An autonomous identification and encrypted observation method for ocean vertical structure based on Argo mode

Abstract

This invention belongs to the field of marine observation and exploration technology, and discloses a method for autonomous identification and intensive observation of marine vertical structures based on the Argo model, including the following steps: S1. The submersible operates at a preset hovering depth according to the Argo model, and begins to ascend after reaching a preset time; S2. During the ascent, the temperature, salinity, and pressure profiles of seawater are collected according to a conventional sampling strategy to obtain profile data; S3. After the submersible completes conventional sampling, the profile data obtained in this cycle is preprocessed and analyzed to identify the locations of key structural layers in multiple key structural layers in the ocean; S4. An intensive observation window for the descent phase is generated based on the location of the key structural layers, and a corresponding descent observation strategy is generated; S5. The submersible conducts targeted intensive observations near the key structural layers during the descent according to the descent observation strategy, and after completing the intensive observation, it continues to descend to the preset hovering depth and enters the next cycle.

Description

Technical Field

[0001] This invention belongs to the field of marine observation and marine exploration technology, specifically relating to a method for autonomous identification and intensified observation of marine vertical structures based on the Argo model. Background Technology

[0002] Argo buoys and their derivative submersibles typically employ a cyclical operating mode, rising from a hovering position to conduct profiling observations, communicating at the surface, and then descending back to the hovering layer. This is a crucial technology for long-term, continuous ocean temperature and salinity profiling. These observation platforms usually drift with the current at a predetermined hovering depth, surfacing after a certain period to measure marine environmental parameters such as temperature, salinity, and pressure. From the surface, they transmit data and receive control commands via satellite links before descending again to the predetermined hovering depth to begin the next cycle. This mode has become an important component of modern ocean observation systems and is widely used in fields such as marine environmental monitoring and climate change research.

[0003] Existing Argo and similar submersible observation models typically employ a pre-defined fixed sampling strategy in their engineering implementation. This involves conducting regular observations during the ascent phase according to pre-set depth intervals, pressure intervals, or stratified sampling tables. This type of method has advantages such as simple control logic and mature program implementation, and can meet the needs of large-scale, long-term standardized ocean profile observations. However, with the increasing demand for detailed ocean structure observations, the limitations of fixed sampling strategies are becoming increasingly apparent.

[0004] Traditional Argo models typically employ a fixed procedure for surfacing observation, surface communication, and dive return, lacking the ability to dynamically adjust sampling strategies for the next stage based on profile information acquired in the current cycle. While most existing systems can transmit data and switch tasks at the sea surface, their observation parameters are often pre-set before deployment or can only be adjusted by shore-based systems over a longer timescale. This makes it difficult to identify key structural layers in the current water body in real time based on the newly acquired profile results within a single cycle and immediately translate these identification results into targeted, intensified sampling strategies for the dive phase. Furthermore, considering the characteristics of ocean profile data, identifying key structural layers is not a simple matter of determining single-parameter thresholds but involves a comprehensive analysis of multiple physical quantities such as temperature, salinity, pressure, and their derived sound velocity. Different structural layers correspond to different physical mechanisms: mixing layers are often identified by density or temperature changes relative to a reference layer threshold; thermoclines are typically identified by extreme values ​​of temperature or temperature gradients and their continuous bands; and sound velocity minima usually require a comprehensive assessment combining the effects of temperature, salinity, and pressure on the sound velocity profile. In addition, measured profile data often suffer from near-surface disturbances, local noise, and fine-scale oscillations. Existing fixed profile processing and fixed-layer-interval sampling methods are insufficient to meet the real-time identification and observation control requirements of key structural layers for online decision-making by autonomous underwater vehicles.

[0005] In summary, while existing technologies have the capability for long-term autonomous ocean profile observation based on the Argo model, they still have shortcomings in the following aspects: first, they lack real-time identification capabilities for key structural layers; second, they lack the ability to dynamically generate submersible intensive observation strategies using the results of this cycle of observations; and third, they lack a key structural layer identification method that balances physical significance and platform feasibility. Therefore, there is an urgent need to propose a method for identifying and intensively observing key ocean structural layers suitable for Argo model submersibles, in order to improve the observation resolution and resource utilization efficiency near key structural layers without altering the basic operating mode of the existing platform. Summary of the Invention

[0006] This invention aims to address the shortcomings of existing technologies and provides the following solutions: An autonomous identification and encrypted observation method for marine vertical structures based on the Argo model includes the following steps: S1. Make the submersible operate at the preset hovering depth in Argo mode, and start to rise after reaching the preset time; S2. During the ascent, the temperature, salinity, and pressure profiles of seawater are collected using conventional sampling strategies to obtain profile data; S3. After the submersible completes routine sampling, the profile data obtained in this cycle is preprocessed and analyzed to identify the locations of key structural layers in multiple key structural layers in the ocean; S4. Generate an encrypted observation window for the diving phase based on the location of the key structural layer, and generate a corresponding diving observation strategy; S5. The submersible conducts targeted and intensive observations near the key structural layer during the descent according to the descent observation strategy. After completing the intensive observations, it continues to descend to the preset hovering depth and enters the next cycle.

[0007] Preferably, S1 includes: The submersible first drifts or hovers at a preset Argo mode, the hovering depth being preset according to mission requirements, the observed sea area environment, and the platform's operating mode; The submersible maintains a preset dwell time at the hovering depth. When any preset condition is met, the hovering phase ends and the ascent phase begins. The preset conditions include: reaching a preset cycle time, meeting a preset trigger condition, or receiving a corresponding control command.

