Amphibious ship target detection system and method

Abstract

The invention relates to the technical field of target detection, in particular to an amphibious ship target detection system and method. The method comprises the following steps: acquiring monitoring data of a multi-modal sensor; performing monitoring signal phase alignment processing on the monitoring data of the multi-modal sensor to generate phase alignment data of the multi-modal sensor; constructing a ship space three-dimensional monitoring field according to the phase alignment data of the multi-modal sensor to obtain a ship three-dimensional space monitoring field; performing ship motion trail analysis on the ship three-dimensional space monitoring field to generate ship motion trail analysis data; performing avoidance path planning on the ship movement track analysis data to generate ship avoidance path data; therefore, by integrating the multi-modal data processing and intelligent path planning technology, the problems that a traditional ship target detection system is low in precision, slow in response and poor in environmental adaptability are solved, and the target recognition capacity and the obstacle avoidance capacity of the ship in multiple environments are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of target detection, and in particular to an amphibious ship target detection system and method. Background Art

[0002] The existing amphibious ship target detection system mainly relies on traditional radar or single sensor for target detection, which is easily disturbed by noise in complex marine environments, resulting in reduced detection accuracy, especially poor performance in bad weather or long-distance detection. In addition, the single-mode data processing method cannot effectively integrate multi-source information, fails to fully utilize the complementary advantages of sensors, and affects the reliability of detection results. In terms of target recognition, traditional feature extraction and pattern matching methods are difficult to cope with dynamically changing ship behaviors and complex backgrounds, especially when identifying new targets and handling ship avoidance behaviors. There is a lack of real-time update and adjustment mechanisms, resulting in inaccurate behavior prediction. The existing system relies on manually set rules and thresholds, lacks adaptability and robustness, cannot cope with environmental changes, and reduces execution efficiency. In data processing and model training, traditional methods rely on manual annotation and static analysis, lack automation and intelligent support, increase the complexity of deployment and maintenance, and limit the ability to process large-scale data. Especially when real-time code generation and automatic tuning are required, traditional methods are cumbersome and inefficient, and cannot meet the needs of modern complex tasks. Summary of the invention

[0003] Based on this, it is necessary to provide an amphibious ship target detection system and method to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for detecting an amphibious ship target is provided, the method comprising the following steps: Step S1: acquiring multimodal sensor monitoring data; performing monitoring signal phase alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; constructing a three-dimensional ship space monitoring field according to the multimodal sensor phase alignment data to obtain a three-dimensional ship space monitoring field; Step S2: analyzing the ship motion trajectory of the ship's three-dimensional space monitoring field to generate ship motion trajectory analysis data; performing avoidance path planning on the ship motion trajectory analysis data to generate ship avoidance path data; Step S3: acquiring land area data; performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; performing a confidence assessment model on the land area data and the ship trajectory cross analysis data to generate an amphibious ship target detection model; Step S4: Perform target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; embed the object detection target data into the radar system to perform the amphibious ship target detection task.

[0005] The beneficial effect of the present invention is that by acquiring multimodal sensor monitoring data and performing phase alignment processing on these data, the consistency of the monitoring signals of each sensor in time sequence is ensured, thereby reducing the data deviation caused by signal delay or synchronization error between different sensors. Then, based on the phase-aligned data, the three-dimensional space monitoring field of the ship is constructed, and a comprehensive and accurate ship monitoring model is formed through the fusion of spatial data, which provides high-quality basic data for subsequent trajectory analysis and path planning. In the second step, by analyzing the motion trajectory of the three-dimensional space monitoring field of the ship, detailed ship motion trajectory data is generated, and the avoidance path planning is implemented in combination with the motion law of the ship and the environmental changes, ensuring that the ship can autonomously avoid obstacles or other ships in different complex environments, thereby improving the safety and maneuverability of the ship. Next, by acquiring the land area data and cross-analyzing it with the ship's motion trajectory data, the ship trajectory cross-analysis data is formed, the ship path planning is further optimized, and a confidence assessment model is constructed based on this to ensure the accuracy and reliability of the data in practical applications, and finally an amphibious ship target detection model is generated. The model combines the target recognition algorithm to learn and identify the detection data, thereby realizing real-time monitoring and identification of targets on the water and land. Finally, by embedding the object detection target data into the radar system, a complete detection system is formed to perform the amphibious ship target detection task. Through this series of technical means, the present invention improves the target detection capability and avoidance efficiency of ships in complex environments. Therefore, by integrating multimodal data processing and intelligent path planning technology, the present invention solves the problems of low accuracy, slow response and poor environmental adaptability of traditional ship target detection systems, and improves the target recognition and obstacle avoidance capabilities of ships in multiple environments.

[0006] Preferably, step S1 comprises the following steps: Step S11: Acquire multimodal sensor monitoring data; Step S12: performing monitoring signal alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; performing monitoring object morphology recognition on the multimodal sensor phase alignment data to obtain monitoring object morphology recognition data; Step S13: constructing a three-dimensional ship space monitoring field according to the monitored object morphology recognition data to obtain a three-dimensional ship space monitoring field.

[0007] The present invention obtains monitoring data from different types of sensors (such as radar, optical sensor, sonar, etc.). Since the data sources of different sensors have different timing, frequency and spatial resolution, signal alignment processing is required to ensure the consistency and comparability of the data. Signal alignment processing includes timing synchronization, frequency adjustment and spatial position registration. Through these processing means, the relative time deviation and spatial error between sensors can be eliminated, thereby generating accurate multimodal sensor phase alignment data. The data set after phase alignment provides a reliable basis for subsequent object morphology recognition, and then the morphology recognition of the monitored object is performed on this basis. This process uses advanced image processing and pattern recognition technology to analyze the morphological features of different objects in the sensor data (such as edges, corners, textures, etc.), identify and classify the target objects in the monitoring area, and obtain the morphological recognition data of the monitored objects. Through morphological recognition, the system can extract feature data with key information from a large amount of data, so as to accurately judge the morphology and position of ships or other important objects. Finally, based on morphological recognition data. This process uses geometric modeling and data fusion technology to integrate the recognition data of various sensors according to spatial coordinates, and through data fitting and optimization, form an accurate three-dimensional monitoring field of the ship and its surrounding environment. This three-dimensional monitoring field can provide key support for subsequent ship trajectory analysis, target detection and path planning, etc. Through this technical solution, the present invention significantly improves the accuracy and real-time performance of ship monitoring, and can achieve efficient and stable target monitoring and identification in complex marine environments.

[0008] Preferably, step S13 comprises the following steps: Step S131: performing geometric edge computing on the monitored object morphology recognition data to generate monitoring sensor edge computing data; Step S132: Analyze the geometric constraint relationship of the monitoring sensor edge computing data based on the synchronous positioning and mapping technology to generate the monitoring sensor geometric constraint data; perform spatiotemporal compensation on the monitoring sensor common view area of ​​the monitoring sensor geometric constraint data to generate the monitoring sensor spatiotemporal calibration data; Step S133: Acquire the location of the coastal feature point; perform spatial-temporal regional mapping between the monitoring sensor spatial calibration data and the coastal feature point location, and use particle swarm optimization to correct the sensor registration error to generate ship space sensor data; Step S134: Acquire a BEV bird's-eye view; use the BEV bird's-eye view to perform multimodal distance fusion on the ship space sensor data, and construct a three-dimensional monitoring field of the ship space to obtain a three-dimensional monitoring field of the ship space.

