A siltation early warning method based on fusion of a towed sonar and multiple source hydrological parameters

Abstract

This invention discloses a siltation early warning method based on the fusion of towed sonar and multi-source hydrological parameters, belonging to the field of underwater topographic monitoring and engineering safety early warning technology. The method first plans the survey lines for the target area and simultaneously collects multibeam sonar bathymetry data, CTD multi-parameter data, platform attitude data, tow cable status data, positioning and navigation data, and time synchronization information during towed platform operation. Then, it performs time alignment, outlier removal, missing value compensation, and noise suppression on the multi-source data. A sound velocity profile is constructed based on the CTD multi-parameter data, and the sound velocity of the multibeam sonar bathymetry data is corrected. Geometric compensation is also performed by combining platform attitude data, tow cable status data, and positioning and navigation data. Finally, the corrected bathymetry data is... The results are mapped to a preset coordinate system to construct an underwater cross-section model, a grid elevation model, or a bed sediment point cloud model for the current time period. Historical baseline models are retrieved for coordinate unification, overlapping area identification, and adaptive registration. Based on this, elevation difference analysis, slope change analysis, regional thickness statistics, and calculations of sediment thickness, changed area, and sediment volume are performed. Furthermore, short-term trend predictions are made by combining historical observations from multiple time periods with environmental driving factors such as flow velocity, tide level, temperature, salinity, turbidity, and sediment content. Finally, a warning level is generated based on a risk threshold, and warning information is output. This invention improves the comparability of multi-time period towed sonar measurement results and the accuracy of sediment identification, enabling continuous monitoring, quantitative analysis, and proactive warning of bed sedimentation status in target areas. It is applicable to port channels, inland waterways, estuaries, reservoir areas, and near-shore engineering areas.

Description

Technical Field

[0001] This invention relates to the fields of underwater topographic monitoring, acoustic detection, hydrological environment perception, and engineering safety early warning technology. Specifically, it relates to a sedimentation early warning method based on the fusion of towed sonar and multi-source hydrological parameters, which is applicable to the monitoring, risk identification, and graded early warning of sedimentation in port channels, inland waterways, estuaries, reservoir areas, and nearshore engineering areas. Background Technology

[0002] The underwater bed morphology of port channels, inland waterways, estuaries, reservoirs, and near-shore engineering areas is constantly evolving due to factors such as inflow and sediment conditions, tidal changes, engineering disturbances, ship navigation, and localized siltation. When rapid siltation, cross-sectional contraction, or abnormal accumulation occurs in a localized area of ​​the bed, it not only reduces navigation capacity but may also adversely affect dredging scheduling, hydraulic structure operation, safety management, and engineering maintenance. Therefore, continuous monitoring, quantitative analysis, and early warning of underwater topographic changes in target areas have become crucial technical requirements for waterway operation and engineering management.

[0003] In existing technologies, underwater sedimentation monitoring typically employs manual bathymetry, fixed-section measurement, single-beam bathymetry, multi-beam bathymetry, side-scan sonar detection, or other underwater topographic surveying methods to obtain seabed elevation information at a specific time or during a particular expedition. Regional erosion and sedimentation changes are then assessed by comparing measurement results from different periods. Some technical solutions also incorporate hydrological parameters such as temperature, salinity, and depth to correct acoustic measurement results and improve bathymetry accuracy. While these methods have some application value in basic surveys and local topographic investigations, they still have significant shortcomings in continuous sedimentation monitoring, short-term change identification, and proactive early warning. On the one hand, most existing measurement methods focus on single-operation results, emphasizing obtaining the seabed morphology at a specific moment, lacking the ability to uniformly model continuous temporal changes. Especially in towed measurement scenarios, changes in the measurement platform's attitude, tow cable disturbances, track deviations, flow field influences, and changes in the acoustic environment can all introduce geometric distortions and positioning errors, making it difficult to directly compare measurement results from different times with high precision. If only simple superposition or cross-sectional subtraction methods are used, measurement errors can easily be misjudged as actual sedimentation changes, thus reducing the reliability of the analysis results. On the other hand, existing technologies utilize historical data in a rather crude manner, typically using historical measurement results only as a reference background, lacking unified coordinate constraints, source region identification, and adaptive registration processing for the relationship between historical baselines and current observations. Since the location, sampling density, attitude, and coverage of survey lines often differ between different expeditions, it is difficult to accurately extract local bed uplift, slope changes, cross-sectional contraction, and anomalous accumulation areas without first establishing comparable baseline relationships. It is also difficult to further achieve stable estimations of sedimentation thickness, changed area, and sedimentation volume.

[0004] Most existing technologies remain at the stage of "post-event measurement" and "post-event analysis," mainly used to generate measurement result maps or explain existing topographic changes, lacking the ability to predict short-term trends within pre-set time windows. For waterway maintenance, dredging scheduling, and risk control, simply knowing whether siltation has occurred is often insufficient to support proactive decision-making. It is more necessary to combine recent multi-time observation results with environmental factors such as flow velocity, temperature, salinity, and turbidity to judge the siltation development trend in the future and generate actionable tiered early warning information. Existing technologies have not yet formed a complete technical chain in this regard, from data acquisition, cross-section reconstruction, baseline registration, change detection, trend prediction to early warning output. Therefore, existing technologies still lack a siltation early warning method that can be applied to towed sonar operation scenarios, collaboratively integrate multibeam bathymetry data with multi-source hydrological parameters such as CTD, and on this basis, achieve continuous underwater cross-section reconstruction, accurate historical baseline registration, quantitative detection of siltation changes, short-term trend prediction, and tiered early warning output. To address the aforementioned issues, it is necessary to propose a new technical solution to improve the comparability of measurement results from different time points, the accuracy of siltation identification, and the initiative and practicality of early warning decisions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, such as insufficient comparability of repeated measurements from towed sonar, difficulty in unifying and registering historical measurement results, low accuracy in identifying sedimentation changes, and a lack of short-term trend prediction and proactive early warning capabilities, this invention provides a sedimentation early warning method based on the fusion of towed sonar and multi-source hydrological parameters. This method constructs a complete technical chain encompassing towed multibeam sonar bathymetry, CTD multi-parameter acquisition, sound velocity correction, attitude and track compensation, underwater cross-section or point cloud reconstruction, historical baseline registration, cross-section change detection, sedimentation estimation, short-term trend prediction, and tiered early warning output. This enables continuous monitoring, quantitative analysis, and early warning of sedimentation conditions on the seabed in target water areas.