[0008] Preferably, in the conventional sampling strategy, the pressure profile is sampled using segmented fixed pressure interval sampling: , in, This represents pressure sampling data. Indicates the first r The initial pressure of each pressure segment Indicates the first r Each pressure segment corresponds to a fixed pressure sampling interval. n Indicates the first r The sampling sequence number of each pressure segment.

[0009] Preferably, the method for preprocessing and analyzing the profile data includes: The original observation sequence was constructed based on the profile data: , in, Indicates pressure coordinates, Indicates the first i Temperature at each measuring point Indicates the first i Salinity at each measuring point Indicates the first i Pressure at each measuring point Indicates the first i The observation time at each measuring point N Indicates the number of observation points; The original observation sequence was subjected to quality control and monotonicity adjustment to obtain effective profile data: , in, Indicates pressure coordinates, Indicates the first j Temperature at each measuring point Indicates the first j Salinity at each measuring point Indicates the first j Pressure at each measuring point M Indicates the number of valid observation points; The effective profile data is mapped to a unified standard grid to obtain the profile quantity to be smoothed. ; The profile to be smoothed is then smoothed to obtain the smoothed profile: , in, This represents the smoothed profile measurement. Indicates the preset smoothing operator, m This indicates the half-width of the window.

[0010] Preferably, the method for identifying the location of the key structural layer includes: For the hybrid layer, the reference layer is used as the standard value. The density increment of the density at different depths relative to the reference layer is calculated. When the density increment reaches or exceeds a preset threshold and the continuity condition is met, the position where the threshold is first crossed is determined as the hybrid layer depth, and the hybrid layer position is obtained. For the thermocline, the temperature profile is smoothed, the vertical gradient of the smoothed temperature is calculated, the gradient band where the main thermocline is located is identified within a preset search depth range, and the center depth of the gradient band where the main thermocline is located is determined as the thermocline depth to obtain the thermocline position. For the sound velocity minima, a sound velocity profile is calculated based on temperature, salinity, and pressure, and the sound velocity profile is smoothed. Local minima are searched in the smoothed sound velocity profile. When a local minima satisfies that the rise on both sides is greater than a preset significance threshold, the corresponding depth is determined as the sound velocity minima depth, and the location of the sound velocity minima is obtained.

[0011] Preferably, the method for obtaining the encrypted observation window includes: For any critical structural layer The recognition result is represented as follows: , and ,in, MLD Indicates a hybrid layer. TMD Indicates the thermocline. MSL This indicates the layer where the speed of sound is minimum. Indicates the upper boundary of the critical structural layer. Indicates the central layer of the critical structural layer. Indicates the lower boundary of the critical structural layer; The basic encrypted observation window corresponding to the identification result is: , in, Indicates the basic encrypted observation window; Due to uncertainties in structural layer identification, and the presence of significant transitional features near the upper and lower boundaries of key structural layers, a certain margin is added outside the basic encrypted observation window to obtain the final encrypted observation window: , in, Indicates an encrypted observation window. This represents the upward expansion margin of the critical structural layer. This represents the downward expansion margin of the critical structural layer.

[0012] Preferably, during the descent, when the current depth enters the upper boundary of the window, the sampling mode is switched from default sampling to encrypted sampling; when the submersible leaves the lower boundary of the window, the default sampling strategy is restored. If multiple encrypted observation windows have been merged in the previous stage, the submersible will continuously perform encrypted sampling within the merged unified window to avoid frequently switching sampling modes between adjacent windows; if multiple encrypted observation windows are independent of each other, the submersible will trigger the encrypted sampling of the corresponding windows in order of depth.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention improves the observation resolution near key structural layers in the ocean. By acquiring conventional temperature, salinity, and pressure profiles during the submersible's ascent phase, identifying the locations of key structural layers such as the mixing layer, thermocline, and minimum sound velocity layer during the sea surface phase, and then conducting targeted and intensive observations around these key structural layers during the descent phase, this invention can concentrate limited observational resources on key layers and their adjacent areas, thereby improving the ability to characterize the vertical fine structure and stratification features of the ocean.

[0014] This invention enables adaptive allocation of observation resources. Unlike traditional fixed sampling strategies, this invention utilizes the profile information acquired during the ascent phase of the current cycle to dynamically generate observation windows and sampling schemes for the descent phase, allowing the sampling resolution to be adjusted according to the current vertical structure of the ocean. This prioritizes sampling resources in water layers with significant changes in physical structure, reducing redundant sampling in areas with weaker structural changes and improving overall observation efficiency.

[0015] This invention balances robustness in identification with engineering feasibility. The key structural layer identification method of this invention has clear physical meaning and good interpretability; by introducing mechanisms such as profile preprocessing, smoothing, and continuity constraints, the algorithm's adaptability to measured profile noise, local anomalies, and discrete sampling conditions is improved, thus making it suitable for real-time discrimination and control of underwater vehicles in online operating environments.