[0009] The present invention performs geometric edge calculation on the morphological recognition data of the monitored object. This process extracts the edge information of the object by analyzing the geometric features in the recognition data, and then generates the edge computing data of the monitoring sensor, which provides a basis for the subsequent spatial relationship analysis. Based on the simultaneous localization and mapping (SLAM) technology, the edge computing data of the monitoring sensor is subjected to geometric constraint relationship analysis. The SLAM technology can synchronously calculate the spatial position of the sensor with the surrounding environment characteristics, optimize the position relationship of the sensor in space through geometric constraints, and thus generate the geometric constraint data of the monitoring sensor. In order to further improve the spatiotemporal accuracy of the data, the spatiotemporal compensation of the sensor common view area is performed on these data, that is, the spatiotemporal calibration is performed in the observation area of ​​the sensor to ensure the consistency between different sensors in time and space, and generate the spatiotemporal calibration data of the monitoring sensor. This calibration process is to compensate and correct the spatiotemporal errors of the sensor through algorithm optimization, thereby improving the accuracy of multi-sensor data fusion. The system obtains the position data of the coastal feature points, which have a significant geographical identification function in the marine environment. Then, the spatiotemporal area mapping technology is used to align the spatiotemporal calibrated sensor data with the position of the coastal feature points to ensure that the sensor data can be accurately mapped to the actual geographical area. On this basis, the particle swarm optimization (PSO) algorithm is used to correct the sensor registration error. The particle swarm optimization can efficiently search for the optimal solution, correct the registration error caused by differences between sensors or environmental changes, and finally generate ship space sensor data to provide accurate sensor positioning information for the construction of a three-dimensional space monitoring field. The system obtains a bird's-eye view (BEV) image, and optimizes the spatial registration between the sensor data by multimodal distance fusion of the BEV image and the ship space sensor data, thereby constructing a three-dimensional monitoring field for the ship space. The BEV image can provide rich ground perspective information, and the combination with the sensor data can help further enhance the spatial perception capability of the ship and its surrounding environment. Through this series of technical means, the present invention can achieve accurate three-dimensional modeling of ships and their surroundings, and improve the real-time monitoring accuracy and stability of ships in complex waters and shoreline environments.

[0010] Preferably, step S2 comprises the following steps: Step S21: Performing spatial motion analysis of the target object according to the three-dimensional spatial monitoring field of the ship to obtain spatial movement data of the object; Step S22: performing ship motion trajectory analysis on the object space movement data to generate ship motion trajectory analysis data; Step S23: Calculate the motion fluid disturbance of the object space movement data and the ship motion trajectory analysis data to generate ship collision probability disturbance data; perform avoidance path planning based on the ship collision probability disturbance data to generate ship avoidance path data.

[0011] The present invention first performs a spatial motion analysis of the target object based on the data of the three-dimensional spatial monitoring field of the ship. This process mainly dynamically analyzes the motion state of the target object (such as other ships, buoys, marine obstacles, etc.) in the monitoring field, extracts the spatial movement characteristics of the object, including motion parameters such as speed, direction, acceleration, and then generates the spatial movement data of the object. These data provide a basis for subsequent trajectory analysis and avoidance path planning. By further analyzing the spatial movement data of the object, the motion trajectory of the ship is calculated. Using the spatial movement data of the target object, combined with the current motion state of the ship, the trajectory prediction model is used to predict the motion trajectory of the ship and its surrounding objects, and the ship motion trajectory analysis data is generated. The analysis data can show the relative position change between the ship and the target object, and provide data support for the next step of collision avoidance decision. The spatial movement data of the object is combined with the ship motion trajectory analysis data to calculate the disturbance of the moving fluid. The calculation estimates the probability of collision of the ship under different motion states by analyzing the influence of the external fluid environment such as water flow and waves on the ship during navigation, as well as the disturbance caused by the interaction between the ship and the surrounding objects. By using fluid mechanics models and computational fluid dynamics (CFD) technology, the disturbance effect of water flow on the navigation trajectory of the ship can be quantified, thereby generating disturbance data on the probability of ship collision. Based on these data, avoidance path planning is further performed. By establishing an avoidance path planning algorithm and combining the collision probability disturbance data, the movement path of the ship is intelligently optimized to generate ship avoidance path data. This process takes into account the dynamic changes of the ship's speed, direction, navigation environment, and obstacles, and can provide the ship with an optimal avoidance path to minimize the risk of collision. Through this series of technical means, the present invention realizes accurate trajectory analysis and intelligent avoidance decision-making of ships in complex dynamic environments, significantly improving the navigation safety and adaptability of ships.

[0012] Preferably, step S22 includes the following steps: Step S221: Perform nonlinear Kalman filtering on the spatial movement data of the object to construct a motion state matrix, and generate a ship motion state prediction matrix; Step S222: performing frequency domain energy spectrum analysis on the ship motion state prediction matrix, extracting the main frequencies of roll and surge, and generating a frequency domain energy distribution spectrum of the ship; Step S223: Perform ship motion stability boundary analysis on the ship motion state prediction matrix through the Lyapunov index stability criterion to generate an acceleration increment safety threshold range; perform dynamic truncation correction processing on the ship frequency domain energy distribution spectrum according to the acceleration increment safety threshold range to generate ship motion trajectory analysis data.

[0013] The present invention constructs a motion state matrix by performing nonlinear Kalman filtering on the spatial movement data of the object. Kalman filtering is a dynamic signal processing method that iteratively updates the estimated value, which can effectively eliminate noise interference and accurately estimate the motion state of the ship. Nonlinear Kalman filtering further expands this technology and can handle complex, nonlinear dynamic systems. In this process, the Kalman filter adjusts the predicted value in real time according to the sensor data, and gradually generates a ship motion state prediction matrix, which contains the motion characteristics of the ship at each moment, such as speed, position, acceleration, etc., and is used for subsequent stability analysis and trajectory prediction. The ship motion state prediction matrix is ​​subjected to frequency domain energy spectrum analysis to extract the main frequency of roll and surge, and generate a frequency domain energy distribution spectrum of the ship. Frequency domain energy spectrum analysis is a technology that converts the signal into the frequency domain for processing, which can reveal the periodicity and fluctuation characteristics in the ship's motion. Roll and surge are the main motion modes of the ship in the lateral and longitudinal directions, and they play an important role in the navigation process of the ship. By extracting these main frequencies, the dynamic response characteristics of the ship under different navigation conditions can be identified, and the corresponding frequency domain energy distribution spectrum can be generated, providing a basis for further stability analysis and path correction. The Lyapunov index stability criterion is applied to analyze the ship motion state prediction matrix to generate the acceleration increment safety threshold range. The Lyapunov index is an indicator to measure the stability of a dynamic system, and can determine whether the system is stable by analyzing the changes in the behavior of the system over time. On the basis of the ship motion state prediction matrix, the Lyapunov index criterion is used to quantitatively analyze the motion stability of the ship, thereby determining the safety boundary of the ship under different motion states and generating the acceleration increment safety threshold range. This process provides a quantitative standard for the safety control of the ship. Finally, according to this safety threshold range, the ship frequency domain energy distribution spectrum is dynamically truncated and corrected to generate the ship motion trajectory analysis data. This correction process can filter out the energy components that exceed the safety range to ensure that the ship's motion trajectory meets the safety and stability requirements. Through this series of technical means, the present invention can achieve accurate prediction, dynamic correction and stability analysis of the ship's motion state, thereby effectively improving the safety and stability of the ship in a complex navigation environment.

[0014] Preferably, performing ship motion stability boundary analysis on the ship motion state prediction matrix by using the Lyapunov index stability criterion comprises the following steps: Extract the maximum Lyapunov exponent of the ship motion state prediction matrix to generate the sensitivity representation of the ship's initial conditions; The gradient field is calculated by characterizing the sensitivity of the ship's initial conditions to generate the flow field velocity gradient tensor; According to the velocity gradient tensor of the flow field, the implicit function relationship of the critical instability acceleration of the ship is constructed to generate the implicit function of the critical instability of the ship; The motion stability boundary analysis is carried out based on the critical instability implicit function of the ship, and the acceleration increment safety threshold range is generated.

[0015] The present invention uses the Lyapunov index to quantitatively measure the sensitivity of the ship motion system when extracting the maximum Lyapunov index of the ship motion state prediction matrix. The Lyapunov index is a method for describing the stability of a dynamic system, which can measure the sensitivity of the ship motion system to changes in initial conditions. By extracting the maximum Lyapunov index, the stability of the ship motion under initial conditions can be quantified, which provides a basis for subsequent stability analysis and path correction, and generates data representing the sensitivity of the ship's initial conditions. Then, based on the characterization of the sensitivity of the ship's initial conditions, the gradient field is calculated to generate a flow field velocity gradient tensor. In this process, the calculation method of the gradient field is used to evaluate the velocity change characteristics of the fluid environment around the ship, especially the influence of the velocity gradient of the fluid on the ship's motion state during the ship's motion. The flow field velocity gradient tensor can describe the change of the flow field in space, reflect the interaction between the ship and the surrounding fluid, and then provide data support for the construction of the implicit function relationship of the critical instability acceleration. Through the flow field velocity gradient tensor, the implicit function relationship of the critical instability acceleration of the ship is constructed. During the movement of a ship, it is affected by fluid disturbances. When the acceleration exceeds a certain threshold, the ship will enter a critical instability state. This process can be described by establishing an implicit function. The implicit function can reflect the influence of factors such as the fluid environment, ship structure, and motion state on the instability acceleration through a series of complex nonlinear relationships, and provide a mathematical basis for the safety boundary of the ship's motion. Based on the critical instability implicit function of the ship, the motion stability boundary analysis is performed to generate the acceleration increment safety threshold range. Through the analysis of the implicit function, the critical stability boundary of the ship under different motion states is determined, that is, the maximum acceleration range in which the ship can safely navigate under a specific fluid environment and motion state. This acceleration increment safety threshold range provides a clear standard for the dynamic control of the ship, which can effectively avoid instability or accidents caused by excessive acceleration during the navigation process.