[0006] The purpose of this invention is to provide a siltation early warning method applicable to port channels, inland waterways, estuary channels, reservoir areas, and near-shore engineering areas. By synergistically fusing towed sonar bathymetry data with CTD hydrological parameters, platform attitude parameters, track positioning parameters, and historical baseline data, an underwater topographic representation model that can be uniformly compared across different voyages and time scales is constructed. Based on this model, the method accurately extracts the areas of sedimentary changes, siltation thickness, siltation area, siltation volume, and their evolution trends. An early warning level is then generated based on preset thresholds, thereby improving the initiative and reliability of waterway operation and maintenance management, navigation safety assurance, and dredging scheduling decisions.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a siltation early warning method based on towed sonar and multi-source hydrological parameter fusion, comprising the following steps: S1, conducting target area task planning and multi-source data synchronous acquisition. Based on the spatial range of the water area to be monitored, the location of key cross-sections, historically silt-prone areas, and operational requirements, the towed survey line, speed range, sampling frequency, and repeated observation cycle are pre-set; during the operation of the towed platform, multi-beam sonar bathymetry data, CTD multi-parameter data, platform attitude data, tow cable status data, positioning and navigation data, and time synchronization information are synchronously acquired. The CTD multi-parameter data includes at least one or more of temperature, salinity, and water depth; the platform attitude data includes one or more of roll, pitch, heading, and heave; and the positioning and navigation data includes one or more of global navigation satellite positioning data, inertial navigation data, speed information, and track information.

[0008] S2. Perform raw data preprocessing and quality control. Time alignment, outlier removal, missing value compensation, and noise suppression are performed on the collected raw sonar echo data, multi-source attitude data, and hydrological parameter data. A unified time reference is established based on the sampling frequency differences of different sensors. Data segments affected by transient disturbances, local loss of lock, abnormal attitude changes, or invalid echoes are marked, masked, or weighted to obtain a standardized input dataset that can be used for subsequent calculations.

[0009] S3. Perform sound velocity correction and geometric compensation based on CTD parameters. Calculate the water sound velocity profile based on CTD multi-parameter data, or invert the sound velocity distribution based on measured temperature, salinity, and pressure information, and perform sound ray propagation correction and measurement distance correction on the multibeam bathymetry data; simultaneously, combine platform attitude data, towing status data, and positioning and navigation data to compensate for sonar installation offset, roll error, pitch error, heading deviation, heave effects, and track drift, eliminating bathymetry distortion caused by platform movement, tow body sway, and changes in the water environment.

[0010] S4. Reconstruct underwater cross-sections or seabed point clouds. Map the sound velocity correction and geometric compensation results to a preset coordinate system to generate a cross-section model, raster elevation model, or seabed point cloud model for the target area at the current time. During reconstruction, outlier removal, overlapping survey line fusion, local smoothing, sparse region interpolation, and boundary clipping can be further performed to improve cross-section continuity, point cloud integrity, and local terrain representation accuracy. Depending on the task requirements, the reconstruction results can be output as a single cross-section, a set of cross-sections, a two-dimensional depth map, or a three-dimensional point cloud.

[0011] S5. Perform historical baseline retrieval and adaptive registration. Retrieve historical cross-section models, historical point cloud models, or historical topographic baseline data corresponding to the current target area, and establish a registration relationship between the current time-based model and the historical baseline model. The registration process includes one or more of the following: coordinate unification, temporal attribute association, overlapping area identification, stable reference area extraction, feature point extraction, cross-section morphology matching, and iterative optimization correction. The stable reference area is a region with relatively small changes in the bedrock that can serve as a registration anchor point. The feature points include slope breakpoints, valley points, peak points, boundary points, and other representative geometric feature points. Through the registration process, a benchmarked cross-section model or benchmarked point cloud model that can be compared with the same source is obtained.

[0012] S6. Perform cross-sectional change detection and sedimentation estimation. Based on the registered current model and historical baseline model, perform elevation difference analysis, local slope change analysis, cross-sectional contraction analysis, and regional thickness statistics on the corresponding cross-section, grid, or point set to identify subgrade uplift zones, subgrade subsidence zones, abnormal deposition zones, and possible scour zones. Further calculate the average sedimentation thickness, maximum sedimentation thickness, changed area, local sedimentation volume, and total regional sedimentation based on the spatial range and elevation difference of the changed area, and form a sedimentation change result set.

[0013] S7. Construct multi-time series data and predict short-term trends. Correlate the current sedimentation change results with historical observations from multiple time periods to construct a time series change dataset for the target area, target section, or key risk zone. Further integrate information on flow velocity, tide level, temperature, salinity, turbidity, sediment content, operational disturbances, and other environmental driving factors to extract sedimentation growth rate, duration of change, acceleration of change, fluctuation range, and risk evolution characteristics. Based on this, establish a short-term trend prediction model to estimate the sedimentation development trend within a preset time window, preferably 24 to 72 hours.

[0014] S8. Conduct risk assessment and graded early warning output. Based on the target area's siltation thickness, siltation volume, cross-sectional change rate, development trend, duration, and environmental driving conditions, compare with preset early warning thresholds to generate corresponding risk levels. These risk levels can be divided into normal, attention, early warning, and severe early warning, or according to business needs, into blue, yellow, orange, and red early warning levels. Finally, output the early warning results, change location, trend information, recommended handling level, and related monitoring maps to the upper-level business platform, monitoring terminal, or management system to achieve visualized display and linked alarms for siltation risks.

[0015] Preferably, in step S1, the synchronous acquisition of multi-source data adopts a unified timestamp mechanism or a master-slave clock calibration mechanism to ensure the data correspondence between multi-beam sonar, CTD, inertial navigation, GNSS and towed state sensors; when there are multiple data streams with inconsistent sampling frequencies, a unified data sequence is established by interpolation mapping, nearest neighbor alignment or weighted alignment.