[0016] Finally, this invention is well-suited to the operational procedures of existing Argo-mode underwater vehicles. This invention does not alter the original basic mission structure of the underwater vehicle; instead, while maintaining the original periodic autonomous observation mode, it introduces key structural layer identification and adaptive adjustment mechanisms in the surfacing post-processing and descent sampling control stages. Therefore, this invention has good compatibility with existing Argo-mode platforms, facilitating engineering implementation and widespread application. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram of segmented fixed pressure interval sampling according to an embodiment of the present invention; Figure 3 This is a schematic diagram of hybrid layer recognition according to an embodiment of the present invention; Figure 4 This is a schematic diagram of thermocline identification according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the sound velocity minima identification in an embodiment of the present invention; Figure 6 This is a schematic diagram of the overlapping and merging of encrypted observation windows according to an embodiment of the present invention, wherein (a) is a schematic diagram before merging and (b) is a schematic diagram after merging. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example: In this embodiment, as Figure 1 As shown, a method for autonomous identification and intensified observation of marine vertical structures based on the Argo model includes the following steps: S1. Make the submersible operate at the preset hovering depth in Argo mode, and start to rise after reaching the preset time.

[0022] S1 includes: first, the submersible drifts or hovers at a preset Argo mode at a hovering depth, the hovering depth being preset according to mission requirements, the observed sea area environment, and the platform's operating mode; the submersible maintains a preset hovering time at the hovering depth, and when any preset condition is met, the hovering phase ends and the ascent phase begins, the preset conditions including: reaching a preset cycle time, meeting a preset trigger condition, and receiving a corresponding control command.

[0023] In this embodiment, the submersible first drifts or hovers at a preset Argo mode. The hovering depth can be preset according to mission requirements, the observed sea area environment, and the platform's operating mode, and is denoted as [insert denoting depth here]. The submersible maintains a preset dwell time at this depth, and when the preset cycle time is reached... When the preset trigger conditions are met, or when a corresponding control command is received, the hovering phase ends and the floating phase begins.

[0024] For example, the submersible operates according to the core Argo mode, with a cycle time set to approximately 10 days. Hover depth set to The maximum profile depth is set to After completing the sea surface communication in the previous cycle, the submersible descended to a depth of [depth missing]. It then drifts and hovers; after reaching the preset cycle time, it continues to descend from the hovering layer to... This serves as the starting depth for profiling, from which the submersible rises to the sea surface to collect conventional temperature, salinity, and pressure profiles. After sea surface communication concludes, the submersible dives back down to begin the next cycle.

[0025] S2. During the ascent, the temperature, salinity, and pressure profiles of seawater are collected according to the conventional sampling strategy to obtain profile data.

[0026] In this embodiment, the submersible rises to the sea surface from its hovering depth or the initial profile depth, and during the ascent, it performs profile observations of seawater temperature, salinity, and pressure according to a preset conventional sampling strategy, forming ascent observation profile data. The conventional sampling strategy can be preset according to the platform's mission configuration, denoted as... .

[0027] In conventional sampling strategies, pressure profiles are sampled using segmented fixed pressure intervals. This involves employing different pressure sampling intervals within different pressure ranges to balance the accuracy of identifying key upper ocean structural layers with cost control for mid-to-deep ocean observations. Let the pressure be P, then segmented fixed pressure interval sampling can be expressed as: , in, This represents pressure sampling data. Indicates the first r The initial pressure of each pressure segment Indicates the first r Each pressure segment corresponds to a fixed pressure sampling interval. n Indicates the first r The sampling sequence number for each pressure segment. Different pressure segments can use different sampling sequences. This constitutes a segmented conventional sampling profile.

[0028] During the ascent phase, the submersible employs the following segmented fixed pressure interval sampling strategy: sampling at 5-dbar intervals in the 0–100 dbar range, at 10-dbar intervals in the 100–200 dbar range, at 20-dbar intervals in the 200–500 dbar range, at 50-dbar intervals in the 500–1500 dbar range, and at 100-dbar intervals below 1500 dbar. Figure 2 As shown.

[0029] Correspondingly, the observed pressure sequence can be represented in segmented form as follows:

[0030] in, , , , and These represent the initial pressure values ​​for the corresponding segments. The temperature, salinity, and pressure measurements acquired by the submersible during its ascent are denoted as follows: T , S and P This corresponds to the formation of a temperature profile. Salinity profile and pressure profile The surface observation profile data reflects the current cycle's vertical thermohaline structure and pressure distribution characteristics of the ocean, serving as the fundamental input for subsequent key structural layer identification and the generation of submerged intensive observation strategies.

[0031] S3. After the submersible completes routine sampling, the profile data obtained in this cycle is preprocessed and analyzed to identify the locations of key structural layers in multiple key structural layers in the ocean.

[0032] After the submersible completes routine sampling, the ascent observation profile data acquired in this cycle is preprocessed to form a standardized profile input suitable for identifying key structural layers. Preprocessing includes: selecting valid ascent profile segments, removing outliers, missing values, and invalid measurement points, adjusting the monotonicity of the observed pressure sequence, resampling or interpolating the original profile according to a preset standard pressure layer, and smoothing temperature, salinity, or their derivatives.

[0033] In this embodiment, the method for preprocessing and analyzing profile data includes: Construct the original observation sequence based on the profile data: , in, Indicates pressure coordinates, Indicates the first i Temperature at each measuring point Indicates the first i Salinity at each measuring point Indicates the first i Pressure at each measuring point Indicates the first i The observation time at each measuring point N This indicates the number of observation points.

[0034] The original observation sequence was subjected to quality control and monotonicity processing to obtain effective profile data: , in, Indicates pressure coordinates, Indicates the first j Temperature at each measuring point Indicates the first j Salinity at each measuring point Indicates the first j Pressure at each measuring point M This indicates the number of valid observation points.

