) Gravity anomaly difference entropy ( ):

[0031] (9),

[0032] in, Local difference probability, Gravity anomaly difference value;

[0033] (4) Gravity anomaly correlation coefficient ( ):

[0034] (10) ,

[0035] (11),

[0036] (12),

[0037] in, and represent the correlation coefficients in longitude and latitude respectively, is the gravity anomaly correlation coefficient after integrating the two directions;

[0038] (5) The integral of the gravity anomaly standard deviation and the absolute roughness ( ):

[0039] (13),

[0040] (14),

[0041] (15),

[0042] (16),

[0043] The gravity anomaly features are normalized to remove the dimensional influence between different features, and then the k-means algorithm based on Euclidean distance can divide the sample area into a set of adaptation areas when the label is unknown. A and non-adaptive zone setB , and gives the cluster center of the corresponding category.

[0044] In step 4 of the present invention, the adaptation zone reclosing mechanism includes the following steps:

[0045] Step 4-1: Reclosing of the adaptation area: for a given set of adaptation areas A ,Will A The process of expanding each point m into a structural element st is recorded as the expansion operation:

[0046] (17), record the structural element st after it is translated by one unit and still included in A The points that meet the conditions are B right A The erosion operation corresponds to the expression: (18), then A The closing operation with st is defined as:

[0047] (19), where the structural element st is:

[0048] (20),

[0049] After the above-mentioned closing operation is performed on the original adaptation zone samples, the problems of the breaks in the adaptation zone and the holes inside the adaptation zone are alleviated, which is conducive to enhancing the overall connectivity of the adaptation zone. For isolated samples in the adaptation zone, the sample points with only one or two adaptation zones in the eight neighborhoods of the current area are deleted. At this time, most of the newly added areas are selected because they are adjacent to the adaptation zone.

[0050] Step 4-2: Update the adaptability of the closed area: After the closing operation, the non-adaptive samples are converted into the adapted samples, and the set of changed samples is recorded as ,in , the cluster centers corresponding to the adapted and non-adapted samples are and ,calculate and Euclidean distance In the adaptation zone closure mechanism, the newly added area is obtained because the neighborhood location information is obtained and most of them are located on one side of the adaptation zone. With The fan-shaped area with a radius of 2 keeps the newly added samples that are closer to the fitting area as the fitting samples, while the area outside the fan-shaped area is far away from the fitting samples and has a large difference in feature distribution from the fitting samples. Based on this, the newly added samples are calculated. Euclidean distance from the center of the adapted cluster ,when When the added samples Although the samples in the more suitable area deviate from the original ones, they are still within the acceptable range. , indicating the added samples If the vehicle deviates too much from the center of the adaptation cluster and directly navigates in this area, it will cause a large matching positioning error.

[0051] Compared with the existing technology, the present invention takes the directional adaptability and regional connectivity of the underwater gravity navigation area as the dual research objectives, and proposes an adaptation zone aggregation method based on a new curvature direction feature. Specifically, the directional characteristics of the gravity area are evaluated by the maximum and minimum curvature projection values ​​on the horizontal plane. At the same time, Gaussian curvature is introduced to measure the regional gravity change to construct the curvature direction feature. Secondly, the curvature direction feature is combined with multiple traditional gravity anomaly feature parameters to divide the adaptation area and the non-adaptation area by the k-means clustering method. Furthermore, in order to eliminate isolated adaptation areas, connect broken adaptation blocks and fill the internal holes of adaptation blocks, an adaptation zone re-closing mechanism is designed based on neighborhood position information and morphological closing operation, and clustering is used to calculate the curvature direction feature. The adaptability of the closed area of ​​the information update is verified; the simulation results show that the CD-MAC algorithm can select an adaptation area with better directional adaptability and greater connectivity. The curvature direction feature proposed in the present invention is effectively matched in all 6 navigation directions, and achieves the best average matching performance compared with the traditional gravity anomaly feature parameters; under the connectivity marks of 4 neighborhoods and 8 neighborhoods, the adaptation area reclosing mechanism optimizes the connectivity performance by 72.73% and 60.00% respectively; navigation verification is carried out on the selected adaptation area in 180 directions, of which 151 directions are successfully matched, indicating that the selected area not only has strong connectivity, but also has good directional adaptability, which can ensure the positioning accuracy of the underwater submersible in multiple directions. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of the CD-MAC algorithm in the present invention.

