Ship navigation monitoring management system and method based on edge calculation

Abstract

The invention relates to the technical field of ship safety, in particular to a ship navigation monitoring management system and method based on edge computing, and the system comprises seven cooperation modules: a data collection module synchronously obtains ship multi-source data, and a processing storage module achieves the data compression, marking and storage. The encryption security module completes full-process AES-256 encryption through a hardware encryption chip, the dual-power supply module guarantees stable power supply, the satellite positioning module obtains accurate positioning through Beidou / GPS dual modes, the self-inspection maintenance module monitors faults in real time, and the AI edge calculation unit achieves channel object recognition, track analysis and violation warning through a deep neural network model. According to the method, closed-loop processes of data acquisition, encryption, processing, intelligent analysis and fault handling are completed through module cooperation; the invention aims to solve the problems of high monitoring delay, insufficient data security, unreliable power supply and the like of the existing system.

Description

Technical Field

[0001] This invention relates to the field of ship safety technology, and in particular to a ship navigation monitoring and management system and method based on edge computing. Background Technology

[0002] As the core equipment for achieving this goal, the ship navigation monitoring system needs to acquire key data such as ship position, speed, and course in real time, and complete functions such as risk identification, status assessment, and alarm push, providing decision support for crew operations and shore-based management. It has become an important part of the modern shipping safety system.

[0003] However, existing ship navigation monitoring systems still face numerous technical bottlenecks in practical applications: most systems employ a shore-based centralized data processing architecture, requiring ship-collected navigation data to be transmitted to a shore-based platform for analysis via communication networks. In scenarios such as offshore and inland waterways where network signals are unstable or interrupted, data transmission delays and monitoring interruptions are prone to occur, leading to delayed alarms for violations and failing to meet real-time safety management requirements. Furthermore, data security protection measures are inadequate, relying heavily on software encryption or encrypting only the transmission process, lacking end-to-end security guarantees from data collection and transmission to storage, thus increasing the risk of ship navigation data leakage. The system faces several challenges: risks of exposure and tampering; a single power supply method, with traditional systems relying heavily on the ship's main power circuit, leading to system shutdown in the event of a main power failure; the use of a single GPS or BeiDou positioning system, which is susceptible to interference in scenarios such as bridge obstruction and complex terrain, resulting in insufficient positioning stability and accuracy; a lack of full-module self-checking and fault handling mechanisms, making it difficult to detect and address module failures in a timely manner, thus affecting the long-term reliable operation of the system; insufficient precision in intelligent monitoring, with risk identification relying heavily on simple threshold judgments, weak ability to identify complex objects in waterway scenarios, and room for improvement in the accuracy of trajectory analysis and violation determination.

[0004] To address the aforementioned issues, a patent with publication number CN117854324A was found, entitled "A Method and System for Shipboard Safety Risk Identification Based on Edge Computing." This solution acquires safety risk data through shipboard communication and navigation equipment and sensors, uses edge computing to categorize and identify personnel, ship, and environmental risk events, and pushes alarms through both shipboard and shore-based platforms, thus mitigating the latency problem of centralized shore-based processing to some extent and improving the real-time performance of risk identification. However, its core shortcomings lie in the lack of a dedicated hardware encryption module; data security protection relies solely on communication link encryption, failing to achieve end-to-end security. Furthermore, it lacks dual-power redundant supply and full-module self-testing and maintenance design, resulting in insufficient system power supply reliability and fault handling capabilities. The positioning method also does not explicitly adopt a dual-mode design, and its positioning stability in complex scenarios needs further verification. Existing solutions have failed to fully address the collaborative optimization issues in various aspects, such as monitoring latency, data security, power supply reliability, positioning stability, fault handling, and intelligent accuracy. Therefore, it is necessary to build an integrated ship navigation monitoring and management system that integrates edge computing, full-process encryption, redundant power supply, dual-mode positioning, and self-inspection and maintenance to meet the high reliability, high security, and high intelligence monitoring requirements of actual shipping scenarios. Summary of the Invention

[0005] This invention provides a ship navigation monitoring and management system and method based on edge computing, addressing the technical problems of existing ship navigation monitoring systems, such as high monitoring latency due to shore-based centralized processing, insufficient data transmission and storage security protection, poor power supply reliability, unsatisfactory positioning stability, difficulty in timely detection and handling of module failures, and insufficient intelligent monitoring accuracy.

[0006] To solve the above problems, the technical solution adopted by the invention is as follows:

[0007] A ship navigation monitoring and management system based on edge computing includes a data acquisition module, a processing and storage module, an encryption and security module, a dual power supply module, a satellite positioning module, a self-testing and maintenance module, and an AI edge computing unit.

[0008] The data acquisition module integrates multiple types of sensors and a universal interface to synchronously acquire the ship's position, speed, heading, and communication data.

[0009] The processing and storage module uses an embedded processor and a removable storage medium to achieve real-time compression, timestamp marking, and storage management of the collected data;

[0010] The encryption security module uses a hardware encryption chip to achieve AES-256 encryption protection for the entire process of data transmission and storage within the system;

[0011] The dual power supply module includes a main power supply circuit and a lithium battery backup power supply, which is used to provide stable power supply to each module of the system.

[0012] The satellite positioning module integrates a Beidou / GPS dual-mode transmitter, which is used to acquire ship positioning data in real time and transmit it to the processing and storage module;

[0013] The self-testing and maintenance module is used to monitor the operating status of each module of the system in real time and perform fault diagnosis.

[0014] The AI ​​edge computing unit is equipped with a multi-object detection algorithm engine and a built-in deep neural network model, which is used to realize object recognition, ship trajectory analysis and alarm of illegal navigation behavior in waterway scenarios;

[0015] The outputs of the data acquisition module and the satellite positioning module are both connected to the input of the processing and storage module. The processing and storage module communicates bidirectionally with the encryption security module and the AI ​​edge computing unit. The self-testing and maintenance module's detection end is connected to each module, and the output of the dual power supply module is connected to the power supply end of each module, forming a ship navigation monitoring and management architecture that works collaboratively on the edge side.

[0016] A ship navigation monitoring and management method based on edge computing includes the following steps:

[0017] Step 1: After the system starts up, the dual power supply module supplies power to each module, and the self-test and maintenance module performs initialization tests. After confirming that the status of each module is normal, it enters the monitoring mode.

[0018] Step 2: The data acquisition module and the satellite positioning module start synchronously. The data acquisition module collects ship speed, heading, and communication data through multiple types of sensors and general interfaces, while the satellite positioning module acquires ship positioning data through a Beidou / GPS dual-mode transmitter.

[0019] Step 3: The encryption security module performs AES-256 encryption on the collected raw data, and the encrypted ciphertext data is transmitted to the processing and storage module;

[0020] Step 4: After the storage module decrypts the encrypted data, the embedded processor performs real-time compression and timestamp marking, then encrypts and stores the processed data, and synchronously transmits it to the AI ​​edge computing unit.

[0021] Step 5: The AI ​​edge computing unit identifies objects in the waterway scene through a deep neural network model, analyzes the ship's trajectory using a trajectory prediction algorithm, and determines whether there is any illegal navigation behavior. If so, it immediately triggers an audible and visual alarm.

[0022] Step 6: The self-test and maintenance module monitors the operating status of each module in real time, continuously generates operating logs, and performs corresponding repairs or alarm handling when a fault is detected;

[0023] Step 7: The system continuously cycles through steps 2-6 to achieve real-time monitoring and management of ship navigation.