[0035] Convert the effective profile data to a unified standard grid to obtain the profile quantity to be smoothed. In this embodiment, taking pressure as an example, a standard pressure layer is used, which can be represented as follows:

[0036] in, Indicates the first k The pressure value corresponding to each standard pressure layer; The starting pressure of the standard pressure grid can be taken as the minimum pressure value in the effective profile or as a preset shallow starting pressure. k Indicates the standard pressure layer number; K Indicates the maximum number of the standard pressure layer; This indicates the preset grid interval. Original temperature, salinity, pressure, and other parameters can be mapped to the standard grid via interpolation or binning averaging. For temperature, salinity, sound velocity, or their derived profiles used for identifying critical structural layers, local smoothing can be performed to reduce the impact of local noise and high-frequency disturbances on layer determination.

[0037] The profile to be smoothed is then smoothed to obtain the smoothed profile: , in, This represents the smoothed profile measurement. Indicates the preset smoothing operator, m This indicates the half-width of the window. The above preprocessing improves the stability and consistency of subsequent identification of the mixing layer, thermocline layer, and minimum sound velocity layer.

[0038] In a specific implementation, after the submersible reaches the sea surface, the ascent observation profile acquired in this cycle is preprocessed according to the following steps: First, the effective ascent segment is extracted from the entire round of observation records. Temperature, salinity, and pressure observation data obtained by the submersible during its continuous ascent from the profile's initial depth to the sea surface are selected, and discontinuous ascent data points caused by short-term platform descent, stagnation, or repeated sampling are removed. Second, the extracted ascent segment data undergoes quality control, deleting measurement points with missing values, values ​​outside the physical range, or bad instrument markings in temperature, salinity, or pressure, while also removing isolated anomalous jump points. Then, the remaining effective observation points are monotonically sorted according to pressure coordinates and resampled according to a unified standard pressure layer, so that the profile data is mapped onto a preset standard pressure grid. The standard pressure grid is consistent with the conventional ascent observation strategy. For observation values ​​that do not fall exactly on the standard pressure layer, linear interpolation can be used to project temperature, salinity, and pressure onto the standard pressure layer, thereby obtaining a standardized profile.

[0039] Furthermore, when gradient calculation, threshold search, or layer interpolation is required, a unified analysis grid is constructed based on the standardized profile described above, without altering the amount of original observation information. A 1-dbar analysis pressure grid is used to linearly interpolate the standardized profile to facilitate subsequent continuous determination of structural layers. Finally, the derived profiles used for identifying key structural layers are smoothed.

[0040] After profile preprocessing, the locations of three key oceanic structural layers—the mixing layer, thermocline, and sonic minima—were identified based on temperature, salinity, and pressure profiles. Methods for identifying the locations of these key structural layers included: (1) Hybrid layer recognition For the hybrid layer, the reference layer is used as the standard value. The density increment of the density at different depths relative to the reference layer is calculated. When the density increment reaches or exceeds the preset threshold and the continuity condition is met, the position where the threshold is first crossed is determined as the hybrid layer depth, and the hybrid layer position is obtained.

[0041] Let the density of a pressure layer in a standardized profile be... The reference layer depth is The density increment relative to the reference layer can be expressed as: , in, Indicates density increment, This represents the density value at the reference layer. When a certain depth exists... Make If, after a certain depth, several consecutive measurement points still satisfy the threshold condition, then the position where the threshold is first crossed is taken as the mixing layer depth (MLD). This represents the density increment threshold.

[0042] like Figure 3 As shown, the reference layer can be a 10m depth layer or other preset shallow reference depths; the density increment threshold can be any one of 0.01kg / m³, 0.03kg / m³, or 0.05kg / m³; the continuity condition can be that after the threshold is first exceeded, at least two or more consecutive measuring points continue to meet the threshold condition. By introducing the reference layer and the continuity condition, misjudgment of the mixed layer depth due to local disturbances in the near-surface layer or a single abnormal measuring point can be avoided.

[0043] Determining the central layer of the mixed layer Based on this, the upper and lower boundaries of the mixing layer transition zone are further identified. The bottom of the mixing layer is not a single depth point in a strict sense, but rather a depth range where homogeneous mixed water gradually transitions to stable stratified water. Therefore, providing only a single mixing layer depth value is insufficient to fully describe this structure. This embodiment divides the density difference transition zone into three parts: the upper boundary, the central layer, and the lower boundary. Using this as the central criterion, the lower and higher proportion values ​​are selected as the criteria for the start and end of the transition zone of the hybrid layer, respectively. Upper boundary of the hybrid layer. Central layer and lower boundary They respectively satisfy:

[0044] in, , The typical value in this embodiment is: , That is, the density increment corresponding to the upper boundary of the mixed layer reaches... The initial position indicates the beginning of the transition zone at the bottom of the mixed layer; the density increment at the central layer reaches [a certain value]. The position indicates the location of the main body of the hybrid layer; the lower boundary corresponds to the density increment reaching The location indicates that the transition from the hybrid layer to the underlying stable junction region is essentially complete. In this way, the hybrid layer identification result is no longer just a single depth value, but can provide the range of the transition zone at the bottom of the hybrid layer.

[0045] In practical calculations, since the profile data is obtained through discrete sampling, the threshold crossing position usually does not fall exactly on a particular observation layer. To improve the continuity of layer location, this embodiment uses linear interpolation to process the threshold crossing interval. The threshold crossing position is taken as the center layer of the mixed layer. For example, let's assume the depth There are: , In depth There are: , The central layer of the mixed layer can then be calculated as follows: , Similarly, the upper boundary of the hybrid layer and lower boundary The threshold can also be replaced with... and Positioning is achieved by using linear interpolation between adjacent measuring points.