[0016] Preferably, in step S2, the quality control of the raw sonar data also includes the identification and removal of abnormal echo intensity, isolated depth points, repeated jump points, discontinuous echo bands, and obvious out-of-limit values; for locally missing areas, neighborhood constraint compensation, trajectory correlation compensation, or multi-flight fusion compensation can be used to improve data integrity.

[0017] Preferably, in step S3, the sound velocity profile obtained based on CTD data is used not only to correct the depth measurement distance, but also to correct the refraction deviation caused by the sound ray propagation path at different depth layers, so as to reduce the measurement error in the edge area of ​​multibeam coverage under complex hydrological conditions; for cases where there is a clear layered structure in different water layers, segmented compensation can be performed according to the layered sound velocity model.

[0018] Preferably, in step S4, the underwater cross-section or bottom point cloud reconstruction process adopts a combination of survey line-level fusion and region-level fusion. First, a local cross-section result is formed for a single survey line, and then multiple adjacent survey lines are spliced ​​together as a whole to improve the continuity of regional topographic representation. For key risk areas, a local high-resolution cross-section model can also be constructed to extract cross-section morphological changes in a more refined manner.

[0019] Preferably, in step S5, the adaptive registration not only considers the overall coordinate transformation relationship, but also further considers the local morphological differences of the cross section. Through reference area constraints, feature point matching, local translation correction, local elevation deviation calibration and iterative convergence optimization, the spatial alignment accuracy between results at multiple times is improved. When significant sedimentation changes occur in some areas, stable areas are selected first to participate in the registration, so as to avoid incorrectly including the real changed areas in the baseline correction process, thereby affecting the change detection results.

[0020] Preferably, in step S6, the sedimentation volume estimation adopts one or more of the following methods: integration based on cross-sectional differences, volume accumulation based on grid elevation differences, or integration based on point cloud local surface differences; for the target cross-section, the average uplift, maximum uplift, net cross-sectional change, and cross-sectional navigation margin change can be output; for the target area, the total sedimentation volume of the area, the local hotspot sedimentation volume, and the migration trend of the sedimentation center can be output.

[0021] Preferably, in step S7, the short-term trend prediction model can be an extrapolation model based on historical time-series change characteristics, a regression prediction model based on environmental driving factors, a trend determination model based on rule thresholds, or a combination of the above methods; the model input includes at least the results of sedimentation changes and environmental parameter changes over multiple consecutive time periods, and the model output includes the sedimentation growth value, growth rate, risk probability, or risk level change trend within a preset time window.

[0022] Preferably, in step S8, the warning threshold is set using a single-index threshold or a multi-index fusion threshold. The single-index threshold is used for independent judgment of siltation thickness, cross-sectional change rate, or regional volume, while the multi-index fusion threshold is used to comprehensively consider siltation scale, growth rate, duration, and environmental conditions to improve the stability and operational adaptability of the warning judgment. When the target area meets the warning conditions, the system can automatically generate warning records, change maps, cross-sectional comparison maps, and disposal suggestions.

[0023] This invention can be further described as a siltation early warning system, comprising: a multi-source data acquisition unit, a preprocessing and quality control unit, a sound velocity correction and geometric compensation unit, a cross-section or point cloud reconstruction unit, a historical baseline management and registration unit, a change detection and siltation estimation unit, a trend prediction unit, and an early warning output unit; the units are sequentially connected or logically interact with each other according to a preset data flow to jointly implement the above-mentioned method steps. The system can be deployed on a towed platform, a shore-based processing terminal, a mobile workstation, a cloud-based business platform, or a combination thereof.

[0024] The present invention may further include an electronic device and a computer-readable storage medium, the electronic device including a processor and a memory, the memory storing a computer program, which, when executed by the processor, is used to implement the siltation early warning method described above.

[0025] Compared with the prior art, the present invention has at least the following beneficial effects: First, the present invention integrates towed sonar depth sounding data with CTD multi-parameter data, attitude data and positioning and navigation data, and improves the authenticity and stability of depth sounding results in complex towed operation scenarios through sound speed correction, attitude compensation and track correction, and reduces the error accumulation caused by environmental and attitude disturbances of single measurement data.

[0026] Second, by establishing an adaptive registration relationship between historical baselines and current observation results, this invention solves the problem of difficulty in direct comparison caused by incomplete overlap of survey lines from different voyages, different sampling densities, changes in attitude state, and local geometric distortions. This enables multi-time section results to have a unified reference basis, thereby significantly improving the accuracy and reliability of sedimentation change identification.

[0027] Third, this invention can not only identify whether the substrate has changed, but also perform quantitative analysis on the range, thickness and volume of the change. It can output key indicators such as average sediment thickness, maximum sediment thickness, change area and sediment volume, improving from a qualitative judgment of "whether it has changed" to a quantitative expression of "how much it has changed, where it has changed, and how much it has affected".

[0028] Fourth, based on change detection, this invention further constructs a multi-time change sequence and introduces environmental driving factors such as flow velocity, tide level, temperature, salinity, and turbidity to achieve short-term prediction of siltation development trends within a preset time window. This extends the monitoring results from post-event analysis to pre-event warning, enhancing the system's support capabilities for waterway maintenance, dredging operations, and engineering operation management.

[0029] Fifth, this invention forms a complete closed-loop technology chain from task planning, multi-source acquisition, data correction, cross-section reconstruction, historical registration, change detection, trend prediction to early warning output. It can adapt to various application scenarios such as ports and waterways, estuaries, inland waterways, reservoirs and nearshore engineering, and has strong engineering applicability, scalability and platform linkage capabilities.

[0030] Sixth, this invention allows for flexible setting of early warning thresholds for different regions, cross sections, and levels according to business needs, and supports key area monitoring, local hotspot tracking, and multi-level linkage alarms. Therefore, it is more conducive to forming a refined, dynamic, and hierarchical siltation risk management mechanism.