[0024] The principle and advantages of this scheme are as follows:

[0025] The core principle of this solution is to construct a collaborative edge-side ship navigation monitoring and management architecture, achieving closed-loop monitoring throughout the entire process through the organic linkage of seven functional modules. The data acquisition module, utilizing multiple types of sensors and universal interfaces, simultaneously captures ship position, speed, heading, and communication data, providing a comprehensive data source for monitoring. The satellite positioning module acquires precise positioning data via a BeiDou / GPS dual-mode transmitter, transmitting it along with the acquired data to the processing and storage module. The processing and storage module employs an embedded processor and removable storage media to perform real-time data compression, timestamping, and storage management, ensuring orderly data flow. The encryption and security module, based on a hardware encryption chip, performs AES-256 encryption protection throughout the entire process from data acquisition and transmission to storage, eliminating the risk of data leakage and tampering. The dual power supply module provides continuous and stable power to each module through redundant design of the main power supply circuit and lithium battery backup power supply. The self-checking and maintenance module monitors the operating status of all modules in real time, promptly detecting and diagnosing faults. The AI ​​edge computing unit, equipped with a multi-target detection algorithm engine and a deep neural network model, performs object recognition in the waterway scene, ship trajectory analysis, and violation alarms locally on the ship, without relying on a shore-based platform.

[0026] Compared to existing technologies, this solution completely solves the latency problem of traditional shore-based centralized processing through an edge-side collaborative architecture. The AI ​​edge computing unit realizes real-time data analysis and alarms locally on the ship. Compared to the shore-based processing mode in existing technologies, the alarm response speed is reduced from seconds to milliseconds. It can still work normally even in network interruption scenarios such as in the open sea, avoiding monitoring failure. The end-to-end hardware encryption design has significant advantages. Existing technologies mostly use software encryption or partial encryption, while this solution uses a hardware encryption chip to achieve AES-256 end-to-end encryption, which has stronger anti-cracking capabilities and can effectively protect the security of sensitive information such as navigation data, solving the pain point of incomplete data protection in existing solutions. The collaborative design of power supply and self-test enhances system reliability. The seamless switching design of the dual power supply module solves the problem of easy interruption in traditional single power supply. The self-test and maintenance module enables early detection and handling of faults. Compared with existing systems that lack self-test functions, it significantly reduces operation and maintenance costs and the probability of monitoring interruption. Fourth, dual-mode positioning and AI intelligent analysis improve monitoring accuracy. The Beidou / GPS dual-mode design avoids the signal interference problem of single positioning mode. The positioning stability is far superior to existing single-mode systems. The object recognition and trajectory analysis driven by the deep neural network model are more adaptable to complex waterway scenarios than traditional threshold judgment methods, and the accuracy of violation recognition is significantly improved.

[0027] Furthermore, the removable storage medium of the processing storage module is an SSD solid-state drive or a TF memory card. The embedded processor uses the LZ77 lossless compression algorithm to compress the collected data in real time. The compressed data carries a unique timestamp and is stored in chronological order. SSD solid-state drives and TF memory cards combine high read and write speeds, small size, and stable performance, making them suitable for the complex environment of ship navigation and meeting the high-speed storage requirements of multi-source monitoring data. At the same time, their portability facilitates offline data export and backup. The LZ77 lossless compression algorithm can significantly reduce data volume without sacrificing data accuracy, thus reducing the space occupied by the storage medium. This reduces storage costs, and the algorithm has low complexity and high computational efficiency. It can be used with embedded processors to achieve real-time data compression, avoiding the impact of data backlog on the overall system operating efficiency. The unique timestamp and time-series storage design can accurately mark the collection time of each set of data, ensuring data traceability. This facilitates subsequent review and analysis of ship navigation trajectories, equipment operating status, etc., and provides a time-series coherent and logically clear data foundation for trajectory analysis of AI edge computing units. At the same time, the ordered storage mode can improve data retrieval speed, allowing the system to quickly locate and extract historical data when needed, further ensuring the real-time performance of monitoring and analysis.

[0028] Furthermore, the hardware encryption chip of the encryption security module is a high-performance encryption chip. Its full-process encryption process includes: after data acquisition, the hardware encryption chip generates a random key and encrypts the original data using the AES-256 algorithm to obtain ciphertext; after the ciphertext is transmitted to the processing and storage module, the hardware encryption chip calls a preset key to decrypt it; the processed valid data is then encrypted again and stored in the storage medium. The computing efficiency of the high-performance hardware encryption chip far exceeds that of software encryption, and it can complete encryption and decryption operations without affecting the real-time data processing speed of the system, avoiding data backlog caused by excessive resource consumption during the encryption process; the generation of a random key for encryption after data acquisition ensures that the encryption key for each batch of original data is unique, and even if the key for a single batch of data is leaked, it will not affect the security of other data. With its comprehensive and robust encryption capabilities, coupled with the high-strength encryption features of the AES-256 algorithm, it effectively resists attacks such as brute-force attacks and data tampering. Data is transmitted in encrypted form during transmission, eliminating the risk of theft or alteration during inter-module transmission. Further encryption after processing ensures full lifecycle security for the data, addressing the vulnerability of some existing solutions that only encrypt the transmission or storage stage. The design with a pre-set key for decryption ensures that the processing and storage modules can read data for compression, tagging, and other operations, while hardware chip-specific key management prevents key leakage. This ensures that sensitive ship navigation data is securely protected at every stage from collection and transmission to storage, fully meeting the stringent data security requirements of the shipping industry.

[0029] Furthermore, the dual power supply module also includes a power management unit, which collects the voltage signal of the main power supply circuit in real time. When the main power supply voltage is detected to be lower than a preset threshold, the power management unit will... At that time, the lithium battery backup power supply is triggered and a seamless switch is completed, with a switching delay. Furthermore, the lithium battery backup power supply can ensure continuous system operation. The power management unit's real-time monitoring function enables accurate prediction of the main power supply status, preventing system shutdown caused by sudden interruption of the main power supply. The ultra-short switching latency ensures uninterrupted operation of core processes such as data acquisition, encryption processing, and AI analysis, preventing data loss or monitoring lag due to power supply switching. Compared to existing power supply solutions without fast switching capabilities, it solves the problem of monitoring failure during power transition. The ≥4-hour backup endurance provides sufficient emergency monitoring time for ships after a main power supply failure, ensuring timely restoration of the main power supply during near-shore navigation and meeting the needs of continuous monitoring during long-distance navigation. It avoids the risks of navigation data gaps and missed reports of violations caused by sudden failures in traditional single power supply systems, significantly improving the system's power supply reliability and emergency support capabilities in complex shipping scenarios.

[0030] Furthermore, the deep neural network model of the AI ​​edge computing unit is an improved YOLOv7 model, and the multi-object detection algorithm engine realizes ship trajectory analysis through a trajectory prediction algorithm. The trajectory prediction algorithm formula is:

[0031]

[0032] In the formula, Let be the predicted position coordinates of the ship at time t. for The ship's actual position coordinates at that moment for The ship's actual speed at that moment The actual heading angle of the ship at time t-1. The time interval between two consecutive data collections. This is an environmental correction factor, dynamically adjusted based on channel wind speed and current velocity, with a value ranging from 0.9 to 1.1; the actual coordinates of the ship at time t. Deviation value from predicted location coordinates >Preset threshold When this occurs, it is determined to be a trajectory anomaly, where the formula for calculating the deviation value is:

[0033] The improved YOLOv7 model combines high detection accuracy with fast computing speed, adapting to the local processing needs of edge computing units. It can accurately identify various objects such as ships, buoys, and obstacles in the waterway, providing a reliable foundation for trajectory analysis. The trajectory prediction algorithm fully combines the ship's motion characteristics with the influence of the waterway environment. Compared with the traditional trajectory judgment method that only relies on position data, the prediction accuracy is greatly improved. The environmental correction coefficient can effectively compensate for the interference of wind and water flow on ship navigation and avoid misjudgment caused by environmental factors. The deviation value quantification judgment standard is clear and specific. The preset threshold can not only capture real trajectory anomalies such as deviation from the course and sudden turning in time, but also filter out invalid alarms caused by minor position fluctuations. Compared with the existing fuzzy risk identification scheme, the accuracy and reliability of trajectory anomaly judgment are significantly improved. The entire analysis process is completed locally on the ship, without relying on the computing power support of the shore-based platform. The response speed is fast, and it can still work stably even in the case of network interruption. It can detect ship trajectory anomalies and trigger alarms in the first time, giving the crew time to adjust the navigation status in time and avoid collision or grounding risks, greatly improving the safety and controllability of ship navigation.