[0016] Preferably, step S3 comprises the following steps: Step S31: Acquire land area data; Step S32: performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; Step S33: performing early data level fusion on the land area data to generate amphibious ship model input data; performing model training based on the amphibious ship model input data, and calibrating the model confidence parameters to generate a confidence amphibious detection model; Step S34: Perform late feature fusion based on the ship trajectory cross analysis data to generate amphibious ship feature fusion data; perform trajectory matching weighted optimization on the amphibious ship feature fusion data for the confidence amphibious detection model to generate an amphibious ship target detection model.

[0017] The present invention obtains land area data, which includes terrain information, geographical location, coastline characteristics, etc., and provides necessary environmental background information for ship trajectory analysis and collision avoidance path planning. Trajectory correlation cross analysis is performed based on ship motion trajectory analysis data and ship avoidance path data. This analysis generates ship trajectory cross analysis data by comprehensively considering the spatial and temporal relationship between the ship and other objects (such as land obstacles, other ships, etc.). This data can reveal the potential risk of ships crossing land areas or other objects during navigation, and provide an important basis for ship target detection. Early data level fusion is performed based on land area data to generate amphibious ship model input data. This process uses multi-source data fusion technology to comprehensively process the geographical data of the land area with information such as ship motion trajectory and environmental changes, and extracts feature data with high correlation and high credibility as input to ensure data quality and reliability during model training. Based on these data, model training is performed and confidence parameter calibration is performed to generate a confidence amphibious detection model. The model can realize the recognition and detection of amphibious ship targets by learning a large amount of historical data and environmental features, and further improve the accuracy and robustness of the detection results through the calibrated confidence parameters. Late feature fusion is performed based on the cross-analysis data of ship trajectories to generate amphibious ship feature fusion data. Late feature fusion further processes and optimizes the analysis data obtained in the early stage, integrates multi-dimensional spatiotemporal information, motion characteristics, environmental factors, etc., to form a comprehensive feature data set. These data sets can more accurately reflect the movement patterns and potential dangers of ships in complex environments. Then, these amphibious ship feature fusion data are used to perform trajectory matching weighted optimization on the confidence amphibious detection model. The optimization process enhances the model's response to key features through a weighted method, so that the model can give priority to important risk factors when facing a complex and changing environment, thereby generating the final amphibious ship target detection model. The model can not only effectively identify targets in a static environment, but also intelligently predict and adjust the ship's motion path in a dynamic environment, thereby improving the accuracy and response speed of target detection.

[0018] Preferably, step S32 includes the following steps: Step S321: performing ship avoidance marking on the ship avoidance path data to generate ship avoidance mark data; performing deceleration-avoidance time feature two-dimensional feature extraction on the ship avoidance mark data to generate ship avoidance mark feature data; Step S322: performing time window matching on the ship avoidance mark feature data and the ship motion trajectory analysis data to generate ship time window avoidance-path matching data; performing Euclidean distance trajectory similarity metric matching on the ship time window avoidance-path matching data to generate ship motion trajectory-avoidance path matching data; Step S323: performing ship behavior pattern recognition on the ship motion trajectory-avoidance path matching data to generate ship motion behavior pattern feature data; performing avoidance type trajectory association cross analysis based on the ship motion behavior pattern feature data to generate ship trajectory cross analysis data.

[0019] The present invention generates ship avoidance mark data by marking the ship avoidance path data. This step is intended to identify the time and position where the ship needs to avoid during navigation, and by marking the path data, key nodes are provided for subsequent analysis. Then, the ship avoidance mark data is subjected to two-dimensional feature extraction of deceleration-avoidance time features to generate ship avoidance mark feature data. The feature extraction process focuses on capturing the time information required for deceleration and avoidance of the ship during the avoidance process, which is of great significance for predicting the time and spatial distribution of the ship avoidance behavior. The extraction of deceleration and avoidance time features provides a reliable data basis for further trajectory matching and path optimization by quantifying the process of ship deceleration and the duration of avoidance action. The ship avoidance mark feature data is matched with the ship motion trajectory analysis data by time window to generate ship time window avoidance-path matching data. This step synchronizes the ship avoidance action with the motion trajectory data by setting the time window, so as to more accurately dock the ship's motion and avoidance behavior. Through the matching of the time window, the ship's avoidance trajectory can be more clearly defined in time and space, ensuring the consistency between the avoidance path and the motion trajectory. Subsequently, the ship time window avoidance-path matching data is matched by Euclidean distance trajectory similarity measurement to generate ship motion trajectory-avoidance path matching data. The Euclidean distance measurement method is used to evaluate the trajectory similarity of the ship in different time periods, so as to determine whether the ship motion trajectory matches the avoidance path, and optimize the path selection through the similarity measurement to ensure that the ship can complete the avoidance safely and timely. The ship motion trajectory-avoidance path matching data is subjected to ship behavior pattern recognition to generate ship motion behavior pattern feature data. This step analyzes the motion behavior of the ship in different situations through the pattern recognition algorithm, identifies the typical behavior pattern of the ship in different avoidance scenarios, and generates corresponding feature data to provide a basis for subsequent avoidance decisions. By identifying the ship motion behavior pattern, the motion law of the ship in a specific avoidance situation can be revealed, providing guidance for the dynamic adjustment and optimization of the avoidance path. Then, based on the ship motion behavior pattern feature data, the avoidance type trajectory association cross analysis is performed to generate ship trajectory cross analysis data. By cross-analyzing different types of avoidance behaviors, we can further explore the intrinsic connection between ship motion and avoidance paths, generate ship trajectory cross-analysis data, and thus optimize the selection and adjustment of avoidance paths, thereby improving the intelligence and real-time performance of ship avoidance decisions.

[0020] Preferably, step S4 comprises the following steps: Step S41: performing target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; Step S42: Automatically generate code according to the object detection target data to generate an amphibious ship target detection module; Step S43: embed the amphibious ship target detection module into the radar system to perform the amphibious ship target detection mission.

[0021] The present invention obtains object detection target data by performing target detection and recognition learning based on an amphibious ship target detection model. This step uses deep learning technology to train a large amount of sensor data (including radar, infrared, vision, etc.), and automatically identifies and classifies different target objects through feature extraction and pattern recognition. Through an efficient target recognition algorithm, the ship's targets in a complex environment can be accurately classified and located, thereby generating object detection target data, which is essential for further target verification and task execution. Automatic code generation is performed based on the object detection target data to generate an amphibious ship target detection module. Through the automatic code generation technology, the target detection results are converted into a code module that can actually run. The automatic code generation technology automatically creates corresponding functional modules by parsing the target detection data, and these modules can be connected to a larger system for data interaction and task execution. This process not only improves development efficiency, but also ensures the high consistency and accuracy of the target detection module, avoids potential errors or inconsistencies in manually written codes, and enhances the stability and maintainability of the system. The generated amphibious ship target detection module is embedded in the radar system to perform amphibious ship target detection tasks. By embedding the generated target detection module into the existing radar system, the system can perform target detection tasks in real time. The radar system receives and processes data from different sensors, and uses the automatic recognition and processing capabilities of the detection module to classify, locate and track potential targets. This process greatly improves the accuracy and efficiency of target detection, and can respond to various target changes in real time in a dynamic environment to ensure the safe navigation of the ship.