[0046] (2) Thermocline identification For the thermocline, the temperature profile is smoothed, the vertical temperature gradient after smoothing is calculated, the gradient band where the main thermocline is located is identified within a preset search depth range, and the center depth of the gradient band where the main thermocline is located is determined as the thermocline depth, thus obtaining the location of the thermocline. Figure 4 As shown.

[0047] Let the temperature at a certain depth in the standardized profile be... To reduce the impact of local noise and fine-scale oscillations on gradient calculation, the system first performs local smoothing on the temperature profile to obtain a smoothed temperature profile. .

[0048] This embodiment can employ local smoothing methods such as moving average and median smoothing. Taking moving average as an example, for...i Each depth layer is assigned a corresponding temperature or temperature value. The smoothed value is Then we have: , in, Indicates the length of the smooth window. When... When the temperature profile exceeds the boundary range, methods such as shortening the window, endpoint padding, or boundary value extension can be used. The smoothing window should not be too small, otherwise it will be difficult to reduce local noise; nor should it be too large, otherwise it may excessively weaken the gradient characteristics of the thermocline itself. Considering both the vertical resolution of the Argo profile and the stability of layer identification, this embodiment uses a 5-point moving average to smooth the temperature profile in a typical setting.

[0049] Its vertical gradient can be expressed as: , in, This represents the vertical gradient of temperature change with depth after smoothing. This represents the temperature after smoothing. In a coordinate system where depth is positive downwards, if the temperature decreases with increasing depth, the temperature gradient is usually negative. When the temperature drops rapidly within a certain depth range, this region corresponds to a significant negative gradient band, i.e., the region where the main thermocline is located.

[0050] To avoid interference from near-surface disturbances in thermocline identification, the system searches for the main thermocline within a preset search depth range. Let the thermocline search range be: ,in, Indicates the upper bound of the search range. This indicates the lower bound of the search range. The upper bound of the search range can be set to a certain distance below the depth of the blending layer, for example... It can also be set to below a fixed shallow threshold according to task requirements; the lower limit of the search range can be set to 300m, 500m or other upper ocean depth limits.

[0051] Within this search range, the system first seeks the location of the strongest negative gradient in the smooth temperature profile. That is: , in, This indicates the location of the strongest negative gradient. This indicates the depth location of the strongest negative gradient. This location reflects the layer where the temperature decreases most rapidly with depth and can serve as a candidate center for the main thermocline. In actual ocean profiles, the thermocline is usually not a single depth point but a continuous gradient band with a certain thickness. Therefore, this embodiment further searches for the location of the strongest negative gradient within a preset search depth range and identifies its neighboring continuous main negative gradient bands centered on this location. The upper endpoint of the main negative gradient band is defined as the upper boundary of the thermocline. The lower endpoint is defined as the lower boundary of the thermocline. The midpoint between the two is defined as the central layer of the thermocline. ,Right now: , in, This indicates the location where the main thermocline begins to appear. This indicates the location where the main thermocline ends. This indicates the location of the main body of the thermocline. Thus, the thermocline identification result not only includes a center depth but also provides the vertical thickness range of the thermocline.

[0052] The dominant negative gradient band can be determined by the proportion threshold of the strongest negative gradient. Let the strongest negative gradient be... Since it is a negative value, the continuous depth range adjacent to the strongest negative gradient and still maintaining a significant negative gradient can be considered as the main range of the thermocline. In this embodiment, the main negative gradient band can be determined using the following criterion: , in, This indicates the preset scaling factor. This represents a certain percentage threshold relative to the strongest negative gradient. In this embodiment, we take... The continuous negative gradient interval where the gradient intensity reaches 70% or more of the strongest negative gradient is identified as the main thermocline gradient zone. The upper and lower endpoints of this interval are respectively designated as... and The central position as .

[0053] In actual cross-sections, if multiple local negative gradient bands exist, the system preferentially selects the point containing the strongest negative gradient. The continuous gradient band is chosen as the main thermocline, rather than other weaker gradient bands. This avoids misidentifying localized small-scale temperature fluctuations as the main thermocline. For cases where the gradient band boundary falls between adjacent discrete depth layers, linear interpolation can also be used. and Fine-grained positioning is used to improve the continuity of boundary recognition.

[0054] (3) Sound velocity minima layer recognition For the sound velocity minima, a sound velocity profile is calculated based on temperature, salinity, and pressure, and then smoothed. Local minima are searched within the smoothed sound velocity profile. When a local minima satisfies the condition that the rise on both sides exceeds a preset significance threshold, the corresponding depth is determined as the sound velocity minima depth, thus obtaining the location of the sound velocity minima. Figure 5 As shown.

[0055] In this embodiment, the sound velocity at a certain depth in the standardized profile is assumed to be... Using empirical formulas, the sound velocity profile can be expressed as: , in, , and Representing depth z Temperature, salinity, and pressure at the location, This represents the preset sound speed calculation function.

[0056] In this embodiment, the speed of sound can be calculated using an empirical formula: , Because measured temperature-salt pressure profiles may contain local noise or single-point anomalies, directly searching for minima in the original sound velocity profile can easily lead to misjudgments. Therefore, before identifying the sound velocity minima, the system first performs local smoothing on the sound velocity profile. Let the smoothed sound velocity profile be... The system first searches for candidate points of local minima within a preset search depth range. When a certain depth... satisfy At that time, it can be These are considered as candidate locations for local minima in the sound velocity profile. This indicates the adjacent search step size or local neighborhood scale. This condition is used to determine whether the sound velocity at a certain depth point is lower than the sound velocity of the water layers above and below it, thereby initially screening possible candidate points for sound velocity minima.