[0034] Furthermore, the aforementioned violations include abnormal navigation, speeding, and insufficient safe distance. The criterion for determining speeding is the actual speed of the vessel. , Based on the waterway classification plan, the inland waterway is classified as Class III. =12km / h, coastal Class II waterway =18km / h; Insufficient safe distance is determined by the following formula:

[0035] In the formula, The minimum safe distance is defined as follows: L is the overall length of the vessel, B is the maximum width of the vessel, v is the current speed of the vessel, and β is the channel classification coefficient: β=1.5 for Class I channels, β=1.3 for Class II channels, and β=1.1 for Class III and below channels; the actual distance between the vessel and other objects within the channel. When deemed insufficient safe distance, the categorization of violations covers the most common safety risks in ship navigation and clarifies the judgment logic, avoiding the problems of vague definitions, omissions, and misjudgments found in existing schemes. The speeding threshold, combined with differentiated settings based on waterway grade, aligns with the navigation capacity and safety requirements of different waterways, making it more reasonable and applicable than a uniform speed limit standard. It effectively constrains ship speeds in various waterway scenarios. The safe distance calculation formula comprehensively considers the ship's overall length L, maximum width B, current speed v (higher speeds require longer braking distances and higher safety redundancy), and the waterway grade coefficient β (Class I waterways β=1.5, Class II waterways β=1.3, Class III and below β=1.1), adapting to the ship density and navigation complexity of different waterways to ensure a minimum safe distance. The calculations are more in line with actual navigation scenarios and are more scientific than fixed safety distance standards. The quantitative judgment criteria and calculation formulas make the identification of violations more accurate and objective, reducing the subjectivity of human judgment. Combined with the real-time analysis capabilities of the AI ​​edge computing unit, it can quickly capture various violations and trigger alarms.

[0036] Furthermore, the BeiDou / GPS dual-mode transmitter of the satellite positioning module supports automatic satellite selection and signal switching, improving positioning accuracy. The location data update frequency is 10Hz; when the signal strength of one of the location modes is lower than... When the time is right, it automatically switches to another positioning mode, and the response time changes accordingly. The dual-mode collaborative design breaks through the signal coverage limitations of a single positioning system. In complex scenarios such as bridge obstruction, mountainous waterways, and open seas, it can obtain stable signals through automatic satellite selection and optimization. Even if the single-mode signal is attenuated or interrupted, it can quickly switch to ensure positioning continuity, completely solving the positioning failure problem that is prone to occur in traditional single-mode positioning systems. Its high-precision positioning and 10Hz high update frequency provide accurate and real-time location data support for ship trajectory analysis and violation judgment, making trajectory deviation calculation and safety distance assessment more reliable and avoiding misjudgment or omission due to positioning errors; the ultra-short switching response time of ≤500ms ensures that the positioning data has no obvious gaps and there will be no data delay due to mode switching. Combined with the full-scene signal adaptation capability, it realizes stable and accurate positioning throughout the ship's navigation.

[0037] Furthermore, the self-testing and maintenance module includes a status monitoring submodule and a fault handling submodule. The status monitoring submodule collects the power supply voltage, operating current, and data transmission rate of each module in real time and generates an operating status log. The fault handling submodule performs automatic repair for minor faults and generates fault codes for serious faults, which are then sent to the shore-based management platform via satellite communication. The status monitoring submodule collects the power supply voltage, operating current, and data transmission rate of each module in real time and generates an operating status log, enabling comprehensive detection of early signs of abnormal module operation. Compared to traditional systems lacking active monitoring, this achieves early fault detection and early warning. The fault handling submodule is designed for... Minor faults such as loose connections and signal attenuation are automatically repaired, allowing the system to quickly return to normal operation and preventing small faults from escalating into major monitoring problems. For serious faults such as module power failure and chip damage, a unique fault code is generated and pushed to the shore-based management platform via the satellite communication module, enabling shore-based personnel to accurately locate the fault type and location, significantly shortening troubleshooting and repair time. The entire design reduces the workload of crew members in operation and maintenance, and avoids problems such as data acquisition interruption, decreased monitoring accuracy, or alarm failure caused by untimely handling of module faults. It significantly improves the stability and maintainability of the system in long-term navigation scenarios, ensuring the continuous reliability of ship navigation monitoring.

[0038] Furthermore, the AI ​​edge computing unit also includes a model optimization submodule, which dynamically updates the parameters of the deep neural network model using an online incremental learning algorithm. The update formula is as follows: ,

[0039] In the formula, Let be the model parameters updated at time t. Here are the model parameters at time t-1, and η is the learning rate, ranging from 0.001 to 0.01. For loss function exist gradient at, The number of samples for a single incremental learning iteration. Let i be the true label of the i-th sample. The probability of the model predicting the i-th sample is calculated by the deviation between the model's prediction and the actual waterway scenario data. This submodule is based on the actual waterway scenario data and uses a loss function. The deviation between the model's predicted results and the true labels is calculated, and the model parameters are dynamically updated based on a learning rate η within a reasonable range (0.001-0.01). Compared to traditional neural network models with fixed parameters, this new model achieves adaptive optimization for complex waterway environments. It can complete online fine-tuning without manual intervention and continuously absorb data information such as object features and navigation trajectory patterns in new scenarios, effectively solving the problem of decreased recognition accuracy of traditional models in unfamiliar waterways and special weather conditions. As navigation data accumulates, the model's object recognition accuracy, trajectory analysis precision, and violation judgment reliability will continue to improve, allowing the system's performance to be continuously optimized during long-term use. Compared to existing technologies that require offline model updates, this significantly reduces operation and maintenance costs while ensuring the accuracy and stability of monitoring in different navigation scenarios. Attached Figure Description

[0040] Figure 1 This is a modular flowchart of the system of the present invention.

[0041] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0042] Example 1

[0043] As attached Figure 1-2 As shown, a ship navigation monitoring and management system based on edge computing includes a data acquisition module, a processing and storage module, an encryption and security module, a dual power supply module, a satellite positioning module, a self-testing and maintenance module, and an AI edge computing unit.

[0044] The data acquisition module integrates multiple types of sensors and a universal interface to synchronously acquire the ship's position, speed, heading, and communication data.

[0045] The processing and storage module uses an embedded processor and a removable storage medium to achieve real-time compression, timestamp marking, and storage management of the collected data;

[0046] The encryption security module uses a hardware encryption chip to achieve AES-256 encryption protection for the entire process of data transmission and storage within the system;

[0047] The dual power supply module includes a main power supply circuit and a lithium battery backup power supply, which is used to provide stable power supply to each module of the system.

[0048] The satellite positioning module integrates a Beidou / GPS dual-mode transmitter, which is used to acquire ship positioning data in real time and transmit it to the processing and storage module;

[0049] The self-testing and maintenance module is used to monitor the operating status of each module of the system in real time and perform fault diagnosis.