[0031] This invention organically integrates towed sonar depth sounding, CTD hydrological sensing, historical baseline registration, cross-sectional change analysis, and short-term trend prediction, overcoming the shortcomings of existing technologies that can only perform single observations, rough comparisons, and post-event analyses. It realizes an active, continuous, and quantitative technical solution for siltation risk identification and early warning decision-making, and has significant technological advancements and practical application value. Attached Figure Description

[0032] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0034] Appendix Figure 1 This is a flowchart of the overall process of a siltation early warning method based on the fusion of towed sonar and multi-source hydrological parameters according to the present invention.

[0035] Appendix Figure 2 This is a flowchart illustrating the core technologies of this invention, which are based on historical baseline registration, cross-sectional change detection, and short-term trend prediction. Detailed Implementation

[0036] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0037] It should be noted that all directional indications in the embodiments of this invention are only used to explain the relative relationships, data flow, or processing order between components in a specific state, and should not be construed as specific limitations on the invention. When the corresponding operating posture, installation method, or deployment environment changes, the relevant directional expressions should also be interpreted accordingly.

[0038] Furthermore, the descriptions used in this invention, such as "first," "second," "current," "historical," "baseline," and "target area," are only used to distinguish different objects, stages, or processes, and should not be construed as limitations on relative importance, sequence, or quantity. Therefore, technical features marked with "first" or "second" may explicitly or implicitly include at least one of those features. The technical solutions of the various embodiments can be combined with each other, but this should be based on the premise that those skilled in the art can implement them; when there are obvious contradictions or impossibilities between the technical solutions, it should be understood that such a combination does not exist and is not within the scope of protection of this invention.

[0039] In underwater sedimentation monitoring and risk early warning scenarios, the continuity, accuracy, and comparability of the system are crucial technical indicators. Faced with the influencing factors such as attitude fluctuations, tow cable swaying, track drift, and changes in sound velocity distribution across different water layers caused by towed platforms under complex flow field conditions, the core problem this invention aims to solve is how to stably obtain high-precision bed cross-sections while ensuring operational efficiency, and how to unify multi-time measurement results under the same comparison benchmark. Especially in port channels, estuaries, reservoir areas, and near-shore engineering areas, bed sedimentation often exhibits characteristics such as continuous evolution, local abrupt changes, and short-term intensification. Traditional single-measurement or simple cross-section subtraction methods are insufficient to meet the engineering requirements for refined monitoring and proactive early warning.

[0040] To address this, this invention proposes a siltation early warning method based on the fusion of towed sonar and multi-source hydrological parameters. By constructing a closed-loop processing flow of "multi-source synchronous acquisition—sound velocity correction and geometric compensation—section / point cloud reconstruction—historical baseline registration—change detection—trend prediction—early warning output," it achieves continuous perception and quantitative analysis of the siltation status of the target area's seabed. This invention not only improves the comparability of measurement results across multiple time periods but also assesses siltation risks within a predetermined time window based on historical change patterns and environmental driving factors, thus providing more proactive and reliable technical support for waterway maintenance, dredging scheduling, engineering operation and maintenance, and safety management.

[0041] Next, we will combine the attached Figure 1 and attached Figure 2 The following provides a detailed description of two embodiments of the present invention. Embodiment 1 mainly corresponds to the overall process of the present invention, focusing on the complete implementation process from task planning to early warning output; Embodiment 2 mainly corresponds to the core technical process of the present invention, focusing on the specific implementation methods of historical baseline registration, cross-sectional change detection, sedimentation estimation, and short-term trend prediction. Example 1

[0042] This embodiment is combined with the appendix Figure 1 This paper describes the overall process of a siltation early warning method based on towed sonar and multi-source hydrological parameter fusion according to the present invention. This embodiment uses a port channel prone to siltation as an example, but the present invention is not limited to this application scenario, and can also be applied to inland waterways, estuaries, reservoirs, and nearshore engineering projects, as well as other areas that require bottom sedimentation monitoring and risk early warning.