[0022] In this specification, an amphibious ship target detection system is provided, which is used to perform the above-mentioned amphibious ship target detection method. The amphibious ship target detection system includes: The data acquisition and space construction module is used to obtain multi-modal sensor monitoring data; perform monitoring signal phase alignment processing on the multi-modal sensor monitoring data to generate multi-modal sensor phase alignment data; construct the ship's three-dimensional space monitoring field according to the multi-modal sensor phase alignment data to obtain the ship's three-dimensional space monitoring field; The motion trajectory analysis and path planning module is used to analyze the ship's motion trajectory in the three-dimensional space monitoring field of the ship and generate ship motion trajectory analysis data; perform avoidance path planning on the ship's motion trajectory analysis data and generate ship avoidance path data; The trajectory association and target detection model building module is used to obtain land area data; perform trajectory association cross analysis based on ship motion trajectory analysis data and ship avoidance path data to generate ship trajectory cross analysis data; perform a confidence assessment model on land area data and ship trajectory cross analysis data to generate an amphibious ship target detection model; The target detection and execution module is used to perform target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; the object detection target data is embedded into the radar system to execute the amphibious ship target detection task.

[0023] The beneficial effect of the present invention is that by acquiring multimodal sensor monitoring data and performing phase alignment processing on these data, the consistency of the monitoring signals of each sensor in time sequence is ensured, thereby reducing the data deviation caused by signal delay or synchronization error between different sensors. Then, based on the phase-aligned data, the three-dimensional space monitoring field of the ship is constructed, and a comprehensive and accurate ship monitoring model is formed through the fusion of spatial data, which provides high-quality basic data for subsequent trajectory analysis and path planning. In the second step, by analyzing the motion trajectory of the three-dimensional space monitoring field of the ship, detailed ship motion trajectory data is generated, and the avoidance path planning is implemented in combination with the motion law of the ship and the environmental changes, ensuring that the ship can autonomously avoid obstacles or other ships in different complex environments, thereby improving the safety and maneuverability of the ship. Next, by acquiring the land area data and cross-analyzing it with the ship's motion trajectory data, the ship trajectory cross-analysis data is formed, the ship path planning is further optimized, and a confidence assessment model is constructed based on this to ensure the accuracy and reliability of the data in practical applications, and finally an amphibious ship target detection model is generated. The model combines the target recognition algorithm to learn and identify the detection data, thereby realizing real-time monitoring and identification of targets on the water and land. Finally, by embedding the object detection target data into the radar system, a complete detection system is formed to perform the amphibious ship target detection task. Through this series of technical means, the present invention improves the target detection capability and avoidance efficiency of ships in complex environments. Therefore, by integrating multimodal data processing and intelligent path planning technology, the present invention solves the problems of low accuracy, slow response and poor environmental adaptability of traditional ship target detection systems, and improves the target recognition and obstacle avoidance capabilities of ships in multiple environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the steps of an amphibious ship target detection method; Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart; Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. Figure 4 for Figure 1 Detailed implementation steps of step S4 in FIG. The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0025] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0026] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0027] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve this, please refer to Figures 1 to 4 , an amphibious ship target detection method, the method comprising the following steps: Step S1: acquiring multimodal sensor monitoring data; performing monitoring signal phase alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; constructing a three-dimensional ship space monitoring field according to the multimodal sensor phase alignment data to obtain a three-dimensional ship space monitoring field; Step S2: analyzing the ship motion trajectory of the ship's three-dimensional space monitoring field to generate ship motion trajectory analysis data; performing avoidance path planning on the ship motion trajectory analysis data to generate ship avoidance path data; Step S3: acquiring land area data; performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; performing a confidence assessment model on the land area data and the ship trajectory cross analysis data to generate an amphibious ship target detection model; Step S4: Perform target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; embed the object detection target data into the radar system to perform the amphibious ship target detection task.

[0029] The beneficial effect of the present invention is that by acquiring multimodal sensor monitoring data and performing phase alignment processing on these data, the consistency of the monitoring signals of each sensor in time sequence is ensured, thereby reducing the data deviation caused by signal delay or synchronization error between different sensors. Then, based on the phase-aligned data, a three-dimensional spatial monitoring field of the ship is constructed, and a comprehensive and accurate ship monitoring model is formed through the fusion of spatial data, which provides high-quality basic data for subsequent trajectory analysis and path planning. By analyzing the motion trajectory of the three-dimensional spatial monitoring field of the ship, detailed ship motion trajectory data is generated, and the avoidance path planning is implemented in combination with the motion law of the ship and the environmental changes, ensuring that the ship can autonomously avoid obstacles or other ships in different complex environments, thereby improving the safety and maneuverability of the ship. Next, by acquiring the land area data and cross-analyzing it with the ship's motion trajectory data, the ship trajectory cross-analysis data is formed, the ship path planning is further optimized, and a confidence assessment model is constructed based on this to ensure the accuracy and reliability of the data in practical applications, and finally an amphibious ship target detection model is generated. The model combines the target recognition algorithm to learn and identify the detection data, thereby realizing real-time monitoring and identification of targets on the water and land. Finally, by embedding the object detection target data into the radar system, a complete detection system is formed to perform the amphibious ship target detection task. Through this series of technical means, the present invention improves the target detection capability and avoidance efficiency of ships in complex environments. Therefore, by integrating multimodal data processing and intelligent path planning technology, the present invention solves the problems of low accuracy, slow response and poor environmental adaptability of traditional ship target detection systems, and improves the target recognition and obstacle avoidance capabilities of ships in multiple environments.

[0030] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a step flow chart of an amphibious ship target detection method of the present invention. In this example, the amphibious ship target detection method includes the following steps: Step S1: acquiring multimodal sensor monitoring data; performing monitoring signal phase alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; constructing a three-dimensional ship space monitoring field according to the multimodal sensor phase alignment data to obtain a three-dimensional ship space monitoring field; In an embodiment of the present invention, multimodal sensor monitoring data is obtained. These data come from different types of sensors, such as radar, infrared, sonar, visual sensors, etc. These sensors can provide multi-dimensional information about the ship and its surrounding environment, including position, speed, shape and other related features. By comprehensively utilizing these multimodal data, more comprehensive and accurate ship information can be obtained, laying the foundation for subsequent analysis and modeling. After the data is acquired, the monitoring signal phase alignment processing is performed on the data from different sensors to generate multimodal sensor phase alignment data. Phase alignment processing refers to aligning the data collected by different sensors in time and space, so that the data from different sensors can be effectively compared and analyzed in the same coordinate system and time frame. This processing can eliminate the time delay and spatial error between sensors, so that the timing information of multiple data sources is synchronized, thereby ensuring the accuracy and consistency of subsequent analysis results. Next, based on the multimodal sensor phase alignment data, the three-dimensional monitoring field of the ship space is constructed. In this process, by combining the spatial distribution and time series changes of sensor data, a three-dimensional modeling algorithm (such as multi-view geometric modeling, spatial interpolation, etc.) is used to convert the multimodal sensor data into a three-dimensional spatial model of the environment where the ship is located. This three-dimensional spatial monitoring field can accurately reflect the ship's relative position, motion trajectory and its relationship with the surrounding environment, providing detailed spatial structure information for subsequent monitoring and prediction tasks.

[0031] Step S2: analyzing the ship motion trajectory of the ship's three-dimensional space monitoring field to generate ship motion trajectory analysis data; performing avoidance path planning on the ship motion trajectory analysis data to generate ship avoidance path data; In an embodiment of the present invention, the motion trajectory of the ship can be extracted and modeled by trajectory analysis technology. The data level technology used includes a trajectory prediction algorithm based on time series, such as Kalman filtering, particle filtering or Bayesian filtering, etc. These methods can effectively handle noise and uncertainty, thereby accurately predicting the future motion trajectory of the ship. By processing the spatial monitoring data, the motion law of the ship within a given time range can be extracted, thereby providing the required dynamic information for the subsequent avoidance path planning. After generating the ship motion trajectory analysis data, the next step is to perform avoidance path planning based on these data. The core task of avoidance path planning is to calculate an optimal avoidance path based on the motion state of the ship and the surrounding environment. This process relies on a path planning algorithm, such as an algorithm based on the Dijkstra algorithm or an algorithm based on optimization theory, by extracting the spatial relationship between the ship and the obstacle in the ship motion trajectory analysis data, and generating an avoidance path according to the avoidance requirements and the dynamic characteristics of the ship. In order to further optimize the path planning results, different constraints are taken into account in this process, such as the minimum turning radius of the ship, speed limit, avoidance time window, etc., so as to ensure the executability and safety of the path.