[0050] The AI ​​edge computing unit is equipped with a multi-object detection algorithm engine and a built-in deep neural network model, which is used to realize object recognition, ship trajectory analysis and alarm of illegal navigation behavior in waterway scenarios;

[0051] The outputs of the data acquisition module and the satellite positioning module are both connected to the input of the processing and storage module. The processing and storage module communicates bidirectionally with the encryption security module and the AI ​​edge computing unit. The self-testing and maintenance module's detection end is connected to each module, and the output of the dual power supply module is connected to the power supply end of each module, forming a ship navigation monitoring and management architecture that works collaboratively on the edge side.

[0052] A ship navigation monitoring and management method based on edge computing includes the following steps:

[0053] Step 1: After the system starts up, the dual power supply module supplies power to each module, and the self-test and maintenance module performs initialization tests. After confirming that the status of each module is normal, it enters the monitoring mode.

[0054] Step 2: The data acquisition module and the satellite positioning module start synchronously. The data acquisition module collects ship speed, heading, and communication data through multiple types of sensors and general interfaces, while the satellite positioning module acquires ship positioning data through a Beidou / GPS dual-mode transmitter.

[0055] Step 3: The encryption security module performs AES-256 encryption on the collected raw data, and the encrypted ciphertext data is transmitted to the processing and storage module;

[0056] Step 4: After the storage module decrypts the encrypted data, the embedded processor performs real-time compression and timestamp marking, then encrypts and stores the processed data, and synchronously transmits it to the AI ​​edge computing unit.

[0057] Step 5: The AI ​​edge computing unit identifies objects in the waterway scene through a deep neural network model, analyzes the ship's trajectory using a trajectory prediction algorithm, and determines whether there is any illegal navigation behavior. If so, it immediately triggers an audible and visual alarm.

[0058] Step 6: The self-test and maintenance module monitors the operating status of each module in real time, continuously generates operating logs, and performs corresponding repairs or alarm handling when a fault is detected;

[0059] Step 7: The system continuously cycles through steps 2-6 to achieve real-time monitoring and management of ship navigation.

[0060] The principle of this solution is as follows: the data acquisition module uses multiple types of sensors and universal interfaces to simultaneously capture ship position, speed, heading, and communication data, providing a comprehensive data source for monitoring; the satellite positioning module acquires precise positioning data through a Beidou / GPS dual-mode transmitter and transmits it to the processing and storage module along with the acquired data; the processing and storage module uses an embedded processor and removable storage media to complete real-time data compression, timestamp marking, and storage management, ensuring orderly data flow; the encryption and security module, based on a hardware encryption chip, performs AES-256 encryption protection on the entire process of data acquisition, transmission, and storage, eliminating the risk of data leakage and tampering; the dual power supply module provides continuous and stable power to each module through the redundant design of the main power supply circuit and the lithium battery backup power supply; the self-testing and maintenance module monitors the operating status of all modules in real time, promptly detecting and diagnosing faults; the AI ​​edge computing unit is equipped with a multi-object detection algorithm engine and a deep neural network model, completing object recognition in the waterway scene, ship trajectory analysis, and violation alarms locally on the ship, without relying on a shore-based platform.

[0061] Compared to existing technologies, this solution completely solves the latency problem of traditional shore-based centralized processing through an edge-side collaborative architecture. The AI ​​edge computing unit realizes real-time data analysis and alarms locally on the ship. Compared to the shore-based processing mode in existing technologies, the alarm response speed is reduced from seconds to milliseconds. It can still work normally even in network interruption scenarios such as in the open sea, avoiding monitoring failure. The end-to-end hardware encryption design has significant advantages. Existing technologies mostly use software encryption or partial encryption, while this solution uses a hardware encryption chip to achieve AES-256 end-to-end encryption, which has stronger anti-cracking capabilities and can effectively protect the security of sensitive information such as navigation data, solving the pain point of incomplete data protection in existing solutions. The collaborative design of power supply and self-test enhances system reliability. The seamless switching design of the dual power supply module solves the problem of easy interruption in traditional single power supply. The self-test and maintenance module enables early detection and handling of faults. Compared with existing systems that lack self-test functions, it significantly reduces operation and maintenance costs and the probability of monitoring interruption. Fourth, dual-mode positioning and AI intelligent analysis improve monitoring accuracy. The Beidou / GPS dual-mode design avoids the signal interference problem of single positioning mode. The positioning stability is far superior to existing single-mode systems. The object recognition and trajectory analysis driven by the deep neural network model are more adaptable to complex waterway scenarios than traditional threshold judgment methods, and the accuracy of violation recognition is significantly improved.

[0062] The removable storage medium of the processing storage module is an SSD solid-state drive or a TF memory card. The embedded processor uses the LZ77 lossless compression algorithm to compress the collected data in real time. The compressed data carries a unique timestamp and is stored in chronological order. SSD solid-state drives and TF memory cards combine high read and write speeds, small size, and stable performance, making them suitable for the complex environment of ship navigation and meeting the high-speed storage requirements of multi-source monitoring data. Their portability also facilitates offline data export and backup. The LZ77 lossless compression algorithm can significantly reduce data volume without sacrificing data accuracy, reducing the space occupied by the storage medium and lowering costs. With low storage costs, low algorithm complexity, and high computational efficiency, it can be used with embedded processors to achieve real-time data compression, avoiding the impact of data backlog on the overall system operating efficiency. The unique timestamp and time-series storage design can accurately mark the collection time of each set of data, ensuring data traceability. This facilitates subsequent review and analysis of ship navigation trajectories, equipment operating status, etc., and provides a time-coherent and logically clear data foundation for trajectory analysis of AI edge computing units. At the same time, the ordered storage mode can improve data retrieval speed, allowing the system to quickly locate and extract historical data when needed, further ensuring the real-time performance of monitoring and analysis.

[0063] The hardware encryption chip of the encryption security module is a high-performance encryption chip. Its full-process encryption includes: after data acquisition, the hardware encryption chip generates a random key and encrypts the original data using the AES-256 algorithm to obtain ciphertext; after the ciphertext is transmitted to the processing and storage module, the hardware encryption chip decrypts it using a preset key; the processed valid data is then encrypted again and stored in the storage medium. The high-performance hardware encryption chip has a much higher computational efficiency than software encryption, and can complete encryption and decryption operations without affecting the system's real-time data processing speed, avoiding data backlog caused by excessive resource consumption during the encryption process; the generation of random keys for encryption after data acquisition ensures that the encryption key for each batch of original data is unique, and even if the key for a single batch of data is leaked, it will not affect the security of other data. Combined with the high-strength encryption of the AES-256 algorithm, it can effectively resist attacks such as brute-force cracking and data tampering. During transmission, it flows in encrypted form, which can eliminate the risk of data being stolen or tampered with in the transmission link between modules. The re-encryption and storage after processing realizes the security protection of the entire data life cycle, which solves the protection vulnerability of some existing solutions that only encrypt the transmission or storage stage. The design of the preset key for decryption processing not only ensures that the processing and storage modules can read the data normally for compression, marking and other operations, but also avoids key leakage through dedicated key management of the hardware chip. This ensures that the sensitive information of ship navigation data is under security protection at every stage from collection, transmission to storage, and fully meets the stringent data security requirements in the shipping scenario.

[0064] The dual power supply module also includes a power management unit, which collects the voltage signal of the main power supply circuit in real time. When the main power supply voltage is detected to be lower than a preset threshold, the power management unit will detect the voltage signal. When the main power supply fails, the lithium battery backup power supply is triggered and seamlessly switched over with a switching delay of ≤10ms. The lithium battery backup power supply can ensure continuous operation of the system for ≥4 hours. The real-time monitoring function of the power management unit enables accurate prediction of the main power supply status, avoiding system shutdown caused by sudden interruption of the main power supply. The ultra-short switching delay of ≤10ms ensures uninterrupted operation of core processes such as data acquisition, encryption processing, and AI analysis, without data loss or monitoring lag due to power supply switching. Compared with existing power supply solutions without fast switching function, it solves the problem of monitoring failure during power supply transition. The backup endurance of ≥4 hours provides sufficient emergency monitoring time for ships after main power supply failure. It ensures timely repair of main power supply during near-shore navigation and meets the needs of continuous monitoring during ocean navigation. It avoids the risks of navigation data gaps and missed reporting of violations caused by sudden failures in traditional single power supply systems, and greatly improves the power supply reliability and emergency support capability of the system in complex shipping scenarios.