[0043] In this embodiment, task planning for the target area is first performed. Based on existing water depth data, historical siltation records, channel boundary ranges, and the distribution of key risk areas, the operating route, repeat survey lines, survey line spacing, operating speed, and observation cycle of the towed platform are set. For example, in a port access channel, several parallel survey lines can be laid out along the main channel axis and its two sides, with additional locally denser survey lines added for bends, berth corners, or historically high siltation areas. Preferably, the repeat observation cycle can be set to once a day, once per tide, or once every few hours, depending on engineering management needs, to form continuous time-series data. During the data acquisition phase, the towed platform operates along the planned survey lines, simultaneously acquiring multibeam sonar bathymetry data, CTD multi-parameter data, platform attitude data, tow cable status data, positioning and navigation data, and unified time synchronization information. The system comprises several components: a multibeam sonar for acquiring bottom echoes and depth information of the target area; a CTD module for collecting parameters such as temperature, salinity, pressure, and depth to establish a sound velocity profile; an attitude sensor for acquiring motion parameters such as roll, pitch, heading, and heave; a tow cable status sensor for reflecting tow body offset, tow cable length, or stress state; and a positioning and navigation unit for providing GNSS positioning, inertial navigation, speed, and track information. To improve the accuracy of multi-source data, a unified master clock is preferred for timestamping, ensuring that different sensor outputs are mapped to the same time reference. During the raw data preprocessing stage, data from different sources undergoes unified processing and quality control. Specifically, invalid pulse identification, abnormal echo intensity removal, isolated point removal, and transition point smoothing are performed on the raw sonar echoes; abnormal sampling detection and depth consistency verification are performed on the CTD data; abrupt change detection, drift constraints, and missing segment compensation are performed on the attitude and navigation data; and synchronization sequences are established for data streams with different sampling frequencies through interpolation mapping, time window alignment, or nearest neighbor matching. For data segments that are significantly distorted due to external interference, their weight can be reduced in subsequent calculations by setting quality markers, or the corresponding segments can be directly masked to avoid the impact of abnormal data on the cross-section reconstruction results. In the sound velocity correction and geometric compensation stage, based on the temperature, salinity, and pressure information obtained from the CTD (Conductivity, Tolerance, and Displacement) data, the sound velocity distribution of the target water area at different depths is calculated, and sound ray propagation correction and refraction compensation are performed on the multibeam sonar bathymetry data accordingly. For water areas with obvious stratification, segmented compensation can be performed according to the stratified sound velocity profiles to reduce the impact of sound velocity differences at different depths on the measurement accuracy of the coverage edge area. Simultaneously, by combining roll, pitch, heading, heave, towed body offset, and track drift parameters, the sonar installation attitude, beam pointing, and bathymetry geometry are corrected to obtain a more accurate set of bed depth points. In the cross-section or point cloud reconstruction stage, the corrected bed depth points are projected onto a preset geographic coordinate system or engineering coordinate system to form a bed point cloud model of the target area at the current time.For scenarios requiring cross-sectional comparison, two-dimensional cross-sectional data can be further extracted according to the survey line direction or a specified cutting direction. For scenarios requiring regional sedimentation estimation, the point cloud can be discretized into a regular grid elevation model. Preferably, during the reconstruction process, local outliers are filtered out again, fusion weighting is performed on overlapping survey line areas, and interpolation compensation is performed on local sparse areas, thereby obtaining a more continuous and complete subsurface representation model. After completing the current time-based model construction, the historical baseline model of the target area is retrieved. The historical baseline model can be the previous measurement result, a reference average model within a certain time period, an acceptance benchmark model after dredging, or a standard cross-sectional model confirmed by the engineering management department. After inputting the current model and the historical baseline model into the registration module, coordinate system one, time attribute association, and spatial coverage matching are first performed, and then the overlapping area and stable reference area between the current measurement result and the historical baseline result are identified. The stable reference area is usually selected from areas with relatively small topographic changes, clear boundary features, and reliable repeated observations, and is used as the registration anchor point between the current model and the historical baseline model. After registration, changes are detected between the current model and the historical baseline model. For cross-sectional data, the elevation difference between each sampling point or sampling interval can be calculated along the corresponding cross-sectional direction to obtain the distribution of cross-sectional uplift and subsidence. For regional raster models, elevation differences can be calculated on the same raster cells to form a difference field. For point cloud models, local surface differences can be calculated using a neighborhood fitting method. Generally, when the current model shows a bed surface uplift relative to the historical baseline model, it can be identified as a siltation trend; when it shows a bed surface subsidence, it can be identified as a scouring trend. Furthermore, indicators such as average siltation thickness, maximum siltation thickness, area of ​​change, location of siltation center, and total siltation volume of the region can be statistically analyzed.

[0044] In this embodiment, to enhance early warning capabilities, the current time-series change results are combined with historical time-series results to form a time-series dataset. For example, the average siltation thickness, maximum siltation thickness, and regional siltation volume of the same target section in the most recent seven observations can be stored in chronological order, and simultaneously correlated with environmental driving factors such as flow velocity, tide level, temperature, salinity, and turbidity. Subsequently, the trend analysis module extracts features such as recent siltation growth rate, growth duration, fluctuation amplitude, and acceleration of change to predict siltation development over the next 24 to 72 hours. In the early warning output stage, the prediction results are compared with pre-set risk thresholds. For example, a state of concern can be triggered when the average siltation thickness in a key waterway area exceeds a first threshold; a state of early warning can be triggered when the siltation volume or maximum siltation thickness exceeds a second threshold and the growth rate continues to rise; and a state of severe early warning can be triggered when the predicted depth will reach a limit affecting navigation safety or engineering operations within a preset time window. The warning results can be output in different levels such as blue, yellow, orange, and red, and at the same time generate cross-section comparison maps, regional difference maps, hot spot location maps and suggested handling instructions, which are pushed to shore-based workstations, monitoring terminals or business management platforms.

[0045] In this embodiment, the method of the present invention can complete offline processing after a single operation, or perform near real-time calculations in a data acquisition and processing mode. For example, during operation, the towed platform can first complete the preprocessing of the raw data, basic sound velocity correction, and local cross-section reconstruction, and then the shore-based processing terminal can perform historical baseline registration, regional sedimentation estimation, and trend prediction to balance processing timeliness and computational complexity. Of course, for mobile platforms with strong processing capabilities, the entire process can also be completed directly on the platform side; the present invention does not limit this.

[0046] As can be seen from this embodiment, the method of the present invention can not only obtain the subsurface morphology of the target area at a certain moment, but also extract the sedimentation trend under a unified reference across multiple time periods, forming a risk warning result for engineering management. Compared with traditional single measurement or simple cross-sectional differential methods, the method described in this embodiment has significant advantages in terms of data comparability, stability of change identification, and proactive warning capability. Example 2

[0047] This embodiment is combined with the appendix Figure 2 This paper provides a detailed explanation of the core technical processes of this invention, including historical baseline registration, cross-sectional change detection, and short-term trend prediction. This embodiment focuses on how, starting from the inputs of the current cross-sectional model and the historical baseline model, reference area extraction, feature matching, registration optimization, homologous cross-section generation, sedimentation estimation, and early warning output are achieved, thus demonstrating the core innovative aspects of this invention.