[0032] Step S3: acquiring land area data; performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; performing a confidence assessment model on the land area data and the ship trajectory cross analysis data to generate an amphibious ship target detection model; In an embodiment of the present invention, by acquiring the land area data, a rich environmental background can be provided for the subsequent ship trajectory analysis and avoidance path planning. Next, based on the ship motion trajectory analysis data and the ship avoidance path data, trajectory correlation cross analysis is performed. This analysis mainly relies on multivariate data fusion and association technology. By comparing the ship's motion trajectory data with the avoidance path data, combined with the land area data, the spatial interaction between the ship and the surrounding environment is analyzed. In order to realize the cross analysis of trajectory and avoidance path, spatiotemporal association algorithm and data mining technology are used to fuse the dynamic motion data of the ship with the static land data, thereby revealing potential collision areas, avoidance opportunities and cross paths. Through this step, ship trajectory cross analysis data can be generated to provide data support for the accuracy of subsequent target detection tasks. A confidence assessment model is constructed for the ship trajectory cross analysis data and the land area data. The confidence assessment model uses machine learning methods, such as support vector machine (SVM), decision tree or random forest, to train and evaluate data. In this process, the model will generate credibility indicators for ship target detection based on the input data (such as trajectory intersection information, environmental characteristics, etc.). These indicators can help the system evaluate the reliability and accuracy of the detection results, thereby providing effective data support for ship target detection. Finally, based on the above analysis results, an amphibious ship target detection model is generated. This model will combine trajectory analysis data, avoidance path data, and confidence assessment results, and further improve the accuracy of ship target detection through data fusion and pattern recognition technology. Through this model, the system can effectively identify potential targets in complex environments and complete the task of accurately detecting amphibious ship targets.

[0033] Step S4: Perform target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; embed the object detection target data into the radar system to perform the amphibious ship target detection task.

[0034] In an embodiment of the present invention, target detection and recognition learning is performed on an amphibious ship target detection model. This process mainly involves training the target detection model through machine learning algorithms, such as convolutional neural networks (CNNs), deep learning algorithms, decision trees, etc. During the training process, the model gradually adjusts internal parameters through an optimization algorithm based on pre-collected object detection target data (such as ship image data, radar echo signals or sensor data, etc.), thereby improving the accuracy and efficiency of target recognition. Through this process, the model can identify different target features from multimodal data, and can handle complex background noise, obstructions and other factors to improve the robustness and accuracy of detection. The key technology of target detection and recognition learning lies in the use of a large amount of labeled data and the adaptive learning mechanism introduced during the training process to ensure that the model can adjust its recognition strategy according to actual conditions and gradually improve the accuracy of target recognition. After completing the learning, the model can generate object detection target data, which includes not only the position and morphological characteristics of the target, but also the dynamic change information of the target, such as speed, direction, etc. Next, these object detection target data will be embedded in the radar system for data fusion and processing. The core technology of the embedded radar system is the design of the data interface and communication protocol, which ensures that the target detection data can be seamlessly connected with other components of the radar system and provide accurate and real-time target information. By receiving these embedded data, the radar system can update the execution status of the ship target detection task in real time according to the spatial position, speed and movement trend of the object. In this process, the radar system will also optimize and adjust based on the detected target data, and further improve the execution effect of the amphibious ship target detection task by continuously updating the object detection data. The real-time detection performance of the radar system depends on the accurate target data processing capability and efficient data fusion technology, so as to ensure the real-time detection and dynamic tracking of ship targets in complex environments.

[0035] Preferably, step S1 comprises the following steps: Step S11: Acquire multimodal sensor monitoring data; Step S12: performing monitoring signal alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; performing monitoring object morphology recognition on the multimodal sensor phase alignment data to obtain monitoring object morphology recognition data; Step S13: constructing a three-dimensional ship space monitoring field according to the monitored object morphology recognition data to obtain a three-dimensional ship space monitoring field.

[0036] In an embodiment of the present invention, monitoring data from a multimodal sensor is obtained. Multimodal sensors include laser radar, infrared imaging, radar, visual sensors, etc., which can collect rich data about target objects from different perception dimensions (such as space, time, spectrum, etc.). These data include multi-dimensional information such as the spatial position, shape, material, temperature, speed, etc. of the object. Multimodal data fusion technology improves the reliability and diversity of information by integrating raw data from different sensors. In practical applications, data synchronization and denoising are very important steps. First, the monitoring signal phase alignment processing is performed on these multimodal sensor monitoring data, the purpose of which is to solve the data inconsistency problem caused by the synchronization error of different sensors in time and space. Phase alignment processing can ensure that the monitoring signals from different sensors are aligned within the same time window, thereby ensuring the consistency and accuracy of the data in subsequent processing. This process uses interpolation algorithms, timestamp matching, or time-space alignment based on synchronous localization and mapping technology (SLAM). After the phase alignment is completed, the generated multimodal sensor phase alignment data provides a highly consistent and accurate data basis for subsequent object recognition and monitoring field construction. Based on the aligned data, the morphological recognition of the monitored object is continued. Through image recognition algorithms and target detection methods (such as convolutional neural networks (CNNs) and deep learning), the data obtained by the sensors can be analyzed to identify the morphological features of objects. These morphological recognition data not only contain the geometric shape of the object, but also other physical features, such as material type, surface reflectivity, etc., which are crucial for subsequent monitoring and identification. Based on the morphological recognition data of the monitored object, the three-dimensional spatial monitoring field of the ship is further constructed. The construction of the three-dimensional spatial monitoring field depends on multi-sensor fusion technology and geometric modeling methods. Using the morphological recognition data and combining it with the spatial positioning data of the sensor, a three-dimensional spatial monitoring environment around the ship is constructed through three-dimensional reconstruction technology (such as stereo vision, point cloud data fusion, spatial transformation algorithm, etc.).

[0037] Preferably, step S13 includes the following steps: Step S131: performing geometric edge computing on the monitored object morphology recognition data to generate monitoring sensor edge computing data; Step S132: Analyze the geometric constraint relationship of the monitoring sensor edge computing data based on the synchronous positioning and mapping technology to generate the monitoring sensor geometric constraint data; perform spatiotemporal compensation on the monitoring sensor common view area of ​​the monitoring sensor geometric constraint data to generate the monitoring sensor spatiotemporal calibration data; Step S133: Acquire the location of the coastal feature point; perform spatial-temporal regional mapping between the monitoring sensor spatial calibration data and the coastal feature point location, and use particle swarm optimization to correct the sensor registration error to generate ship space sensor data; Step S134: Acquire a BEV bird's-eye view; use the BEV bird's-eye view to perform multimodal distance fusion on the ship space sensor data, and construct a three-dimensional monitoring field of the ship space to obtain a three-dimensional monitoring field of the ship space.

[0038] In an embodiment of the present invention, the edge computing data of the monitoring sensor is generated by extracting the geometric edge features of the target object. This process relies on image processing algorithms, such as edge detection algorithms (such as Canny edge detection or Sobel operator), and accurately extracts the geometric features of the object by calculating the spatial boundary information in the sensor data. The edge computing data provides important geometric information in object recognition and subsequent analysis, which helps to build an accurate monitoring environment model. Based on the simultaneous localization and mapping technology (SLAM), the geometric constraint relationship of the monitoring sensor edge computing data is analyzed. SLAM technology combines sensor data and spatial positioning information to analyze the geometric constraint relationship between different sensors, and then generates the geometric constraint data of the monitoring sensor. This process identifies the relative position and posture between sensors through optimization algorithms (such as Kalman filtering or maximum likelihood estimation), thereby accurately describing the geometric constraints between sensors. Next, the monitoring sensor geometric constraint data is subjected to spatiotemporal compensation in the sensor common view area. Through spatiotemporal compensation, it is ensured that the data of different sensors can be calibrated consistently in time and space, and accurate spatiotemporal calibration data of the monitoring sensor is generated. Obtaining the location of coastal feature points is the key spatial reference information. Satellite positioning systems (such as GPS) or geographic information system (GIS) technology are used to determine the location of coastline feature points. Then, the spatiotemporal calibration data of the monitoring sensor is mapped to the location of the coastal feature points in a spatiotemporal area. This process ensures the accuracy of spatial data by aligning the sensor data with the geographic coordinate system. In order to further correct the sensor registration error, the particle swarm optimization (PSO) algorithm is used for optimization. Particle swarm optimization can effectively reduce the registration error caused by sensor error by simulating group intelligence to search for the optimal solution, and finally generate ship space sensor data. By obtaining a bird's eye view (BEV), the ship space sensor data is further fused by multimodal distance. BEV images provide a global perspective from high altitude, which can combine sensor data from different sources (such as lidar, vision, infrared, etc.), and enhance the accuracy and reliability of the data through multimodal data fusion methods (such as weighted average, Kalman filtering, etc.). Finally, the fused data is applied to the construction of the ship space three-dimensional monitoring field, and a precise ship three-dimensional space model is generated using a three-dimensional reconstruction algorithm (such as stereo vision, point cloud processing, etc.).