[0065] The deep neural network model of the AI ​​edge computing unit is an improved YOLOv7 model. The multi-object detection algorithm engine realizes ship trajectory analysis through trajectory prediction algorithm. The trajectory prediction algorithm formula is:

[0066]

[0067] In the formula, Let be the predicted position coordinates of the ship at time t. The coordinates of the ship's actual position at time t-1 The actual speed of the ship at time t-1 The actual heading angle of the ship at time t-1. The time interval between two consecutive data collections. This is an environmental correction factor, dynamically adjusted based on channel wind speed and current velocity, with a value ranging from 0.9 to 1.1; the actual coordinates of the ship at time t. Deviation value from predicted location coordinates >Preset threshold When this occurs, it is determined to be a trajectory anomaly, where the formula for calculating the deviation value is:

[0068] The improved YOLOv7 model combines high detection accuracy with fast computing speed, adapting to the local processing needs of edge computing units. It can accurately identify various objects such as ships, buoys, and obstacles in the waterway, providing a reliable foundation for trajectory analysis. The trajectory prediction algorithm fully combines the ship's motion characteristics with the influence of the waterway environment. Compared with the traditional trajectory judgment method that only relies on position data, the prediction accuracy is greatly improved. The environmental correction coefficient can effectively compensate for the interference of wind and water flow on ship navigation and avoid misjudgment caused by environmental factors. The deviation value quantification judgment standard is clear and specific. The preset threshold can not only capture real trajectory anomalies such as deviation from the course and sudden turning in time, but also filter out invalid alarms caused by minor position fluctuations. Compared with the existing fuzzy risk identification scheme, the accuracy and reliability of trajectory anomaly judgment are significantly improved. The entire analysis process is completed locally on the ship, without relying on the computing power support of the shore-based platform. The response speed is fast, and it can still work stably even in the case of network interruption. It can detect ship trajectory anomalies and trigger alarms in the first time, giving the crew time to adjust the navigation status in time and avoid collision or grounding risks, greatly improving the safety and controllability of ship navigation.

[0069] The aforementioned violations include abnormal navigation, speeding, and insufficient safe distance. The criterion for determining speeding is the actual speed of the vessel. ( Based on the waterway classification plan, the inland waterway is classified as Class III. =12km / h, coastal Class II waterway =18km / h); Insufficient safe distance is determined by the following formula:

[0070] In the formula, The minimum safe distance is defined as follows: L is the overall length of the vessel, B is the maximum width of the vessel, v is the current speed of the vessel, and β is the channel classification coefficient: β=1.5 for Class I channels, β=1.3 for Class II channels, and β=1.1 for Class III and below channels; the actual distance between the vessel and other objects within the channel. When deemed insufficient safe distance, the categorization of violations covers the most common safety risks in ship navigation and clarifies the judgment logic, avoiding the problems of vague definitions, omissions, and misjudgments found in existing schemes. The speeding threshold, combined with differentiated settings based on waterway grade, aligns with the navigation capacity and safety requirements of different waterways, making it more reasonable and applicable than a uniform speed limit standard. It effectively constrains ship speeds in various waterway scenarios. The safe distance calculation formula comprehensively considers the ship's total length L, maximum width B (reflecting the ship's own occupancy), current speed v (higher speeds require longer braking distances and higher safety redundancy), and waterway grade coefficient β (Class I waterways β=1.5, Class II waterways β=1.3, Class III and below β=1.1), adapting to the ship density and navigation complexity of different waterways to ensure a minimum safe distance. The calculations are more in line with actual navigation scenarios and are more scientific than fixed safety distance standards. The quantitative judgment criteria and calculation formulas make the identification of violations more accurate and objective, reducing the subjectivity of human judgment. Combined with the real-time analysis capabilities of the AI ​​edge computing unit, it can quickly capture various violations and trigger alarms.

[0071] The BeiDou / GPS dual-mode transmitter of the satellite positioning module supports automatic satellite selection and signal switching, with a positioning accuracy of ≤1m and a positioning data update frequency of 10Hz; when the signal strength of one of the positioning modes is lower than - When switching to another positioning mode, the system automatically switches to a different mode with a switching response time of ≤500ms. The dual-mode collaborative design overcomes the signal coverage limitations of a single positioning system. In complex scenarios such as bridge obstruction, mountainous waterways, and open seas, it can automatically select and optimize satellites to obtain stable signals. Even if the single-mode signal weakens or is interrupted, it can quickly switch to ensure positioning continuity, completely solving the positioning failure problem that is prone to occur in traditional single-mode positioning systems. The high-precision positioning of ≤1m and the high update frequency of 10Hz can provide accurate and real-time position data support for ship trajectory analysis and violation judgment, making trajectory deviation calculation and safe distance assessment more reliable and avoiding misjudgments and omissions caused by positioning errors. The ultra-short switching response time of ≤500ms ensures that there are no obvious gaps in positioning data and no data delays due to mode switching. Combined with the full-scene signal adaptation capability, it achieves stable and accurate positioning throughout the ship's navigation.

[0072] The self-testing and maintenance module includes a status monitoring submodule and a fault handling submodule. The status monitoring submodule collects the power supply voltage, operating current, and data transmission rate of each module in real time and generates an operational status log. The fault handling submodule automatically repairs minor faults and generates fault codes for serious faults, which are then sent to the shore-based management platform via satellite communication. The status monitoring submodule's real-time collection of power supply voltage, operating current, and data transmission rate of each module and generation of operational status logs allows for comprehensive detection of early signs of module malfunctions. Compared to traditional systems lacking proactive monitoring, this enables early fault detection and warning. The fault handling submodule is designed for systems with loose interfaces. Minor faults such as dynamic and signal attenuation can be automatically repaired, quickly restoring the system to normal operation and preventing small faults from escalating into major problems affecting monitoring. For serious faults such as module power failure and chip damage, a unique fault code is generated and pushed to the shore-based management platform via satellite communication module, allowing shore-based personnel to accurately locate the fault type and location, significantly shortening the time for fault diagnosis and repair. The entire design reduces the workload of crew members in operation and maintenance, and avoids problems such as data acquisition interruption, decreased monitoring accuracy, or alarm failure caused by untimely handling of module faults. It significantly improves the stability and maintainability of the system in long-term navigation scenarios, ensuring the continuous reliability of ship navigation monitoring.

[0073] The AI ​​edge computing unit also includes a model optimization submodule, which dynamically updates the parameters of the deep neural network model using an online incremental learning algorithm. The update formula is as follows: ,

[0074] In the formula, Let be the model parameters updated at time t. Here are the model parameters at time t-1, and η is the learning rate (ranging from 0.001 to 0.01). For loss function exist gradient at, The number of samples for a single incremental learning iteration. Let i be the true label of the i-th sample. The probability of the model predicting the i-th sample is calculated by the deviation between the model's prediction and the actual waterway scenario data. This submodule is based on the actual waterway scenario data and uses a loss function. The deviation between the model's predicted results and the true labels is calculated, and the model parameters are dynamically updated based on a learning rate η within a reasonable range (0.001-0.01). Compared to traditional neural network models with fixed parameters, this new model achieves adaptive optimization for complex waterway environments. It can complete online fine-tuning without manual intervention and continuously absorb data information such as object features and navigation trajectory patterns in new scenarios, effectively solving the problem of decreased recognition accuracy of traditional models in unfamiliar waterways and special weather conditions. As navigation data accumulates, the model's object recognition accuracy, trajectory analysis precision, and violation judgment reliability will continue to improve, allowing the system's performance to be continuously optimized during long-term use. Compared to existing technologies that require offline model updates, this significantly reduces operation and maintenance costs while ensuring the accuracy and stability of monitoring in different navigation scenarios.