[0053] Figure 2 is a schematic diagram of adaptive update of a closed area in an embodiment of the present invention, Figure 2 (a) is the sample status assessment; Figure 2 (b) in the middle is the sample status update.

[0054] Figure 3 is a graph showing different heading matching accuracy in an embodiment of the present invention, wherein Figure 3 (a) is the 1°-90° matching accuracy curve. Figure 3 Middle (b) is the 91°-180° matching accuracy curve. DETAILED DESCRIPTION

[0055] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0056] The algorithm flow of the novel curvature directional feature based matching area clustering method (CD-MAC) proposed in this example is as follows: Figure 1 As shown, the execution steps are as follows:

[0057] Step 1: Obtain a gravity anomaly reference map, grid the gravity anomaly data, and divide it using the sliding window method to obtain a gravity anomaly reference map;

[0058] Step 2: Design the curvature direction feature. Combine the directional characteristics of the maximum and minimum curvatures with the Gaussian curvature's ability to assess regional gravity anomaly changes to construct the curvature direction feature.

[0059] Step 3: Cluster pre-calibration of the adaptation area, calculate the gravity anomaly characteristics in each area: curvature direction characteristics, gravity anomaly standard deviation, roughness, difference entropy, correlation coefficient, integral of standard deviation and absolute roughness, and calibrate the adaptation area and non-adaptation area using the k-means clustering method;

[0060] Step 4: Adaptation zone reclosing mechanism. Morphological closing operations are used to enhance connectivity between adaptation zones and fill holes within the adaptation zone. Isolated adaptation zones are then removed to obtain updated adaptation zones. The distance of the newly added adaptation zones from the cluster center is calculated, and threshold constraints are applied to update the adaptation zone status again, enhancing navigation feasibility within the adaptation zone.

[0061] Step 5: Obtain the final calibration result of the adaptation area. According to the proposed CD-MAC algorithm, the final calibrated adaptation area is obtained and experimental verification is carried out.

[0062] In order to screen out regions with directional adaptability and improve the connectivity of the adaptation zone, this example proposes curvature directional features, adaptation zone pre-calibration and adaptation zone re-closing mechanisms, and then obtains an adaptation zone aggregation method based on the new curvature directional features.

[0063] When underwater submersibles approach the same sea area from different directions, their matching and positioning performance can vary significantly. Traditional gravity feature parameters calibrate the adaptation zone by evaluating the overall gravity variation in the region. These adaptation zones sometimes exhibit significant gravity anomaly variations only in a certain direction. However, in actual navigation, the selected area may not be suitable for the current navigation direction, and its positioning performance may even be worse than that of some non-adaptive zones. For example, the gravity anomaly correlation coefficient is calculated by comprehensively measuring the gravity anomaly variation in longitude and latitude. The selected adaptation zone only has good adaptability within a small range of navigation directions, such as near longitude and latitude. Therefore, it is necessary to design a feature that can evaluate regional directionality.

[0064] Latitude and longitude grid data and gravity anomaly values ​​at corresponding locations can be used to generate a three-dimensional seafloor gravity anomaly surface. The corresponding surface curvature provides information on the variation and direction of the gravity anomaly. Specifically, the Gaussian curvature of a point on the surface measures the degree to which the local surface deviates from the plane, and the two directional principal curvatures provide information on the direction of the point within a certain range. The definition of the three-dimensional surface curvature is as follows:

[0065] Suppose a three-dimensional surface embedded in Euclidean space , represents the longitude grid coordinates corresponding to the parameter domain (u, v), Indicates in the parameter domain The corresponding latitude grid coordinates are Indicates that at the grid point The gravity anomaly at the position, the tangent vectors of the surface S in the u and v directions are , , then the tangent plane is defined as: (1),

[0066] The normal vector of a point P on a surface is defined as: (2),

[0067] Then the point on the surface and normal vector Determined planes can produce intersecting arcs , the degree of curvature is defined as the curvature, and the maximum principal curvature can be obtained by rotating the surface and the minimum principal curvature , then the Gaussian curvature is defined as

[0068] (3);

[0069] In order to ensure the adaptability of the underwater area in the navigation direction, two conditions must be met at the same time: one is that the overall gravity anomaly in the area fluctuates violently, and the other is that the gravity changes in all directions are similar. However, there is no completely isotropic gravity distribution in the actual sea area, so it is only necessary to make the gravity changes in all directions approximately isotropic, which can be intuitively displayed as the gravity anomaly contour lines in the sea area tending to concentric circles. In order to reflect the near-circularity of the gravity anomaly changes in the matching area, the maximum principal curvature and the minimum principal curvature are projected on the horizontal plane. and Construct the direction adaptation feature. Its expression is as follows:

[0070] (4),

[0071] in, M and N Respectively represent the number of longitude and latitude that divide the window, The standard deviation of the Gaussian curvature in the region is used to measure the overall gravity anomaly changes in the region. and They are and The angle between the direction and the horizontal plane, It is used to evaluate the adaptability of each heading in the area. The closer it is to 1, the more consistent the direction adaptability of the area is.