[0057] Relying solely on local minima is insufficient to guarantee that the identification results have clear physical meaning. In actual profiles, local minima may appear due to noise, interpolation errors, or weak fluctuations. To avoid misclassifying such weak minima as effective sound velocity minima, this embodiment further introduces a saliency judgment. Let the representative sound velocities in the adjacent water layers on either side of the local minima be... and Then the rise in sound velocity on both sides of the local minimum can be expressed as: , in, This indicates the magnitude of the sound velocity rise relative to the upper adjacent water layer, representing the minimum value. This represents the rise in sound velocity relative to the adjacent water layer below the minimum value. To ensure that this minimum value has a complete low-sound velocity valley structure, this embodiment requires sufficient rise amplitude on both its left and right sides, i.e.: , in, This represents a preset significance threshold. When a local minimum simultaneously meets the above conditions, the system determines that it is sufficiently significant and identifies its corresponding depth as the central layer of the sound velocity minimum layer, denoted as: In practical calculations, the significance of a local minimum can be represented by the smaller of the rise in sound velocity on either side of it: , in, This indicates the significant validity of the minimum sound speed. Only when... Only then is the candidate minimum value recognized as the center of the effective sound velocity minimum layer. The layer at which the center of the sound velocity minimum layer is determined... Subsequently, this embodiment further identifies the continuous low-sound-velocity intervals corresponding to this layer, and defines the upper and lower endpoints of this interval as the upper and lower boundaries of the sound-velocity minimum layer. Using the local minimum sound speed at a given location as a benchmark, search for a continuous low-sound speed range in its vicinity that satisfies the following formula:

[0058] in, This is a proportionality coefficient used to control the range of expansion in the low-sound velocity range. In this embodiment, we take... The boundary criterion for the low-sound speed interval is defined as half the amplitude of the rise from the local minimum to both sides. A continuous interval satisfying this condition indicates that the sound speed is still near its local low value and can be considered the main range of the sound minimum layer. The upper endpoint of this continuous low-sound speed interval is defined as the upper boundary of the sound minimum layer, denoted as […]. The lower endpoint is defined as the lower boundary of the sound velocity minimum layer, denoted as . .

[0059] The preset search range can be determined based on the typical sound velocity structure of the research sea area and the mission requirements. For example, when focusing primarily on the upper sound velocity structure, the search range can be limited to below the mixing layer, extending to within several hundred meters or a thousand meters; if focusing on deep sound channels or the mid-deep sound propagation environment, the search range can be further extended to deeper water layers. To avoid misjudgments caused by near-surface noise and weak sound velocity disturbances within the mixing layer, local minimum searches are usually limited to below the preset minimum search depth.

[0060] If multiple local minima satisfying the significance condition exist within the search range, the system can, according to... The candidate points are sorted by their magnitude, and the candidate point with the highest significance and clearest structure is selected as the center layer of the final sound velocity minima. If no local minima satisfy the significance threshold, it is determined that there is no effective sound velocity minima in this cycle, and the corresponding sound velocity layer densification observation window is not triggered.

[0061] S4. Generate encrypted observation windows for the diving phase based on the location of key structural layers, and generate corresponding diving observation strategies.

[0062] After identifying the positions of the mixing layer, thermocline layer, and minimum sound velocity layer, the system further generates a encrypted observation window for the descent phase based on the identification results of each key structural layer. In this embodiment, the upper boundary, central layer, and lower boundary of each key structural layer have been determined during the aforementioned identification process. The actual vertical range of the key structural layer can be used as the basis, combined with necessary safety margins and sampling resolution settings, to form an encrypted sampling interval for the submersible's descent.

[0063] In this embodiment, the method for obtaining the encrypted observation window includes: For any critical structural layer The recognition result is represented as follows: , and ,in, MLD Indicates a hybrid layer. TMD Indicates the thermocline. MSL This indicates the layer where the speed of sound is minimum. Indicates the upper boundary of the critical structural layer. Indicates the central layer of the critical structural layer. Indicates the lower boundary of the critical structural layer; The basic encrypted observation window corresponding to the identification result is: , in, This represents the basic encrypted observation window; this window indicates the main vertical range where the key structural layer itself is located. Considering the certain uncertainty in structural layer identification, and the important transitional features near the upper and lower boundaries of the key structural layer, this embodiment adds a certain margin outside the basic window to form the final encrypted observation window. Let the key structural layer be... X The upper and lower expansion margins are respectively and The final encrypted observation window can then be represented as: Due to uncertainties in structural layer identification, and the presence of important transitional features near the upper and lower boundaries of key structural layers, a certain margin is added outside the basic encrypted observation window to obtain the final encrypted observation window: , in, Indicates an encrypted observation window. This represents the upward expansion margin of the critical structural layer. This represents the downward expansion margin of the critical structural layer. and The settings can be configured according to the observation mission requirements and platform resource status. For example, for the mixing layer, the upper part of the water body inside the mixing layer can be covered, and the lower part of the stable stratification initiation region can be covered; for the thermocline layer, the main gradient band and its upper and lower transition regions can be covered; for the sound velocity minima layer, the sound velocity trough and its upper and lower sound velocity recovery regions can be covered. Correspondingly, the densification windows for the three types of key structural layers can be written as follows: , in, , and These represent the submerged intensive observation windows corresponding to the mixing layer, thermocline layer, and minimum sound velocity layer, respectively.