[0075] I. System Hardware Selection and Parameter Configuration

[0076] In this embodiment, the hardware selection for each module of the edge computing-based ship navigation monitoring and management system strictly adheres to the requirements of the ship navigation environment for anti-interference, high and low temperature resistance, and stability. The specific configuration is as follows:

[0077] Data acquisition module

[0078] Sensor type and parameters: Integrated GNSS-assisted sensor (positioning accuracy) Hall effect speed sensor (measurement range) precision ), electronic compass (heading measurement range) resolution VHF communication data acquisition module (operating frequency) It is also equipped with RS485, CAN and Ethernet universal interfaces to support data interaction with existing ship navigation equipment (such as radar and AIS).

[0079] Working environment: operating temperature relative humidity (Non-condensing), IP67 protection rating, meeting the requirements for use in complex environments such as ship decks and engine rooms.

[0080] Processing storage module

[0081] Embedded processor: Employs STM32H743VI high-performance microcontroller, with a main frequency of... It supports single-precision / double-precision floating-point operations and has sufficient computing power to support the LZ77 lossless compression algorithm and real-time data processing.

[0082] Removable storage media: 1TB SSD solid-state drive (read speed...) Write speed It also reserves a TF card interface (supporting up to 256GB TF cards) to achieve dual data backup.

[0083] Compression and storage parameters: The sliding window size of the LZ77 algorithm is set to 32KB, and the dictionary size is 64KB to ensure that compression is achieved without loss of data precision. The data compression rate is high; the timestamps use UTC time format, accurate to the millisecond level (format: YYYY-MM-DDHH:MM:SS.XXX), and the data is stored in a hierarchical directory according to "date-hour" for easy and fast retrieval.

[0084] Encryption security module

[0085] Hardware encryption chip: Utilizes the SM4-AES256 high-performance encryption chip, supporting multiple encryption algorithms such as AES-256 / 128 and SM4, with a high processing speed. Operating temperature It meets military-grade safety standards.

[0086] Key management mechanism: Random keys are generated using a true random number generator (entropy value). The default key is stored in the chip's built-in secure storage area (capacity 16KB), and supports key update and destruction functions. The key update cycle can be remotely configured through the shore-based platform (default 7 days automatic update).

[0087] Dual power supply module

[0088] Main power supply circuit: Input voltage range Output voltage (ripples) Rated output current It is equipped with an EMI filter to suppress power grid interference.

[0089] Lithium battery backup power: uses lithium iron phosphate power battery packs Cycle life Secondly, it is equipped with a battery management system (BMS) to monitor battery voltage, current, and temperature in real time (monitoring accuracy: voltage). Current ,temperature ).

[0090] Power Management Unit: Selected Power management chip, preset threshold (Allowable deviation) ), switching delay When the lithium battery backup power supply is used alone, the system power consumption is controlled within 50W, ensuring continuous operation for ≥4 hours.

[0091] Satellite positioning module

[0092] Dual-mode transmitter: Selected High-precision positioning module, supporting BeiDou-3, Dual-frequency positioning, positioning accuracy , Location data update frequency (Configurable to) ).

[0093] Signal switching parameters: Signal strength monitoring period 100ms; when the signal strength of a certain mode is lower than 1 for 3 consecutive periods... When the time is right, it automatically switches to another mode, changing the response time. During the switching process, a data interpolation algorithm is used to ensure the continuity of positioning data.

[0094] Assisted positioning: Integrated SBAS (Satellite Augmentation System), supports , , This further improves positioning accuracy in complex terrain.

[0095] Self-test and maintenance module

[0096] Status monitoring submodule: adopts Low-power microcontroller, real-time acquisition of power supply voltage (monitoring range 0-30V, accuracy ±0.05V) and operating current (monitoring range) of each module. precision ), data transmission rate (monitoring range) ), sampling period The generated runtime status logs are stored daily and retained for 30 days.

[0097] Fault handling submodule: Built-in fault code library containing 128 common faults. Minor faults, such as loose interfaces or signal attenuation ≤3dB, are automatically repaired by restarting the module or recalibrating parameters, with a high success rate. In case of severe faults, such as module power outages or data transmission rates remaining at 0 for 10 seconds, a 16-bit fault code is generated. The first 8 bits identify the module, and the last 8 bits identify the fault type. This code is then pushed to the shore-based management platform in real time via satellite communication. (Push delay is not specified.) Second.

[0098] AI edge computing unit

[0099] Hardware platform: Selected Edge computing module, equipped with 6 cores 384 cores 8GB of video memory, 21 TOPS computing power, supports The computation meets the real-time inference requirements of the improved YOLOv7 model.

[0100] Deep Neural Network Models: Improved Model input resolution , use , Partial increase structure, Partial optimization loss function (using) The model weight file size is approximately 200MB, and the inference speed is... (After INT8 quantization) Object recognition accuracy (The test set includes 10 types of common objects in waterways, such as ships, buoys, reefs, and bridges.)

[0101] Model optimization submodule: Online incremental learning algorithm parameters are set as follows: learning rate Number of incremental learning samples per session The model parameters are updated every hour, and the loss function L is calculated using... Gradient descent method This ensures that the model can achieve adaptive optimization without interrupting monitoring.

[0102] The parameters and their ranges for the trajectory prediction algorithm are defined as follows to ensure the reproducibility of the calculation logic:

[0103] Coordinate system: The WGS-84 geodetic coordinate system is adopted, and the UTM projection is used for plane coordinate transformation. The projection zone is automatically selected according to the ship's navigation area.

[0104] Key parameters:

[0105] Time interval between two consecutive data acquisitions (Same as the satellite positioning module update frequency);

[0106] Environmental correction factor k: based on channel wind speed Water flow velocity Dynamic adjustment, the adjustment formula is as follows ,in =5m / s (base velocity), the value of k is strictly controlled within the range of... Between (when the calculation result is outside the range, the boundary value is taken);

[0107] Trajectory Anomaly Detection Threshold : Set according to the width of the waterway, inland waterway =5m, coastal waterway =10m, open sea area =15m.

[0108] Assumption The actual position coordinates of the ship at that moment Actual sailing speed (Converted to 2.778 m / s), actual heading angle (Due east direction), environmental correction factor , ,but: , ≈121.6543235° If the actual position coordinates of the ship at time t Then the deviation value (Approximately 0.72m), if it is an inland waterway ( =5m), then If the trajectory is determined to be normal; (Inland waterway) is then judged as an abnormal trajectory.

[0109] Algorithm for Determining Illegal Navigation Behavior

[0110] Overspeeding detection

[0111] Channel speed limit threshold Preset rules:

[0112] Inland waterway Class I: ;

[0113] Inland waterway Class II: ;

[0114] Inland waterway Class III: ;

[0115] Inland waterways of Class IV and below: ;

[0116] First-class coastal waterways: ;

[0117] Second-class coastal waterways: ;

[0118] Coastal waterways of Class III and below: ;

[0119] Open sea area: .

[0120] Judgment logic: The ship's actual sailing speed v exceeds the corresponding waterway for three consecutive data collection cycles (0.3s). At that time, it was determined to be speeding.