[0039] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: Performing spatial motion analysis of the target object according to the three-dimensional spatial monitoring field of the ship to obtain spatial movement data of the object; Step S22: performing ship motion trajectory analysis on the object space movement data to generate ship motion trajectory analysis data; Step S23: Calculate the motion fluid disturbance of the object space movement data and the ship motion trajectory analysis data to generate ship collision probability disturbance data; perform avoidance path planning based on the ship collision probability disturbance data to generate ship avoidance path data.

[0040] In the embodiment of the present invention, by obtaining the position, velocity and acceleration data of the ship in space, combined with the dynamic characteristics of the object and external environmental factors, the dynamic model or spatiotemporal interpolation technology is used to calculate the spatial movement data of the object. These calculations rely on state estimation methods, such as Kalman filtering or particle filtering, which are used to track the motion trajectory of the target object in real time. Through these technologies, the motion trajectory, velocity change and acceleration characteristics of the object in three-dimensional space can be accurately extracted, thereby obtaining the spatial movement data of the object. The object's spatial movement data is further used for ship motion trajectory analysis. Specifically, the historical trajectory data and real-time motion data of the ship are used to model and analyze the motion trajectory of the object through technical means such as trajectory matching and path fitting. Commonly used methods include least squares method, polynomial fitting, etc., and by analyzing the state of the ship at different time points, the ship's motion trajectory analysis data is generated. These data reflect the movement path, speed change and turning point of the ship along the way. Trajectory analysis helps to provide a scientific basis for subsequent path planning, collision warning and other applications. The ship's motion trajectory analysis data is combined with the object's spatial movement data, and the ship's collision probability disturbance data is obtained by calculating the perturbation of the moving fluid. This calculation involves fluid dynamics models, especially when considering the effects of fluid factors such as water flow, wind speed, and tide on ship motion. Using fluid dynamics simulation, combined with the ship's speed, direction, and disturbance factors of the external environment, the disturbance in the ship's motion is calculated and the collision risk is predicted. In order to accurately calculate the collision probability of ships under disturbance conditions, random simulation methods such as Monte Carlo simulation are required to obtain accurate collision probability data. Finally, avoidance path planning is performed based on the disturbance data of the ship's collision probability. This process involves data analysis and optimization algorithms based on collision probability, such as the Dijkstra algorithm, which combines the ship's current position, speed, and avoidance target to calculate the optimal avoidance path.

[0041] Preferably, step S22 includes the following steps: Step S221: Perform nonlinear Kalman filtering on the spatial movement data of the object to construct a motion state matrix, and generate a ship motion state prediction matrix; Step S222: performing frequency domain energy spectrum analysis on the ship motion state prediction matrix, extracting the main frequencies of roll and surge, and generating a frequency domain energy distribution spectrum of the ship; Step S223: Perform ship motion stability boundary analysis on the ship motion state prediction matrix through the Lyapunov index stability criterion to generate an acceleration increment safety threshold range; perform dynamic truncation correction processing on the ship frequency domain energy distribution spectrum according to the acceleration increment safety threshold range to generate ship motion trajectory analysis data.

[0042] In an embodiment of the present invention, the spatial movement data of the object is processed by nonlinear Kalman filtering to construct a motion state matrix and generate a ship motion state prediction matrix. Kalman filtering is an optimal estimation method for processing dynamic systems with noise. In the nonlinear Kalman filtering process, the motion state of the ship is first estimated using a filtering algorithm based on the dynamic model of the ship (such as the motion equation) and the actual observation data. In particular, when the system has nonlinear characteristics, methods such as extended Kalman filtering (EKF) or unscented Kalman filtering (UKF) are used for state prediction. These methods obtain accurate ship motion state prediction matrices by continuously updating the estimated values ​​of state variables, covering important parameters such as the position, velocity and acceleration of the ship. These prediction matrices provide key data support for subsequent motion analysis, trajectory prediction and stability assessment. The ship motion state prediction matrix will be subjected to frequency domain energy spectrum analysis. Frequency domain energy spectrum analysis is based on Fourier transform, which converts time domain signals into frequency domain representations, thereby revealing the main frequency components in the ship motion process. In this process, by calculating the energy spectrum density of the ship motion, especially extracting the frequency components of roll and surge, the motion characteristics of the ship in different directions can be identified. Through energy spectrum analysis, the energy distribution of the ship at different frequencies can be obtained, and the main frequency components can be extracted. These main frequencies reflect the natural vibration mode of the ship's motion. The generated ship frequency domain energy distribution spectrum provides frequency domain information for subsequent stability analysis and motion state prediction, which helps to reveal the dynamic behavior of the ship under external disturbances. The Lyapunov index stability criterion is used to perform motion stability boundary analysis on the ship motion state prediction matrix. The Lyapunov index is an important tool for judging the stability of the system. By analyzing the time evolution of the ship's motion state, the Lyapunov index can reveal whether the ship is in a stable state, especially in the presence of external disturbances. By calculating the Lyapunov index, the stability boundary of the ship's motion can be determined, that is, the maximum acceleration increment that the ship can withstand under different conditions. This safety threshold helps to provide a stability boundary for the ship's motion, avoiding dangers such as loss of control or collision caused by exceeding the safety threshold. Finally, according to the acceleration increment safety threshold range, the ship's frequency domain energy distribution spectrum is dynamically truncated and corrected to further optimize the analysis data of the ship's motion trajectory.

[0043] Preferably, performing ship motion stability boundary analysis on the ship motion state prediction matrix by using the Lyapunov index stability criterion includes the following: Extract the maximum Lyapunov exponent of the ship motion state prediction matrix to generate the sensitivity representation of the ship's initial conditions; The gradient field is calculated by characterizing the sensitivity of the ship's initial conditions to generate the flow field velocity gradient tensor; According to the velocity gradient tensor of the flow field, the implicit function relationship of the critical instability acceleration of the ship is constructed to generate the implicit function of the critical instability of the ship; The motion stability boundary analysis is carried out based on the critical instability implicit function of the ship, and the acceleration increment safety threshold range is generated.

[0044] In an embodiment of the present invention, the maximum Lyapunov exponent is extracted from the ship motion state prediction matrix, aiming to characterize the sensitivity of the initial conditions of the ship through the Lyapunov exponent. As a classic stability analysis tool, the Lyapunov exponent can evaluate the sensitivity of the ship system to the initial conditions, that is, the stability of the system. If the Lyapunov exponent is positive, the system is unstable, and vice versa, if it is negative, the system is stable. By calculating the maximum Lyapunov exponent, the stability analysis results of the ship under the initial conditions can be obtained, which provides basic data support for the subsequent prediction and optimization of the ship motion state. The gradient field calculation is performed using the sensitivity characterization of the initial conditions of the ship to generate the flow field velocity gradient tensor. The purpose of the gradient field calculation is to capture the changes in the flow field of the ship during the movement, especially the velocity distribution and change law at different spatial positions. By calculating the gradient of the velocity field, the velocity gradient tensor of the flow field can be obtained, which reflects the velocity change trend and the velocity change in different directions during the ship's motion. These data are crucial for describing the interaction between the ship and the surrounding fluid and analyzing the stability of the ship's motion. Based on the velocity gradient tensor of the flow field calculated above, the implicit function relationship of the critical instability acceleration of the ship is further constructed. Critical instability acceleration refers to the situation where the ship becomes unstable when it exceeds a certain acceleration threshold under external disturbance. By analyzing the velocity gradient of the flow field and the motion response of the ship, an implicit function relationship between acceleration and ship instability can be established. This implicit function describes the dynamic connection between ship acceleration and instability under specific conditions through mathematical modeling and numerical simulation, and can predict the critical point of instability according to the actual motion state. Finally, based on the implicit function of the critical instability of the ship, the motion stability boundary analysis is performed to generate the acceleration increment safety threshold range. The motion stability boundary analysis is to evaluate the acceleration of the ship under different motion states through the established implicit function relationship, and then determine the safe operation range of the ship. The acceleration increment safety threshold range provides a stable operation boundary for the ship when encountering external disturbances, ensuring that the movement of the ship does not exceed the safety range, thereby avoiding safety hazards such as instability or collision.