[0048] In this embodiment, two basic data objects are first input: one is the current time-bound cross-sectional model or point cloud model reconstructed from the towed sonar data of the current voyage, and the other is the historical baseline cross-sectional model or historical point cloud model corresponding to the target area. For ease of subsequent comparison, it is preferable to first spatially crop and unify the format of the two data objects to ensure consistent coordinate representation and unified regional boundary definitions. If the current data and historical data originate from different voyages or different equipment, they can also be converted to the same coordinate reference using unified coordinate transformation parameters. Then, the reference area extraction stage begins. Since there may be real sedimentation or scour changes within the target area, directly using the entire area for overall registration could easily mistake real changes for geometric deviations, thus weakening the change detection results. Therefore, this embodiment preferably first identifies a stable reference area, that is, selecting an area with relatively small terrain changes, clear boundary contours, and stable measurement coverage from the overlapping area of ​​the two time-bound models as the registration constraint area. The stable reference area can be obtained through manual setting, rule filtering, or automatic identification. For example, stable sections of channel slopes, areas near fixed structures, and stable bedrock areas far from historical siltation hotspots can be prioritized as reference areas. During the coordinate unification and time alignment stage, a unified coordinate benchmark check is performed on the current time-based model and the historical baseline model, and a correspondence is established for data from different time sources. For cross-sectional data, a one-to-one correspondence can be established based on cross-sectional number, survey line number, spatial location, and sampling mileage; for point cloud or raster data, alignment can be performed based on region boundaries, grid index, and spatial neighborhood. After this step, the two models have a basis for comparison based on their common source. The next step is feature matching and registration optimization. In this embodiment, representative geometric feature points or segments are preferably extracted within the stable reference area, such as slope breakpoints, local peaks, local valleys, boundary abrupt change points, and points with significant curvature changes. Using these features as constraints, the translational deviation, rotational deviation, and elevation deviation of the current model relative to the historical baseline model can be calculated, and optimized iteratively to achieve higher consistency between the two time-based models within the reference area. For cases with localized irregular distortions, local registration corrections can be overlaid on the overall registration to fine-tune local areas and further improve the alignment accuracy between regions of the same origin. After registration, a common-origin comparison section or a common-origin region model is generated. A common-origin comparison section refers to a data object that can be directly compared between the current time section and the historical baseline section under unified coordinates, unified reference area constraints, and unified sampling rules. Preferably, the common-origin comparison sections can be resampled at the same mileage interval, or the common-origin region model can be rasterized with the same grid resolution to facilitate subsequent differential analysis. During the elevation difference calculation stage, difference calculations are performed on the common-origin comparison sections or the common-origin region model.If the current elevation value increases relative to the historical baseline elevation value, it indicates that the bed surface has risen, which is a potential siltation; if the elevation value decreases, it indicates that the bed surface has fallen, which is a potential scour. For cross-sectional data, the elevation difference can be calculated point by point to form a difference curve; for regional data, the elevation difference can be calculated grid by grid to form an elevation difference field. To improve robustness, this embodiment can set an ignore threshold for excessively small random fluctuations, that is, when the absolute value of the elevation change is less than a preset noise threshold, it is considered as no significant change. In the local slope and thickness change extraction stage, the difference results are further analyzed for local structure. For example, the average siltation thickness, maximum siltation thickness, slope change rate, navigation clearance change, and local contraction degree are extracted on key cross-sections; the area of ​​siltation patches, hotspot center location, range of local high-value areas, and spatial expansion direction are extracted on the regional model. Different weights can be assigned to change areas at different risk levels to support subsequent risk assessment. In the siltation volume integration stage, the volume of the identified siltation areas is calculated. For regular raster models, the total sedimentation volume of a region can be obtained by multiplying the positive elevation difference by the cell area and summing the results. For cross-sectional data, the net sedimentation area of ​​the cross-section can be obtained by integrating the positive elevation difference curve over the horizontal range, and then the regional volume can be estimated by combining the cross-sectional spacing. For point cloud models, a local surface can be established first, and then the envelope volume between the current surface and historical surfaces can be calculated. Preferably, volume accumulation is only performed on regions where the elevation difference is greater than a preset minimum sedimentation threshold to reduce the impact of random noise on the volume estimation results. In the multi-time series construction stage, the average sedimentation thickness, maximum sedimentation thickness, sedimentation volume, sedimentation hotspot location, and slope change rate obtained in the current time are combined with the corresponding indicators from multiple historical time periods to form a structured sequence in chronological order. At the same time, flow velocity, tidal level, temperature, salinity, turbidity, sediment content, and local engineering disturbance information are incorporated into the same time series framework as auxiliary environmental driving variables. In this way, topographic change information is preserved, and external conditions affecting sedimentation evolution are introduced. In the environmental factor fusion and trend prediction stage, trend modeling is performed on the aforementioned time-series data. Specifically, the growth rate, growth acceleration, and fluctuation range can be calculated based on the changes in sedimentation over multiple consecutive time periods; the probability and possible duration of continued sedimentation can be analyzed based on the changing trends of different environmental driving factors; and the predicted sedimentation thickness, predicted volume increment, or risk level change trend within a future preset time window can be further output. Preferably, if multiple recent time periods show a continuous rise, with weakened flow velocity, increased turbidity, and enhanced sediment-laden conditions, it can be determined that the possibility of continued sedimentation development in the short term is high; conversely, if recent changes slow down and are accompanied by enhanced scouring conditions, the warning level can be appropriately lowered. In the threshold comparison and warning output stage, the prediction results are compared with operational thresholds to form a final warning conclusion. The operational thresholds can be configured differently according to the nature of the target area.For example, for main navigation channels, the focus should be on the average siltation thickness affecting the design water depth and the local minimum clearance; for areas near berths, the focus should be on the location of siltation hotspots and the rate of volume accumulation; for nearshore engineering areas, the focus should be on local abrupt slope changes and scouring and silting changes around the foundations. The final output may include: the current siltation level of the target area, the risk trend within a future preset time window, the location of hotspot areas, the recommended re-measurement time, and recommended remedial measures.

[0049] This embodiment further illustrates a typical application process. For example, during three consecutive days of observation of a port channel, the reconstructed bed model from the first day serves as the initial baseline; new current-time models are acquired on the second and third days. After registration and differencing the second-day data with the initial baseline, the average siltation thickness of the local channel is found to be 0.12 meters, and the maximum siltation thickness is 0.26 meters. A joint comparison of the third-day data with the updated baseline from the second day or with the initial baseline reveals that the same hotspot area continues to rise, with the average siltation thickness increasing to 0.21 meters, and the regional siltation volume continuing to increase. Combined with environmental information such as decreased flow velocity and increased turbidity during the same period, the trend prediction module determines that the area still has a high probability of continued siltation within the next 24 hours. Therefore, the area is upgraded from a state of concern to a state of warning, and the results are pushed to the management platform, prompting for intensified re-measurement or dredging preparation. This process demonstrates that the present invention can not only identify changes that have already occurred but also make more forward-looking risk assessments based on the fusion of multi-time and multi-factor results.