[0121] Insufficient safe distance determination

[0122] Formula parameter definition:

[0123] Overall length of the vessel L: obtained through the vessel's AIS data (if the AIS data is abnormal, the preset value is used: L=50m for inland waterway vessels, L=100m for coastal vessels, and L=200m for ocean-going vessels).

[0124] Maximum beam of the vessel B: obtained through the vessel's AIS data (preset values: B=8m for inland waterway vessels, B=15m for coastal vessels, and B=30m for ocean-going vessels).

[0125] Waterway grade coefficient β: Grade I waterway β=1.5, Grade II waterway β=1.3, Grade III and below waterway β=1.1;

[0126] Minimum safe distance The calculation result should be rounded to one decimal place.

[0127] Vessels navigating a certain coastal secondary waterway, with a total length of Maximum width Current sailing speed waterway grade coefficient ,but: If the actual distance between the vessel and the vessel in front If the distance is insufficient, it is considered that the safe distance is not enough.

[0128] Detailed steps of the system workflow

[0129] Step 1: System Initialization and Startup

[0130] After the system is powered on, the dual power supply module first starts the main power supply circuit, outputting 24VDC voltage to each module. The power management unit monitors the main power supply voltage in real time to confirm that the voltage is stable. Within the range.

[0131] The self-test and maintenance module starts the status monitoring submodule to perform initialization tests on the data acquisition module, processing and storage module, encryption and security module, satellite positioning module, and AI edge computing unit.

[0132] Check the power supply voltage of each module (within normal range) Operating current (the sum of the rated currents of each module) );

[0133] The sensor communication link of the detection data acquisition module (RS485 / CAN bus communication is normal, with no packet loss) is detected.

[0134] Check the key initialization status of the encryption security module (random key generation successful, preset key verification passed).

[0135] Test the signal strength of the satellite positioning module (both BeiDou and GPS signal strengths are ≥-100dBm);

[0136] Detecting the loading status of the AI ​​edge computing unit model (the weight file of the improved YOLOv7 model has been loaded and the inference engine has been successfully initialized).

[0137] If all test items are normal (no fault codes are generated), the self-test and maintenance module sends an "initialization complete" signal to the processing and storage module, and the system enters the monitoring mode. If there is a minor fault (such as a sensor signal attenuation of 2dB), the fault handling submodule performs automatic repair (restarts the sensor), and enters the monitoring mode after successful repair. If there is a serious fault (such as encryption chip initialization failure), a fault code (0x0301) is generated and pushed to the shore-based platform, the system pauses startup, and waits for manual intervention.

[0138] Step 2: Data Acquisition and Location

[0139] After receiving the "Monitoring Mode Activation" signal, the data acquisition module synchronously activates all sensors and general interfaces:

[0140] Hall effect speed sensor per The ship's speed is collected once, and the electronic compass simultaneously collects the heading angle;

[0141] The VHF communication data acquisition module captures communication data between the ship and shore-based and other ships in real time and stores it frame by frame;

[0142] The RS485 interface reads ship AIS data (including ship name, MMSI code, overall length L, maximum width B, etc.), and the Ethernet interface reads radar detection data (channel obstacle distribution information).

[0143] The satellite positioning module activates BeiDou / GPS dual-mode positioning and automatically selects a satellite (prioritizing signal strength). The satellite acquires positioning data (latitude, longitude, altitude, and positioning accuracy factor) every 0.1 seconds. And transmits the location data to the processing and storage module in real time.

[0144] The data collected by the data acquisition module and the satellite positioning module are both marked as "raw data" and temporarily stored in the module's built-in cache (capacity 16MB). When the cache is full, the oldest data is automatically overwritten.

[0145] Step 3: Data Encryption and Transmission

[0146] The encryption security module monitors the buffers of the data acquisition module and the satellite positioning module in real time. When the amount of data in the buffer reaches 1KB or the acquisition time reaches 1 second, the full-process encryption is initiated first if either condition is met.

[0147] The hardware encryption chip generates a random key (128 bits), uses the AES-256 algorithm to encrypt the original data, and uses the CBC (Cipher Block Chaining) encryption mode. The initialization vector (IV) is randomly generated (128 bits).

[0148] The encrypted ciphertext data (containing the original data, random key ID, IV, and timestamp) is transmitted to the processing and storage module via the SPI bus at a transmission rate of [missing information]. During transmission, CRC32 verification is used (the verification code is transmitted along with the ciphertext).

[0149] After the storage module receives the encrypted data, it first performs a CRC32 check to confirm that the data has not been lost or tampered with. If the check fails, it requests the encryption security module to retransmit the data. If the check passes, it sends a "ciphertext received successfully" signal to the encryption security module. The encryption security module then calls a preset key to decrypt the ciphertext and obtain the original data.

[0150] Step 4: Data Processing and Storage

[0151] Embedded processor for processing storage modules ( Process the decrypted raw data:

[0152] The LZ77 lossless compression algorithm is used to compress the data in real time. The compressed data is then given a unique timestamp (UTC time, accurate to milliseconds) and a data type identifier (such as "speed data", "location data", "communication data").

[0153] The compressed data is then standardized (to JSON format), and the resulting valid data is divided into two paths:

[0154] First path: The data is transmitted to the encryption security module, where it is encrypted again using a preset key (AES-256 algorithm, ECB mode) and then stored on the SSD solid-state drive (main storage) and TF card;

[0155] The second route: synchronous transmission to the AI ​​edge computing unit, using TCP / IP as the transmission protocol, with a transmission rate of [missing information]. To ensure data timeliness.

[0156] Step 5: AI Intelligent Analysis and Violation Alerts

[0157] The AI ​​edge computing unit receives the processed, standardized data and initiates the improved YOLOv7 model and trajectory prediction algorithm:

[0158] Object recognition: Performs real-time inference on waterway scene data (radar data + positioning data) to identify objects such as ships, buoys, reefs, and bridges within the waterway, outputting information such as object type, location coordinates, and size. Recognition latency is minimal. ;

[0159] Trajectory analysis: Based on the ship's actual position at time t-1 ,speed Heading angle Combined with the environmental correction factor k, the predicted position at time t is calculated. and the actual position at time t Compare and calculate the deviation value. ;

[0160] Violation determination: Determine whether there are any abnormal trajectories. Overspeeding Insufficient safe distance If any of the violation conditions are met, it will be judged as an illegal navigation behavior.

[0161] If the navigation behavior is determined to be illegal, the AI ​​edge computing unit will immediately trigger an audible and visual alarm:

[0162] The audible and visual alarm is activated (sound intensity ≥100dB, light flashing frequency 2Hz, color red).

[0163] Push violation information (including violation type, current location, and suggested measures, such as "speeding, current speed") to the ship's bridge display terminal. Speed ​​limits in waterways Please slow down).

[0164] Violation information is simultaneously transmitted to the processing storage module, encrypted and stored on an SSD solid-state drive, and retained for 90 days.

[0165] Step 6: Real-time self-test and fault handling

[0166] The self-test and maintenance module status monitoring submodule collects the operating status data of each module every 500ms and generates an operating status log. The log format is as follows: [2024-05-2014:31:00.500] Data acquisition module: voltage 24.2V, current 1.2A, transmission rate 5Mbps, running time 60s, status normal.

[0167] The fault handling submodule analyzes the operation status log in real time and handles faults according to the following rules:

[0168] Minor faults: If the data transmission rate drops to 2Mbps (below the normal threshold of 3Mbps), it is determined to be a loose interface, and an automatic "restart communication interface" operation is performed. After restarting, if the transmission rate recovers to [the specified value], [the system will resolve the issue]. Record the fault repair log; if the problem persists after 3 restarts, it is upgraded to a critical fault.