[0045] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: Acquire land area data; Step S32: performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; Step S33: performing early data level fusion on the land area data to generate amphibious ship model input data; performing model training based on the amphibious ship model input data, and calibrating the model confidence parameters to generate a confidence amphibious detection model; Step S34: Perform late feature fusion based on the ship trajectory cross analysis data to generate amphibious ship feature fusion data; perform trajectory matching weighted optimization on the amphibious ship feature fusion data for the confidence amphibious detection model to generate an amphibious ship target detection model.

[0046] In an embodiment of the present invention, land area data is obtained and processed to extract environmental features related to ship target detection. These data include geographic information, meteorological data, building layout, ground moving objects, etc., which play a key role in subsequent detection and trajectory analysis. Subsequently, trajectory correlation cross analysis is performed based on the ship motion trajectory analysis data and the ship avoidance path data. This analysis identifies the correlation between the ship's motion behavior and the avoidance path under specific environments and conditions through the matching of time and space, and generates ship trajectory cross analysis data. This data provides a basis for subsequent behavior prediction and path optimization, and provides a scientific basis for understanding the ship's motion mode. The land area data is fused at an early data level, with the intention of combining the environmental data of the land area with the ship's motion trajectory data to generate input data for the amphibious ship model. This fusion process involves the integration of multi-source data. Through technical means such as data cleaning, normalization, and time alignment, data from different sources are unified into a processable standard format for subsequent modeling and analysis. Based on these fused input data, the model is trained using a machine learning method, and the reliability of the model is further improved by calibration of confidence parameters. During the training process, the model learns a large amount of historical data to capture the movement patterns of ships in various environments and generate a confidence amphibious detection model. Late feature fusion is performed based on the cross-analysis data of ship trajectories. This process uses various features extracted in the early analysis, such as the movement trajectory of the ship, the avoidance path, and the changes in the surrounding environment, to extract and fuse features, thereby generating amphibious ship feature fusion data. These features include multi-dimensional fusion of time series, spatial coordinates, speed, acceleration and other data to form a comprehensive description of the ship's behavior. Finally, the generated feature fusion data is used to perform trajectory matching weighted optimization on the confidence amphibious detection model to further improve the accuracy of the model and optimize the target detection capability. In this process, a weighted optimization algorithm is used to adjust the weighted importance of different features, so that the model can more accurately identify amphibious ship targets in practical applications.

[0047] Preferably, step S32 includes the following steps: Step S321: performing ship avoidance marking on the ship avoidance path data to generate ship avoidance mark data; performing deceleration-avoidance time feature two-dimensional feature extraction on the ship avoidance mark data to generate ship avoidance mark feature data; Step S322: performing time window matching on the ship avoidance mark feature data and the ship motion trajectory analysis data to generate ship time window avoidance-path matching data; performing Euclidean distance trajectory similarity metric matching on the ship time window avoidance-path matching data to generate ship motion trajectory-avoidance path matching data; Step S323: performing ship behavior pattern recognition on the ship motion trajectory-avoidance path matching data to generate ship motion behavior pattern feature data; performing avoidance type trajectory association cross analysis based on the ship motion behavior pattern feature data to generate ship trajectory cross analysis data.

[0048] In the embodiment of the present invention, by analyzing the path change and the motion characteristics such as speed and acceleration of the ship, the key nodes of the ship avoidance and the occurrence time of the avoidance behavior are calibrated to generate the ship avoidance mark data. Further, the deceleration-avoidance time feature extraction is performed on these mark data, and the two-dimensional features are extracted by combining the deceleration and time dimension changes of the ship during the avoidance process. This process generates ship avoidance mark feature data through the method of time series analysis and statistical feature extraction, which fully reflects the key behavior mode of ship avoidance. The ship avoidance mark feature data and the ship motion trajectory analysis data are matched in time window. The division of the time window ensures comparison and matching within the same time range by segmenting the time series of the ship motion trajectory data, thereby generating the ship time window avoidance-path matching data. In order to further improve the matching accuracy, the Euclidean distance trajectory similarity measurement method is used to match the ship trajectory within the time window. The method measures the trajectory similarity by calculating the Euclidean distance between the trajectories, thereby providing a quantitative basis for the matching between the ship motion trajectory and the avoidance path, and generating the ship motion trajectory-avoidance path matching data. Finally, based on the ship motion trajectory-avoidance path matching data, ship behavior pattern recognition is performed. This process uses machine learning and pattern recognition technology to analyze the matching data and identify the behavior patterns exhibited by the ship during the avoidance process. This analysis not only focuses on the spatiotemporal laws of the avoidance action, but also involves the interaction mode of the ship with other ships and obstacles in a complex marine environment, generating ship motion behavior pattern feature data. Subsequently, through the avoidance type trajectory correlation cross analysis, the correlation relationship between different avoidance types and trajectories is further explored, thereby generating ship trajectory cross analysis data.

[0049] As an example of the present invention, refer to Figure 4 As shown, in this example, step S4 includes: Step S41: performing target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; Step S42: Automatically generate code according to the object detection target data to generate an amphibious ship target detection module; Step S43: embed the amphibious ship target detection module into the radar system to perform the amphibious ship target detection mission.

[0050] In the embodiment of the present invention, the target detection and recognition learning of the amphibious ship target detection model is carried out, and training is carried out based on a large amount of ship target data and marine environment data to extract the feature data of the object. During the training process, the model uses deep learning algorithms, especially convolutional neural networks (CNN) and other technologies to automatically learn the relationship between the ship and the surrounding environment and identify different types of target objects. The target detection and recognition learning obtains the object detection target data by gradually analyzing the training data set, and the data includes different ship types, the spatial position, speed and related feature information of the target object. The core technology of this stage is to optimize the target recognition algorithm through a large amount of labeled data, so that it has high recognition accuracy and generalization ability. Subsequently, according to the object detection target data, automatic code generation is performed to generate an amphibious ship target detection module. In this process, the object detection target data is used as input, and the algorithm code related to target detection is generated through automated programming tools (such as machine learning automation framework, code generation tool chain, etc.). These codes can realize the full process operation from data acquisition to target recognition, including data preprocessing, feature extraction, classification prediction and other links. Through templated and modular design, the automatic code generation enables the target detection system to be quickly deployed and adapt to different application scenarios and technical requirements. The key technology of code generation relies on advanced automated programming technology and the design of model interfaces, which makes the detection module highly adaptive. Finally, the generated amphibious ship target detection module is embedded in the radar system to perform the amphibious ship target detection task. At this stage, the target detection module is embedded in the core processing unit of the radar system. The radar system receives data from the sensor in real time and transmits it to the detection module for processing. At this point, the target detection module can quickly identify and track the detected targets according to the real-time data stream through the previously learned recognition model, generate corresponding target detection results and feed them back to the system, so as to effectively perform the detection task of amphibious ship targets.

[0051] In this specification, an amphibious ship target detection system is provided, which is used to perform the above-mentioned amphibious ship target detection method. The amphibious ship target detection system includes: The data acquisition and space construction module is used to obtain multi-modal sensor monitoring data; perform monitoring signal phase alignment processing on the multi-modal sensor monitoring data to generate multi-modal sensor phase alignment data; construct the ship's three-dimensional space monitoring field according to the multi-modal sensor phase alignment data to obtain the ship's three-dimensional space monitoring field; The motion trajectory analysis and path planning module is used to analyze the ship's motion trajectory in the three-dimensional space monitoring field of the ship and generate ship motion trajectory analysis data; perform avoidance path planning on the ship's motion trajectory analysis data and generate ship avoidance path data; The trajectory association and target detection model building module is used to obtain land area data; perform trajectory association cross analysis based on ship motion trajectory analysis data and ship avoidance path data to generate ship trajectory cross analysis data; perform a confidence assessment model on land area data and ship trajectory cross analysis data to generate an amphibious ship target detection model; The target detection and execution module is used to perform target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; the object detection target data is embedded into the radar system to execute the amphibious ship target detection task.