[0072] When selecting a matching region, multiple gravity characteristic parameters must be comprehensively considered. The interaction of multiple gravity characteristics can comprehensively capture gravity field variations while reducing the impact of noise on matching results. The gravity anomaly standard deviation and gravity anomaly difference entropy are used to quantify the distribution complexity and uncertainty of gravity anomalies across the entire region, but they cannot reflect the spatial distribution characteristics of gravity anomalies. Roughness, gravity anomaly correlation coefficient, and the integral of the gravity anomaly standard deviation and absolute roughness first evaluate the spatial gravity variation in the matching region in longitude and latitude, and then combine them to obtain a certain degree of directional adaptability. The expressions for the five gravity anomaly characteristic parameters are shown below.

[0073] (1) Gravity anomaly standard deviation ( )

[0074] (5),

[0075] in, is the average gravity anomaly in the matching domain.

[0076] (2) Roughness (R)

[0077] (6),

[0078] (7),

[0079] (8),

[0080] (3) Gravity anomaly difference entropy (H)

[0081] (9),

[0082] Among them, the local difference probability , gravity anomaly difference value ;

[0083] (4) Gravity anomaly correlation coefficient (RC)

[0084] (10),

[0085] (11),

[0086] (12),

[0087] in, and represent the correlation coefficients in longitude and latitude respectively, is the gravity anomaly correlation coefficient after integrating the two directions;

[0088] (5) The integral of the gravity anomaly standard deviation and the absolute roughness ( );

[0089] (13),

[0090] (14),

[0091] (15),

[0092] (16),

[0093] The gravity anomaly features are normalized to remove the dimensional influence between different features, and then the k-means clustering algorithm based on Euclidean distance can divide the sample area into two categories when the label is unknown, namely the adaptation area set A and non-adaptive zone set B , and gives the cluster center of the corresponding category.

[0094] Adaptation zone reclosing mechanism:

[0095] The selected underwater gravity adaptation zone has good directional adaptability, which can improve navigation matching accuracy. However, due to issues such as gaps between adaptation blocks, internal holes within adaptation blocks, and partial isolation of adaptation blocks, the connectivity of the adaptation zone is reduced. Gravity outliers exhibit a certain degree of regional continuity, so gaps and internal holes (non-adaptation zones) adjacent to the adaptation zone remain of considerable research interest. Specifically, the cost of direct navigation is relatively lower than the cost of turning and detouring to avoid these holes (reduced matching accuracy, increased navigation time, and a longer route). Therefore, it is necessary to appropriately fill in these holes in the non-adaptation zone to maximize navigation accuracy while enhancing path continuity and navigation efficiency. Isolated adaptation zones will be eliminated, primarily because they have a smaller navigation area and require the underwater vehicle to expend significant resources to enter them. Based on this analysis, a mechanism for reclosing the adaptation zone is proposed.

[0096] (1) Reclosing of the adaptation area:

[0097] Morphological closing is the process of sequentially dilating and eroding an image. Closing can fill small holes and breaks in the adaptation area and connect adjacent areas to make them connected.

[0098] For a given set of adaptation zones A ,Will A The process of expanding each point m into a structural element st is recorded as the expansion operation:

[0099] (17), record that the structural element st is still contained in after being translated by one unit A Points, all points that meet the conditions are recorded as B right A The erosion operation corresponds to the expression: (18), then A The closing operation with st is defined as:

[0100] (19), where the structural element st is:

[0101] (20),

[0102] After the above-mentioned closing operation is applied to the original adaptation zone samples, the problems of gaps in the adaptation interval and holes within the adaptation zone are alleviated, which helps to enhance the overall connectivity of the adaptation zone. In addition, for isolated samples in the adaptation zone, sample points with only one or two adaptation zones in the eight-neighborhood of the current area are deleted.