[0169] Serious fault: such as the satellite positioning module signal strength remaining below a certain level for 3 seconds. The problem persisted even after switching to dual-mode, generating fault code 0x0502 (satellite positioning module failure). This code was then pushed to the shore-based management platform via the satellite communication module, simultaneously triggering a yellow alarm (light flashing at a frequency of 1Hz) to remind the crew to check the antenna connection.

[0170] Step 7: Continuous Cyclic Monitoring

[0171] After the system completes steps 2-6, it will automatically return to step 2 without manual intervention and continue to execute the data collection, encryption, processing, AI analysis, and self-inspection and maintenance process, so as to achieve uninterrupted 24-hour monitoring of ship navigation.

[0172] If a crew member issues a "pause monitoring" command via the bridge control terminal (only allowed when the ship is docked), the system enters sleep mode, with only the self-test maintenance module and the dual power supply module remaining operational, while other modules are powered off. Steps 1-7 are repeated upon restarting.

Claims

1. A ship navigation monitoring and management system based on edge computing, characterized in that, It includes a data acquisition module, a processing and storage module, an encryption and security module, a dual power supply module, a satellite positioning module, a self-testing and maintenance module, and an AI edge computing unit; The data acquisition module integrates multiple types of sensors and a universal interface to synchronously acquire the ship's position, speed, heading, and communication data. The processing and storage module uses an embedded processor and a removable storage medium to achieve real-time compression, timestamp marking, and storage management of the collected data; The encryption security module uses a hardware encryption chip to achieve AES-256 encryption protection for the entire process of data transmission and storage within the system; The dual power supply module includes a main power supply circuit and a lithium battery backup power supply, which is used to provide stable power supply to each module of the system. The satellite positioning module integrates a Beidou / GPS dual-mode transmitter, which is used to acquire ship positioning data in real time and transmit it to the processing and storage module; The self-testing and maintenance module is used to monitor the operating status of each module of the system in real time and perform fault diagnosis. The AI ​​edge computing unit is equipped with a multi-object detection algorithm engine and a built-in deep neural network model, which is used to realize object recognition, ship trajectory analysis and alarm of illegal navigation behavior in waterway scenarios; The outputs of the data acquisition module and the satellite positioning module are both connected to the input of the processing and storage module. The processing and storage module communicates bidirectionally with the encryption security module and the AI ​​edge computing unit. The self-testing and maintenance module's detection end is connected to each module, and the output of the dual power supply module is connected to the power supply end of each module, forming a ship navigation monitoring and management architecture that works collaboratively on the edge side.

2. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The removable storage medium of the processing storage module is an SSD solid-state drive or a TF memory card. The embedded processor uses the LZ77 lossless compression algorithm to compress the collected data in real time. The compressed data carries a unique timestamp and is stored in an ordered manner according to the time sequence.

3. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The hardware encryption chip of the encryption security module is a high-performance encryption chip. Its full-process encryption process includes: after data acquisition, the hardware encryption chip generates a random key and encrypts the original data to obtain ciphertext through the AES-256 algorithm; after the ciphertext is transmitted to the processing and storage module, the hardware encryption chip calls the preset key to decrypt it; the processed valid data is encrypted again and stored in the storage medium to achieve full life cycle security protection of data.

4. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The dual power supply module also includes a power management unit, which collects the voltage signal of the main power supply circuit in real time. When the main power supply voltage is detected to be lower than a preset threshold, the power management unit will detect the voltage signal. At that time, the lithium battery backup power supply is triggered and a seamless switch is completed, with a switching delay. Furthermore, the lithium battery backup power supply can ensure continuous operation of the system for ≥4 hours.

5. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The deep neural network model of the AI ​​edge computing unit is an improved YOLOv7 model. The multi-object detection algorithm engine realizes ship trajectory analysis through trajectory prediction algorithm. The trajectory prediction algorithm formula is: In the formula, Let be the predicted position coordinates of the ship at time t. The coordinates of the ship's actual position at time t-1 The actual speed of the ship at time t-1 The actual heading angle of the ship at time t-1. The time interval between two consecutive data collections. This is an environmental correction factor, dynamically adjusted based on channel wind speed and current velocity, with a value ranging from 0.9 to 1.1; the actual coordinates of the ship at time t. Deviation value from predicted location coordinates >Preset threshold When this occurs, it is determined to be a trajectory anomaly, where the formula for calculating the deviation value is: 。 6. The ship navigation monitoring and management system based on edge computing according to claim 5, characterized in that, The aforementioned violations include abnormal navigation, speeding, and insufficient safe distance. The criterion for determining speeding is the actual speed of the vessel. , Based on the waterway classification plan, the inland waterway is classified as Class III. =12km / h, coastal Class II waterway =18km / h; Insufficient safe distance is determined by the following formula: In the formula, For the minimum safe distance, L is the overall length of the vessel, and B is the maximum width of the vessel. The ship's current speed. This refers to the waterway grade coefficient, specifically for Class I waterways. Class II waterway Class III and below waterways When the actual distance between the ship and other objects in the waterway If the distance is insufficient, it is determined to be a safe distance.

7. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The BeiDou / GPS dual-mode transmitter of the satellite positioning module supports automatic satellite selection and signal switching, and the positioning accuracy is high. The location data update frequency is 10Hz; when the signal strength of one of the location modes is lower than... When the time is right, it automatically switches to another positioning mode, and the response time changes accordingly. .

8. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The self-test and maintenance module includes a status monitoring submodule and a fault handling submodule. The status monitoring submodule collects the power supply voltage, operating current and data transmission rate of each module in real time and generates an operating status log. The fault handling submodule performs automatic repair for minor faults and generates fault codes for serious faults and sends them to the shore-based management platform through the satellite communication module.

9. The ship navigation monitoring and management system based on edge computing according to claim 1, characterized in that, The AI ​​edge computing unit also includes a model optimization submodule, which dynamically updates the parameters of the deep neural network model using an online incremental learning algorithm. The update formula is as follows: , In the formula, for The model parameters are updated over time. for The model parameters at time step 1, where η is the learning rate, ranging from 0.001 to 0.

01. For loss function exist gradient at, The number of samples for a single incremental learning iteration. Let i be the true label of the i-th sample. The probability of the model predicting the i-th sample is calculated by the deviation between the model prediction result and the actual waterway scene data.

10. A ship navigation monitoring and management method based on edge computing, characterized in that, The ship navigation monitoring and management system based on edge computing, as described in any one of claims 1-9, includes the following steps: Step 1: After the system starts up, the dual power supply module supplies power to each module, and the self-test and maintenance module performs initialization tests. After confirming that the status of each module is normal, it enters the monitoring mode. Step 2: The data acquisition module and the satellite positioning module start synchronously. The data acquisition module collects ship speed, heading, and communication data through multiple types of sensors and general interfaces, while the satellite positioning module acquires ship positioning data through a Beidou / GPS dual-mode transmitter. Step 3: The encryption security module performs AES-256 encryption on the collected raw data, and the encrypted ciphertext data is transmitted to the processing and storage module; Step 4: After the storage module decrypts the encrypted data, the embedded processor performs real-time compression and timestamp marking, then encrypts and stores the processed data, and synchronously transmits it to the AI ​​edge computing unit. Step 5: The AI ​​edge computing unit identifies objects in the waterway scene through a deep neural network model, analyzes the ship's trajectory using a trajectory prediction algorithm, and determines whether there is any illegal navigation behavior. If so, it immediately triggers an audible and visual alarm. Step 6: The self-test and maintenance module monitors the operating status of each module in real time, continuously generates operating logs, and performs corresponding repairs or alarm handling when a fault is detected; Step 7: The system continuously cycles through steps 2-6 to achieve real-time monitoring and management of ship navigation.

Images