[0052] The beneficial effect of the present invention is that by acquiring multimodal sensor monitoring data and performing phase alignment processing on these data, the consistency of the monitoring signals of each sensor in time sequence is ensured, thereby reducing the data deviation caused by signal delay or synchronization error between different sensors. Then, based on the phase-aligned data, the three-dimensional space monitoring field of the ship is constructed, and a comprehensive and accurate ship monitoring model is formed through the fusion of spatial data, which provides high-quality basic data for subsequent trajectory analysis and path planning. In the second step, by analyzing the motion trajectory of the three-dimensional space monitoring field of the ship, detailed ship motion trajectory data is generated, and the avoidance path planning is implemented in combination with the motion law of the ship and the environmental changes, ensuring that the ship can autonomously avoid obstacles or other ships in different complex environments, thereby improving the safety and maneuverability of the ship. Next, by acquiring the land area data and cross-analyzing it with the ship's motion trajectory data, the ship trajectory cross-analysis data is formed, the ship path planning is further optimized, and a confidence assessment model is constructed based on this to ensure the accuracy and reliability of the data in practical applications, and finally an amphibious ship target detection model is generated. The model combines the target recognition algorithm to learn and identify the detection data, thereby realizing real-time monitoring and identification of targets on the water and land. Finally, by embedding the object detection target data into the radar system, a complete detection system is formed to perform the amphibious ship target detection task. Through this series of technical means, the present invention improves the target detection capability and avoidance efficiency of ships in complex environments. Therefore, by integrating multimodal data processing and intelligent path planning technology, the present invention solves the problems of low accuracy, slow response and poor environmental adaptability of traditional ship target detection systems, and improves the target recognition and obstacle avoidance capabilities of ships in multiple environments.

[0053] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0054] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for detecting an amphibious ship target, characterized in that: The following steps are involved: Step S1: acquiring multimodal sensor monitoring data; performing monitoring signal phase alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; constructing a three-dimensional ship space monitoring field according to the multimodal sensor phase alignment data to obtain a three-dimensional ship space monitoring field; Step S2: analyzing the ship motion trajectory of the ship's three-dimensional space monitoring field to generate ship motion trajectory analysis data; performing avoidance path planning on the ship motion trajectory analysis data to generate ship avoidance path data; Step S3: acquiring land area data; performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; performing a confidence assessment model on the land area data and the ship trajectory cross analysis data to generate an amphibious ship target detection model; Step S4: performing target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; Object detection target data is embedded into the radar system to perform amphibious ship target detection missions.

2. The amphibious ship target detection method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire multimodal sensor monitoring data; Step S12: performing monitoring signal alignment processing on the multimodal sensor monitoring data to generate multimodal sensor phase alignment data; performing monitoring object morphology recognition on the multimodal sensor phase alignment data to obtain monitoring object morphology recognition data; Step S13: constructing a three-dimensional ship space monitoring field according to the monitored object morphology recognition data to obtain a three-dimensional ship space monitoring field.

3. The amphibious ship target detection method according to claim 2, characterized in that: Step S13 includes the following steps: Step S131: performing geometric edge computing on the monitored object morphology recognition data to generate monitoring sensor edge computing data; Step S132: Analyze the geometric constraint relationship of the monitoring sensor edge computing data based on the synchronous positioning and mapping technology to generate the monitoring sensor geometric constraint data; perform spatiotemporal compensation on the monitoring sensor common view area of ​​the monitoring sensor geometric constraint data to generate the monitoring sensor spatiotemporal calibration data; Step S133: Acquire the location of the coastal feature point; perform spatial-temporal regional mapping between the monitoring sensor spatial calibration data and the coastal feature point location, and use particle swarm optimization to correct the sensor registration error to generate ship space sensor data; Step S134: Acquire a BEV bird's-eye view; use the BEV bird's-eye view to perform multimodal distance fusion on the ship space sensor data, and construct a three-dimensional monitoring field of the ship space to obtain a three-dimensional monitoring field of the ship space.

4. The amphibious ship target detection method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Performing spatial motion analysis of the target object according to the three-dimensional spatial monitoring field of the ship to obtain spatial movement data of the object; Step S22: performing ship motion trajectory analysis on the object space movement data to generate ship motion trajectory analysis data; Step S23: Calculate the motion fluid disturbance of the object space movement data and the ship motion trajectory analysis data to generate ship collision probability disturbance data; perform avoidance path planning based on the ship collision probability disturbance data to generate ship avoidance path data.

5. The amphibious ship target detection method according to claim 1, characterized in that: Step S22 includes the following steps: Step S221: Perform nonlinear Kalman filtering on the spatial movement data of the object to construct a motion state matrix, and generate a ship motion state prediction matrix; Step S222: performing frequency domain energy spectrum analysis on the ship motion state prediction matrix, extracting the main frequencies of roll and surge, and generating a frequency domain energy distribution spectrum of the ship; Step S223: Perform ship motion stability boundary analysis on the ship motion state prediction matrix through the Lyapunov index stability criterion to generate an acceleration increment safety threshold range; perform dynamic truncation correction processing on the ship frequency domain energy distribution spectrum according to the acceleration increment safety threshold range to generate ship motion trajectory analysis data.

6. The amphibious ship target detection method according to claim 5, characterized in that: The ship motion stability boundary analysis of the ship motion state prediction matrix using the Lyapunov index stability criterion includes the following: Extract the maximum Lyapunov exponent of the ship motion state prediction matrix to generate the sensitivity representation of the ship's initial conditions; The gradient field is calculated by characterizing the sensitivity of the ship's initial conditions to generate the flow field velocity gradient tensor; According to the velocity gradient tensor of the flow field, the implicit function relationship of the critical instability acceleration of the ship is constructed to generate the implicit function of the critical instability of the ship; The motion stability boundary analysis is carried out based on the critical instability implicit function of the ship, and the acceleration increment safety threshold range is generated.

7. The amphibious ship target detection method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire land area data; Step S32: performing trajectory correlation cross analysis based on the ship motion trajectory analysis data and the ship avoidance path data to generate ship trajectory cross analysis data; Step S33: performing early data level fusion on the land area data to generate amphibious ship model input data; performing model training based on the amphibious ship model input data, and calibrating the model confidence parameters to generate a confidence amphibious detection model; Step S34: Perform late feature fusion based on the ship trajectory cross analysis data to generate amphibious ship feature fusion data; perform trajectory matching weighted optimization on the amphibious ship feature fusion data for the confidence amphibious detection model to generate an amphibious ship target detection model.

8. The amphibious ship target detection method according to claim 7, characterized in that: Step S32 includes the following steps: Step S321: performing ship avoidance marking on the ship avoidance path data to generate ship avoidance mark data; performing deceleration-avoidance time feature two-dimensional feature extraction on the ship avoidance mark data to generate ship avoidance mark feature data; Step S322: performing time window matching on the ship avoidance mark feature data and the ship motion trajectory analysis data to generate ship time window avoidance-path matching data; performing Euclidean distance trajectory similarity metric matching on the ship time window avoidance-path matching data to generate ship motion trajectory-avoidance path matching data; Step S323: performing ship behavior pattern recognition on the ship motion trajectory-avoidance path matching data to generate ship motion behavior pattern feature data; performing avoidance type trajectory association cross analysis based on the ship motion behavior pattern feature data to generate ship trajectory cross analysis data.

9. The amphibious ship target detection method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; Step S42: Automatically generate code according to the object detection target data to generate an amphibious ship target detection module; Step S43: embed the amphibious ship target detection module into the radar system to perform the amphibious ship target detection mission.

10. An amphibious ship target detection system, characterized in that: Used to perform the amphibious ship target detection method according to claim 1, the amphibious ship target detection system comprises: The data acquisition and space construction module is used to obtain multi-modal sensor monitoring data; perform monitoring signal phase alignment processing on the multi-modal sensor monitoring data to generate multi-modal sensor phase alignment data; construct the ship's three-dimensional space monitoring field according to the multi-modal sensor phase alignment data to obtain the ship's three-dimensional space monitoring field; The motion trajectory analysis and path planning module is used to analyze the ship's motion trajectory in the three-dimensional space monitoring field of the ship and generate ship motion trajectory analysis data; perform avoidance path planning on the ship's motion trajectory analysis data and generate ship avoidance path data; The trajectory association and target detection model building module is used to obtain land area data; perform trajectory association cross analysis based on ship motion trajectory analysis data and ship avoidance path data to generate ship trajectory cross analysis data; perform a confidence assessment model on land area data and ship trajectory cross analysis data to generate an amphibious ship target detection model; The target detection and execution module is used to perform target detection and recognition learning on the amphibious ship target detection model to obtain object detection target data; the object detection target data is embedded into the radar system to execute the amphibious ship target detection task.

Images