[0050] It should be noted that the reference region extraction method, feature matching method, registration optimization strategy, differential threshold setting method, volume integration method, and trend prediction model form described in this embodiment can all be adjusted according to specific application scenarios. For example, the reference region can be determined by manual calibration or by automatic identification; trend prediction can be based on rule-based threshold determination, historical sequence-based fitting extrapolation, or a combination of both. This invention does not limit these aspects; any method that can enhance the comparability between the current time period and the historical baseline, extract siltation changes, and provide early warning output should fall within the protection scope of this invention. Technical terms that require explanation and are helpful in understanding the present invention

[0051] (1) Towed sonar: refers to a sonar device that is deployed behind or below a shipboard platform by a tow body, tow cable or other towing device, and conducts continuous acoustic detection of underwater target areas during navigation. The towed sonar in this invention is preferably a towed multibeam echo sounder, used to acquire bottom echo information, depth information and cross-sectional morphology information of the target area, so as to support subsequent underwater topography reconstruction, cross-sectional change detection and siltation early warning.

[0052] (2) Multibeam echo sounding: This refers to an echo sounding method that uses multiple acoustic beams to perform fan-shaped coverage measurements on the underwater seabed, thereby simultaneously acquiring depth information from multiple measuring points during a single passage. Compared with single-beam echo sounding, multibeam echo sounding has the advantages of large coverage area, high spatial sampling density, and strong terrain representation capability. The multibeam echo sounding data in this invention is used to generate seabed cross-section models, grid elevation models, or point cloud models.

[0053] (3) CTD multi-parameter data: refers to hydrological parameter data collected by conductivity, temperature, and depth measurement devices. In specific applications, it can also be expanded to include salinity, pressure, density, or other aquatic environmental parameters. The CTD multi-parameter data in this invention is mainly used to establish the sound velocity profile of the target water area and to provide data support for sound ray propagation correction, depth sounding accuracy compensation, and environmental driving analysis.

[0054] (4) Sound velocity profile: refers to the distribution relationship of the speed of sound propagation in water as a function of depth. Since factors such as water temperature, salinity, and pressure affect the speed of sound, different depth layers may have different sound velocity values. The sound velocity profile in this invention is used to correct the sound velocity and compensate for refraction in the towed sonar depth sounding results, so as to reduce the measurement error under complex hydrological conditions.

[0055] (5) Attitude data: refers to parameter data characterizing the motion state of the towed platform or towed body, which usually includes one or more of roll, pitch, heading, heave, yaw, roll or pitch. The attitude data in this invention is used to correct the installation attitude, beam pointing and spatial geometry of the towed sonar, thereby improving the accuracy of the bottom depth point positioning.

[0056] (6) Track positioning data: refers to data used to describe the spatial position and motion path of the towed platform or towed body, which typically includes one or more of the following: global navigation satellite positioning data, inertial navigation data, velocity information, heading information, and timestamp information. The track positioning data in this invention is used to achieve spatiotemporal unification between different sensors and to provide position reference for cross-section reconstruction, baseline registration, and comparison of results across multiple time periods.

[0057] (7) Bottom bed point cloud model: refers to a three-dimensional terrain representation model composed of a large number of discrete bottom bed points with spatial coordinates and elevation information. In this invention, the sounding points after sound speed correction, attitude compensation and track correction can be constructed as a bottom bed point cloud model to describe the real terrain state of the target area, and can be further converted into a cross-section model or a grid elevation model.

[0058] (8) Cross-section model: refers to the two-dimensional underwater topographic profile extracted along the preset survey line direction or the specified cutting direction, used to characterize the changes in the seabed elevation at a specific location along the lateral or longitudinal direction. The cross-section model in this invention is used for direct comparative analysis between different time periods, and is used to extract the changes in cross-section uplift, cross-section contraction, local siltation thickness, and other characteristics.

[0059] (9) Historical baseline: refers to a historical reference model, historical cross-section results, historical point cloud results, or standard terrain model used for comparison with the current observation results. The historical baseline can be the measurement results of the previous voyage, the average reference model for a specific period, the acceptance cross-section after dredging, or the confirmed standard terrain data. This invention achieves accurate identification of underwater bed erosion and deposition changes through registration processing between historical baselines and current observation results.

[0060] (10) Registration: refers to the process of unifying topographic data acquired at different times, on different voyages, or from different sources into the same spatial reference frame, so that they have a direct comparable relationship. The registration in this invention includes coordinate unification, overlapping area identification, reference area extraction, feature matching, deviation correction, and iterative optimization, which are used to improve the comparability of measurement results from multiple time periods.

[0061] (11) Stable reference area: refers to an area with relatively small topographic changes, clear boundary features, and suitable as a registration anchor point in multiple observations. In this invention, the stable reference area is used to establish registration constraints between the current model and the historical baseline model, so as to avoid mistakenly eliminating the actual sedimentation changes as geometric deviations.

[0062] (12) Cross-sectional change detection: This refers to the process of performing differential analysis on the current time model and the historical baseline model under unified registration conditions to extract changes in bed elevation, slope, cross-sectional contraction, and local topographical abrupt changes. The cross-sectional change detection in this invention is used to identify siltation areas, scour areas, and abnormal deposition areas, and to provide basic data for subsequent volume estimation and trend prediction.

[0063] (13) Sedimentation thickness: refers to the positive uplift of the current subsurface surface of the target area relative to the historical baseline surface, and is usually used to describe the degree of accumulation at a certain point, a certain section, or a certain area. In this invention, indicators such as average sedimentation thickness, maximum sedimentation thickness, and local sedimentation thickness can be further distinguished.

[0064] (14) Sedimentation volume: refers to the total volume formed by the positive uplift of the current subsurface relative to the historical baseline within the target area. The sedimentation volume in this invention can be obtained by methods such as cross-sectional difference integration, grid elevation difference accumulation, or point cloud curved envelope volume calculation, and is used to quantify the scale of regional sedimentation.

[0065] (15) Short-term trend prediction: This refers to the process of estimating the sedimentation development trend of the seabed within a preset time window based on the cross-sectional change results of multiple consecutive time periods, combined with environmental driving factors such as flow velocity, tide level, temperature, salinity, turbidity, and sediment content. In this invention, short-term trend prediction is mainly used to achieve sedimentation development judgment and risk pre-identification within a range of 24 to 72 hours.