[0064] Regarding sampling resolution settings, different structural layers can adopt different sampling intervals based on their physical significance and observation requirements. The bottom of the mixing layer typically exhibits smaller scale changes and can be set with a higher sampling resolution; the thermocline layer corresponds to the main gradient band where temperature changes rapidly and should also be sampled with a higher resolution; the minimum sound velocity layer is usually related to the underwater acoustic propagation environment, and its depth may be relatively deep, so the window width can be appropriately increased, and the sampling interval can be set to a medium density based on mission requirements and platform resources.

[0065] In a typical configuration of this embodiment, the following sampling strategy can be adopted: the sampling interval within the encryption window of the mixed layer is set to 1m; the sampling interval within the encryption window of the thermocline layer is set to 2m; and the sampling interval within the encryption window of the minimum sound velocity layer is set to 10m. For depths not falling within any encryption window, the submersible still uses the default sampling interval for observation. This forms a submersible observation mode of high-resolution sampling within the critical structural layer and conventional sampling within non-critical water layers.

[0066] Because different key structural layers in actual ocean profiles may be close to each other, especially the bottom of the mixing layer and the upper part of the thermocline, where there is often a continuous transition relationship, the system also needs to merge multiple densification windows. Let the two densification observation windows be: , in, This indicates the first encrypted observation window. This indicates the second encrypted observation window. Display window The upper boundary depth, Display window The lower boundary depth, Display window The upper boundary depth, Display window The lower boundary depth.

[0067] When satisfied or At this time, the two can be merged into a unified window: , in, Indicates the same window, This indicates the preset spacing threshold corresponding to the window merging criterion. The sampling strategy for repeated segments follows the higher-frequency strategy in the original encrypted observation scheme. The purpose of window merging is to avoid the complexity of control logic and redundant observations caused by repeated switching of sampling modes between multiple adjacent windows during the dive. Window merging is as follows: Figure 6 As shown.

[0068] Ultimately, the set of encrypted observation windows generated by the system can be represented as follows: , Each window corresponds to a specific depth range, sampling interval, and trigger source. This set of windows will serve as the direct basis for sampling control during the descent phase. During descent, the submersible determines whether to enter a specific encrypted window based on the current depth: if it enters the window range, it switches to the corresponding encrypted sampling interval; if it is outside the window, it maintains the default sampling strategy.

[0069] S5. The submersible conducts targeted and intensive observations near key structural layers during the descent, in accordance with the descent observation strategy. After completing the intensive observations, it continues to descend to the preset hovering depth and enters the next cycle.

[0070] During the descent, when the current depth enters the upper boundary of the window, the sampling mode is switched from default sampling to encrypted sampling; when the submersible leaves the lower boundary of the window, the default sampling strategy is restored; if multiple encrypted observation windows have been merged in the previous stage, the submersible will continuously perform encrypted sampling within the merged unified window to avoid frequent switching of sampling modes between adjacent windows; if multiple encrypted observation windows are independent of each other, the submersible will trigger encrypted sampling of the corresponding windows in order of depth.

[0071] In this embodiment, after the sea surface profile data processing, key structural layer identification, and encrypted observation window generation are completed, the submersible enters the dive observation execution phase. During this phase, the submersible determines in real-time whether it has entered a specific encrypted observation window based on its current depth. If the current depth is within the encrypted window, it switches to the corresponding high-resolution sampling strategy; if it is outside the window, it maintains the default sampling strategy. The sampling interval can be uniformly represented as: , in, Indicates the first i The sampling interval corresponding to each encryption window This indicates the default sampling interval for the unencrypted range. Normally, The sampling interval is smaller than the default interval to ensure higher vertical observation resolution near key structural layers.

[0072] During execution, the submersible determines in real time whether it has entered a specific encrypted observation window based on current pressure or depth. When the current depth enters the upper boundary of the window, the system switches the sampling mode from default sampling to encrypted sampling; when the submersible leaves the lower boundary of the window, the system reverts to the default sampling strategy. If multiple encrypted observation windows have been merged in the previous stage, the submersible continuously performs encrypted sampling within the merged unified window, avoiding frequent switching of sampling modes between adjacent windows. If multiple windows are independent of each other, the system triggers encrypted sampling of the corresponding windows sequentially according to depth.

[0073] After completing targeted encrypted observations during the descent phase, the submersible continues to move to the preset hovering depth and re-enters a drifting and hovering state. This step marks the end of this round of Argo mode intelligent observation cycle and is also the starting preparation phase for the next observation cycle. Upon returning to the hovering depth, the submersible can simultaneously record the execution status of this round of observation tasks. The recorded information includes whether encrypted observations were triggered in this round, the type of key structural layer triggered, the depth of the mixing layer, the depth of the thermocline layer, the depth of the minimum sound velocity layer, the actual encrypted observation windows executed, the number of sampling points in each window, and the task execution status. This information can be used for the next round of task management, shore-based data analysis, and subsequent control parameter optimization.

[0074] Overall, the return to hovering depth stage ensures the integrity and continuity of the intelligent observation and control process. This stage not only enables the submersible to recover its state after completing a single observation mission but also provides a unified starting point for the next cycle. Combining the aforementioned stages of surfacing observation, key structural layer identification, encrypted window generation, and descent encrypted observation, this embodiment forms a closed-loop intelligent observation process suitable for Argo mode submersibles, providing an implementable control framework for subsequent long-term autonomous observation and engineering applications.