[0103] At this point, newly added areas are mostly selected due to their proximity to the adaptation zone. While the continuity of gravity outliers ensures the adaptability of the matching areas in adjacent locations, it is still necessary to verify whether the added areas deviate too far from the adaptation sample, thereby generating large navigation positioning errors. To this end, relevant clustering information is used to verify whether the newly added areas meet navigation requirements.

[0104] (2) Adaptability update of closed areas

[0105] After the closing operation, the non-adaptive samples are transformed into the adapted samples, and the change sample set is recorded as , where C = A·BA, the cluster centers corresponding to the adapted and non-adapted samples are O1 and O2 respectively, and the Euclidean distance between O1 and O2 is calculated δ In the adaptation zone closure mechanism, the newly added area is obtained due to the neighborhood location information and is mostly located on one side of the adaptation zone. δ The fan-shaped area with a radius of will keep the new samples that are closer to the fitting area as the fitting samples. The area outside the fan-shaped area is far away from the fitting samples and has a large difference in feature distribution with the fitting samples. Based on this, the new samples will be calculated. Euclidean distance from the center of the adapted cluster ,when When the increase Although the sample deviates from the sample in the adaptation zone, it is still within the acceptable range. , indicating the increase The sample deviates too much from the center of the adaptation cluster, and navigating directly in this area will result in a large matching positioning error.

[0106] like Figure 2 As shown in (a), the newly added samples n1 and n2 are located in the fan-shaped area, deviating from the center of the adaptation cluster within the allowable range, and still remain in the adaptation area after the position update. Although samples n3 and n4 are also selected because they are geographically close to the samples in the adaptation area, they deviate too much from the adaptation samples. The spatial distribution characteristics in their area deviate greatly from the actual adaptation area, and the gravity anomaly changes relatively flat. Therefore, the current sample status will be updated and restored to the non-adaptation area, as shown in Figure 2 Middle (b).

[0107] Experimental results and analysis:

[0108] In order to verify the feasibility and effectiveness of the proposed adaptation zone aggregation method based on the new curvature direction feature, experiments were carried out on the directionality and connectivity of the adaptation zone in the sea area.

[0109] Experimental setup:

[0110] Gravity anomaly data with a resolution of 111°E-113°E and 15°N-17°N were selected from the website of the University of California, San Diego and converted into grid resolution data using bilinear interpolation. The average gravity anomaly in this area is 2.70 mGal, with a maximum and minimum of 69.87 mGal and -42.66 mGal, respectively. All experiments were run using Matab2023a software on a computer with a CPU of 3733 MHz and 32 GB of memory. The simulation area was divided into a sliding window of fixed size 100×100 grids and a fixed step size. num ( num =484) subregions. K-means clustering parameters were set to the default parameters in MATLAB software. To test the actual navigation effect of CD-MAC in the selected area, matching and positioning were performed using the TERCOM algorithm.

[0111] Simulation verification and analysis:

[0112] In order to verify the effectiveness of the proposed curvature direction feature in screening the adaptation area, the curvature direction single feature value is arranged in descending order, and the rankings of 1 (area 1), 1 / 4 num (area 2), 1 / 2 num (area 3), 3 / 4 num(area 4), num (area 5) Five representative areas, from six tracks to evaluate the feasibility of the sample area in the navigation direction and the effectiveness of matching positioning, with a grid diagonal length ( m) as the effective matching judgment threshold, 100 independent experiments were conducted on 6 types of tracks in the same area, and the average (mean), standard deviation (std), minimum (min), and maximum (max) of the matching accuracy were used as evaluation indicators. The area with the largest curvature direction characteristic value is Area 1, which is effectively matched in all 6 headings, indicating that Area 1 has good directional feasibility. However, Area 2 only achieves an effective match in track 5, indicating that this area has good positioning performance only in a single heading. In addition, the curvature direction characteristic values ​​gradually decrease from Area 1 to Area 5, and the positioning accuracy of each track and its average value basically shows a gradually increasing trend. This shows that the proposed gravity anomaly feature can not only reflect the near homogeneity of navigation direction, but also take into account the changes in regional gravity anomaly, and has a more comprehensive adaptability assessment capability.

[0113] Arrange each gravity anomaly feature value in descending order and select the optimal region corresponding to the current feature. The optimal region selected by features (1) and (3) is labeled 436, the optimal region corresponding to features (2) and (4) is 437, and feature (5) considers 409 to be the optimal region. The region corresponding to the proposed feature achieves the average optimal matching accuracy in the comprehensive heading. The region that achieves the second-best average matching accuracy is region 437, which can achieve effective matching in 5 headings. The main reason is that features (2) and (4) are constructed by integrating the gravity change features in the longitude and latitude directions. To a certain extent, they can reflect the directionality of the region, but there is a certain distance compared with the evaluation performance of the proposed feature. The remaining features selected by 436 and 409 regions failed to achieve effective matching. The main reason is that features (1), (3) and (5) all evaluate the gravity anomaly changes of the entire region, and the navigation performance in some headings is poor.