[0066] (16) Graded early warning: This refers to a processing mechanism that classifies the risk level of a target area and outputs an alarm result by comparing indicators such as siltation thickness, siltation volume, rate of change, duration and predicted trend with preset thresholds. The graded early warning in this invention can be expressed in four levels: normal, attention, warning and severe warning, or in multiple levels such as blue, yellow, orange and red.

[0067] (17) Risk threshold: refers to the quantitative or qualitative criteria used to determine whether a target area meets the warning conditions. The risk threshold can be a single indicator threshold or a multi-indicator fusion threshold such as sediment thickness, sediment volume, growth rate, duration, and environmental conditions. In this invention, the risk threshold can be set differently according to different scenarios, different regions, and different management needs.

[0068] (18) Target area: refers to the water area in which underwater topographic monitoring, siltation analysis and risk warning need to be carried out when implementing the method of the present invention. It usually includes port channels, inland waterways, estuary channels, reservoir areas, anchorages, berth areas or nearshore engineering areas, etc. The present invention does not limit the specific scope, shape and scale of the target area.

Claims

1. A siltation early warning method based on towed sonar and multi-source hydrological parameter fusion, characterized in that, Includes the following steps: S1. Conduct survey line planning for the target area and simultaneously collect multibeam sonar depth sounding data, CTD multi-parameter data, platform attitude data, towing cable status data, positioning and navigation data, and time synchronization information during the towing platform operation. S2. Perform time alignment, outlier removal, missing value compensation, and noise suppression on the collected multi-source data to obtain a standardized input dataset; S3. Construct a sound velocity profile based on the CTD multi-parameter data, correct the sound velocity of the multibeam sonar depth sounding data, and perform geometric compensation by combining the platform attitude data, tow cable status data, and positioning and navigation data. S4. Map the corrected depth sounding results to the preset coordinate system to construct the underwater cross-section model, grid elevation model or bottom point cloud model for the current time. S5. Retrieve the historical baseline model corresponding to the current time, and perform coordinate unification, overlapping area identification and adaptive registration on the current time model and the historical baseline model to obtain a benchmark model that can be compared from the same source. S6. Based on the registered current time model and historical baseline model, perform elevation difference, slope change analysis and regional thickness statistics, extract the subsoil uplift zone, subsidence zone, abnormal accumulation zone or scour zone, and calculate the siltation thickness, change area and siltation volume. S7. Correlate the current sedimentation change results with the observation results of multiple historical time periods to construct a time-series change dataset, and integrate environmental driving factors to make short-term trend predictions. S8. Generate an early warning level based on the comparison results of siltation thickness, siltation volume, rate of change, duration, and predicted trend with the preset risk threshold, and output the early warning information.

2. The siltation early warning method according to claim 1, characterized in that, The CTD multi-parameter data includes at least one or more of temperature, salinity, and depth; the platform attitude data includes one or more of roll, pitch, heading, and heave; and the positioning and navigation data includes one or more of global navigation satellite positioning data, inertial navigation data, velocity information, and track information.

3. The siltation early warning method according to claim 1, characterized in that, The time alignment in step S2 adopts a unified timestamp mechanism or a master-slave clock calibration mechanism; for data streams with different sampling frequencies, a unified data sequence is established by interpolation mapping, nearest neighbor matching or weighted alignment; the outlier removal includes the identification and removal of abnormal echo intensity, isolated depth points, repeated jump points and invalid data segments.

4. The siltation early warning method according to claim 1, characterized in that, In step S3, the sound velocity distribution at different depth layers is inverted based on the CTD multi-parameter data, and the sound propagation correction and refraction compensation are performed on the multibeam sonar depth sounding data according to the layered sound velocity model; at the same time, compensation is performed for sonar installation offset, roll error, pitch error, heading deviation, heave effect and track drift.

5. The siltation early warning method according to claim 1, characterized in that, In step S4, during the construction of the current time-series model, one or more of the following processes are further performed: outlier removal, overlapping survey line fusion, local smoothing, sparse region interpolation, and boundary clipping; the current time-series model is output in the form of a single cross-section, a set of cross-sections, a two-dimensional depth map, or a three-dimensional point cloud.

6. The siltation early warning method according to claim 1, characterized in that, The adaptive registration in step S5 includes: unifying the coordinates of the current time-based model and the historical baseline model; identifying the overlapping area between the two; extracting a stable reference area from the overlapping area; extracting one or more geometric features from the slope inflection point, valley point, peak point, and boundary point in the stable reference area; and performing feature matching and iterative optimization based on the geometric features to eliminate translational deviation, rotational deviation, and elevation deviation, thereby obtaining a benchmark model that can be compared from the same source.

7. The siltation early warning method according to claim 1, characterized in that, In step S6, the calculation of the sedimentation volume adopts one or more of the following methods: integration based on cross-sectional differences, volume accumulation based on grid elevation differences, or integration based on point cloud local surface differences; and outputs one or more of the following: average sedimentation thickness, maximum sedimentation thickness, variable area, local sedimentation volume, and total sedimentation volume of the region.

8. The siltation early warning method according to claim 1, characterized in that, The environmental driving factors in step S7 include one or more of the following: flow velocity, tide level, temperature, salinity, turbidity, and sediment content; the short-term trend prediction adopts an extrapolation model based on historical time series change characteristics, a regression prediction model based on environmental driving factors, a trend determination model based on rule thresholds, or a combination of the above models, with a prediction time window of 24 hours to 72 hours.

9. The siltation early warning method according to claim 1, characterized in that, The warning level in step S8 can be set using a four-level system: normal, attention, warning, and severe warning, or a four-level system: blue, yellow, orange, and red. The risk threshold can be set using a single-indicator threshold or a multi-indicator fusion threshold.

10. The siltation early warning method according to claim 1, characterized in that, When the target area meets the warning conditions, the output warning information includes one or more of the following: warning level, change location, trend information, cross-sectional comparison map, regional difference map, and handling suggestion information. The warning information is then sent to the upper-level business platform, monitoring terminal, or management system.

Images