[0075] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for autonomous identification and intensified observation of marine vertical structures based on the Argo model, characterized in that, Includes the following steps: S1. Make the submersible operate at the preset hovering depth in Argo mode, and start to rise after reaching the preset time; S2. During the ascent, the temperature, salinity, and pressure profiles of seawater are collected using conventional sampling strategies to obtain profile data; S3. After the submersible completes routine sampling, the profile data obtained in this cycle is preprocessed and analyzed to identify the locations of key structural layers in multiple key structural layers in the ocean; S4. Generate an encrypted observation window for the diving phase based on the location of the key structural layer, and generate a corresponding diving observation strategy; S5. The submersible conducts targeted and intensive observations near the key structural layer during the descent according to the descent observation strategy. After completing the intensive observations, it continues to descend to the preset hovering depth and enters the next cycle.

2. The method for autonomous identification and intensified observation of marine vertical structures based on the Argo model according to claim 1, characterized in that, S1 includes: The submersible first drifts or hovers at a preset Argo mode, the hovering depth being preset according to mission requirements, the observed sea area environment, and the platform's operating mode; The submersible maintains a preset dwell time at the hovering depth. When any preset condition is met, the hovering phase ends and the ascent phase begins. The preset conditions include: reaching a preset cycle time, meeting a preset trigger condition, or receiving a corresponding control command.

3. The method for autonomous identification and intensified observation of marine vertical structures based on the Argo model according to claim 1, characterized in that, In the conventional sampling strategy, the pressure profile is sampled using segmented fixed pressure intervals: ， in, This represents pressure sampling data. Indicates the first r The initial pressure of each pressure segment Indicates the first r Each pressure segment corresponds to a fixed pressure sampling interval. n Indicates the first r The sampling sequence number of each pressure segment.

4. The method for autonomous identification and intensified observation of marine vertical structures based on the Argo model according to claim 1, characterized in that, The methods for preprocessing and analyzing the profile data include: The original observation sequence was constructed based on the profile data: ， in, Indicates pressure coordinates, Indicates the first i Temperature at each measuring point Indicates the first i Salinity at each measuring point Indicates the first i Pressure at each measuring point Indicates the first i The observation time at each measuring point N Indicates the number of observation points; The original observation sequence was subjected to quality control and monotonicity adjustment to obtain effective profile data: ， in, Indicates pressure coordinates, Indicates the first j Temperature at each measuring point Indicates the first j Salinity at each measuring point Indicates the first j Pressure at each measuring point M Indicates the number of valid observation points; The effective profile data is mapped to a unified standard grid to obtain the profile quantity to be smoothed. ; The profile to be smoothed is then smoothed to obtain the smoothed profile: ， in, This represents the smoothed profile measurement. Indicates the preset smoothing operator, m This indicates the half-width of the window.

5. The method for autonomous identification and intensified observation of marine vertical structures based on the Argo model according to claim 1, characterized in that, The method for identifying the location of the key structural layer includes: For the hybrid layer, the reference layer is used as the standard value. The density increment of the density at different depths relative to the reference layer is calculated. When the density increment reaches or exceeds a preset threshold and the continuity condition is met, the position where the threshold is first crossed is determined as the hybrid layer depth, and the hybrid layer position is obtained. For the thermocline, the temperature profile is smoothed, the vertical gradient of the smoothed temperature is calculated, the gradient band where the main thermocline is located is identified within a preset search depth range, and the center depth of the gradient band where the main thermocline is located is determined as the thermocline depth to obtain the thermocline position. For the sound velocity minima, a sound velocity profile is calculated based on temperature, salinity, and pressure, and the sound velocity profile is smoothed. Local minima are searched in the smoothed sound velocity profile. When a local minima satisfies that the rise on both sides is greater than a preset significance threshold, the corresponding depth is determined as the sound velocity minima depth, and the location of the sound velocity minima is obtained.

6. The method for autonomous identification and intensified observation of marine vertical structures based on the Argo model according to claim 1, characterized in that, The methods for obtaining the encrypted observation window include: For any critical structural layer The recognition result is represented as follows: , and ,in, MLD Indicates a hybrid layer. TMD Indicates the thermocline. MSL This indicates the layer where the speed of sound is minimum. Indicates the upper boundary of the critical structural layer. Indicates the central layer of the critical structural layer. Indicates the lower boundary of the critical structural layer; The basic encrypted observation window corresponding to the identification result is: ， in, Indicates the basic encrypted observation window; Due to uncertainties in structural layer identification, and the presence of significant transitional features near the upper and lower boundaries of key structural layers, a certain margin is added outside the basic encrypted observation window to obtain the final encrypted observation window: ， in, Indicates an encrypted observation window. This represents the upward expansion margin of the critical structural layer. This represents the downward expansion margin of the critical structural layer.

7. The method for autonomous identification and intensified observation of marine vertical structures based on the Argo model according to claim 1, characterized in that, During the descent, when the current depth enters the upper boundary of the window, the sampling mode will be switched from default sampling to encrypted sampling. Once the submersible leaves the lower boundary of the window, the default sampling strategy is restored. If multiple encrypted observation windows have been merged in the previous stage, the submersible will continuously perform encrypted sampling within the merged unified window to avoid frequently switching sampling modes between adjacent windows; if multiple encrypted observation windows are independent of each other, the submersible will trigger the encrypted sampling of the corresponding windows in order of depth.

Images