[0114] Combining gravity anomaly characteristics, the k-means clustering method was used to classify subregions into two categories: adaptive and non-adaptive regions. The calibrated adaptive regions largely cover areas with dense contour lines, i.e., areas with intense gravity anomaly fluctuations, demonstrating the accuracy of the proposed calibration model. However, the calibration area clearly exhibits isolated individual adaptive regions, internal voids, and fragmented adaptive regions, which are detrimental to the continuous navigation of underwater vehicles. The calibration results of closing and removing isolated points clearly eliminate isolated points, effectively fill internal voids, and connect some fragmented regions, indicating enhanced overall connectivity within the adaptive region. Analysis of the adaptiveness of the closed region and the removed regions revealed sparse contour lines, which are detrimental to high-precision navigation of underwater vehicles. The proposed CD-MAC algorithm ultimately calibrates the adaptive region, enhancing the overall connectivity of the adaptive region. The added adaptation area can effectively fill the holes inside the adaptation area and connect adjacent adaptation blocks, providing more diverse and high-precision navigation solutions for underwater path planning.

[0115] The effectiveness of the CD-MAC algorithm in improving the connectivity of the adaptation zone was tested using connected domain markings of 4 and 8 neighbors. The proposed algorithm added 14 adaptation zones based on cluster calibration, effectively filling the gaps between the different adaptation zones. Under the 4-connected marking, the adaptation zone reclosing mechanism reduced the number of connected blocks from 22 to 6, achieving an optimization rate of 72.73%. Under the 8-connected marking, the number of connected blocks from 10 to 4 was reduced, achieving an optimization rate of 60.00%.

[0116] In order to explore the direction adaptability and navigation reliability of the increased area during long navigation, the navigation performance will be tested for 1°-90° and 91°-180° headings from the starting point [1600,1100] and [1600,1400] respectively. The navigation accuracy under different trajectories is as follows Figure 3 As shown. Figure 3 As shown, all tracks are located within the adaptive region selected by the CD-MAC algorithm, and some tracks are located within the later-added adaptive region. Statistics show that 151 directions (83.89%) were effectively matched, demonstrating that the CD-MAC algorithm can screen out navigation regions with directional adaptability, and that the subsequent transition from non-adaptive regions to adaptive regions has minimal impact on navigation performance.

Claims

1. A method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation, characterized in that: The following steps are involved: Step 1: Obtain gravity anomaly reference map: grid the gravity anomaly data and divide it using the sliding window method to obtain the gravity anomaly reference map; Step 2: Design curvature directional features: Combine the directional features of maximum and minimum curvatures with the Gaussian curvature's ability to assess regional gravity anomaly changes to construct curvature directional features. Step 3: Cluster pre-calibration of the adaptation area: Calculate the gravity anomaly characteristics in each area, including: curvature direction characteristics, gravity anomaly standard deviation, roughness, difference entropy, correlation coefficient, and the integral of standard deviation and absolute roughness, and calibrate the adaptation area and non-adaptation area using the k-means clustering method; Step 4: Adaptation zone reclosing mechanism: Morphological closing operations are used to enhance the connectivity between adaptation zones and fill holes in the adaptation zone. Isolated adaptation zones are then removed to obtain updated adaptation zones. The distance from the newly added adaptation zones to the cluster center is calculated, and threshold limits are applied to update the adaptation zone status again, enhancing the feasibility of navigation within the adaptation zone. Step 5: Get the final calibrated adaptation area.

2. The method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation according to claim 1, characterized in that: In step 1, the gravity anomaly reference map is a three-dimensional seabed gravity anomaly surface generated by the latitude and longitude grid data and the gravity anomaly values ​​at the corresponding positions. The corresponding surface curvature can provide adaptation information of the gravity anomaly change and direction.

3. The method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation according to claim 1, characterized in that: In step 2, we first define the curvature of the three-dimensional surface: let the three-dimensional surface embedded in the Euclidean space be , Indicates in the parameter domain The corresponding longitude grid coordinates are Indicates in the parameter domain The corresponding latitude grid coordinates are Indicates The gravity anomaly value at the grid position, then the tangent vectors of the surface S in the u and v directions are , , then the tangent plane can be defined as: (1), the vector P at a point on the surface is defined as: (2), then the normal vector N on the surface is p The planes of the plane can produce intersecting arcs α(s), the degree of curvature of which is defined as the curvature, and the maximum principal curvature can be obtained by rotating the surface and the minimum principal curvature , then the Gaussian curvature is defined as: (3) To ensure the adaptability of the underwater area in the navigation direction, two conditions must be met at the same time: one is that the overall gravity anomaly in the area fluctuates violently, and the other is that the gravity changes in all directions are similar. However, there is no completely isotropic gravity distribution in the actual sea area. Therefore, it is only necessary to make the gravity changes in all directions approximately isotropic, which is intuitively displayed as the gravity anomaly contour lines in the sea area tending to concentric circle distribution characteristics. In order to reflect the near circularity of the gravity anomaly changes in the matching area, the maximum principal curvature and the minimum principal curvature are projected on the horizontal plane. and Construct the direction adaptation feature, whose expression is as follows: (4), where M and N Respectively represent the number of longitude and latitude that divide the window, The standard deviation of the Gaussian curvature in the region is used to measure the overall gravity anomaly changes in the region. and They are and The angle between the direction and the horizontal plane, It is used to evaluate the adaptability of each heading in the area. The closer it is to 1, the more consistent the direction adaptability of the area is.

4. The method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation according to claim 1, characterized in that: The expression of gravity anomaly characteristic parameter is as follows: (1) Gravity anomaly standard deviation ( ) (5)， in, is the average gravity anomaly in the matching domain, (2) Roughness (R) (6)， (7)， (8)， (3) Gravity anomaly difference entropy (H) (9)， Among them, the local difference probability , gravity anomaly difference value ; (4) Gravity anomaly correlation coefficient (RC) (10)， (11)， (12)， in, and represent the correlation coefficients in longitude and latitude respectively, and RC is the gravity anomaly correlation coefficient after integrating the two directions; (5) The integral of the gravity anomaly standard deviation and the absolute roughness ( ) (13)， (14)， (15)， (16)， The gravity anomaly features are normalized to remove the dimensional influence between different features, and then the k-means clustering algorithm based on Euclidean distance can divide the sample area into two categories when the label is unknown, namely the adaptation area set A and non-adaptive region set B , and gives the cluster center of the corresponding category.

5. The method for improving underwater navigation adaptability based on curvature direction feature adaptation zone aggregation according to claim 1, characterized in that: In step 4, the adaptation zone reclosing mechanism includes the following steps: Step 4-1: Reclosing of the adaptation area: for a given set of adaptation areas A ,Will A The process of expanding each point m into a structural element st is recorded as the expansion operation: (17), record that the structural element st is still contained in after being translated by one unit A Points, all points that meet the conditions are recorded as B right A The erosion operation corresponds to the expression: (18)， but A The closing operation with st is defined as: (19)， Among them, the structural element st is: (20)， After the above-mentioned closing operation is performed on the original adaptation zone samples, the problems of the breaks in the adaptation zone and the holes inside the adaptation zone are alleviated, which is conducive to enhancing the overall connectivity of the adaptation zone. For isolated samples in the adaptation zone, the sample points with only one or two adaptation zones in the eight neighborhoods of the current area are deleted. At this time, most of the newly added areas are selected because they are adjacent to the adaptation zone. Step 4-2: Update the adaptability of the closed area: After the closing operation, the non-adaptive samples are converted into the adapted samples, and the set of changed samples is recorded as , where C = A·BA, the cluster centers corresponding to the adapted and non-adapted samples are O1 and O2 respectively, and the Euclidean distance between O1 and O2 is calculated δ In the adaptation zone closing mechanism, the newly added area is obtained because of the neighborhood location information and is mostly located on one side of the adaptation zone. With δ The fan-shaped area with a radius of 2 keeps the newly added samples that are closer to the fitting area as the fitting samples, while the area outside the fan-shaped area is far away from the fitting samples and has a large difference in feature distribution from the fitting samples. Based on this, the newly added samples are calculated. Euclidean distance from the center of the adapted cluster ,when When the increase Although the sample deviates from the sample in the adaptation zone, it is still within the acceptable range. , indicating the increase The sample deviates too much from the center of the adaptation cluster, and navigating directly in this area will result in a large matching positioning error.

Images