Image recognition-based open-sea ship personnel danger warning method

Abstract

The invention relates to the technical field of image recognition and ship safety, and discloses an open-sea ship personnel danger warning method based on image recognition, which comprises the following steps: acquiring real-time image data of a sailor operation area; occlusion object detection and complementation are carried out on the image through deep learning; identifying static and dynamic operation states of the sailors by using the attitude estimation model, and performing operation risk assessment by combining real-time external environment data and utilizing a machine learning algorithm; compared with the prior art that traditional sailor danger warning methods mostly depend on single sensor data or manual monitoring, the problem that shelters in sailor operation are difficult to effectively deal with is solved, and particularly under the conditions of complex sea conditions, low illumination, equipment sheltering and the like, the sailor danger warning method has the advantage that the sailor danger warning method cannot effectively deal with the shelters. In order to solve the technical problem that the existing method cannot accurately identify and evaluate the operation risk of the sailors, the invention ensures that an alarm can be accurately sent in real time in a complex environment by combining image visual processing, and enhances the intelligence and safety protection capability of ship operation.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition and ship safety, and in particular relates to a method for warning personnel danger on a high-seas ship based on image recognition. Background Art

[0002] Currently, monitoring hazardous operations performed by crew members on offshore vessels primarily relies on manual inspections or simple sensor-based data monitoring. While these methods can provide warnings to a certain extent, they still have many shortcomings. For example, existing sensor-based data monitoring methods only provide limited information and cannot accurately determine the crew's dynamic operational status, such as fatigue, improper posture, or operational risks caused by equipment obstruction. Furthermore, traditional methods often rely on a single data source (such as meteorological data or simple camera monitoring) and lack comprehensive analysis of crew movements and environmental changes. Existing technologies cannot fully meet the needs of real-time monitoring and efficient early warning of crew operational safety, especially in harsh sea conditions, low light, or equipment obstruction. Therefore, there is an urgent need for a crew hazard warning method based on image recognition and deep learning technologies. This method can accurately identify the crew's operational status in complex environments such as complex sea conditions, low light, and equipment obstruction, acquire and analyze operational environment data in real time, and provide precise hazard warning information. This approach can not only monitor crew posture and movements in real time, but also dynamically assess the potential impact of environmental factors on crew operations, thereby improving the safety and intelligence of ship operations. Summary of the Invention

[0003] In response to the above-mentioned technical deficiencies, the purpose of the present invention is to propose a method for warning personnel danger on offshore vessels based on image recognition, aiming to solve the technical problem that traditional crew danger warning methods in the existing technology mostly rely on single sensor data or manual monitoring, and are difficult to effectively deal with the problem of obstructions in crew operations, especially under complex sea conditions, low light and equipment obstruction. The existing methods are unable to accurately identify and assess the crew's operational risks.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a method for warning personnel danger on offshore vessels based on image recognition.

[0005] The method for warning personnel danger on a high-sea vessel based on image recognition comprises:

[0006] Step S10: acquiring real-time image data of the crew's operating area based on a camera on the ship, removing noise from the real-time image data and enhancing the image brightness to obtain a preliminary clear image;

[0007] Step S20: performing occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image;

[0008] Step S30: introducing a posture estimation model, applying the posture estimation model based on the completed clear image to identify the static working state and dynamic working state of the crew member, and generating the current working state according to the static working state and the dynamic working state;

[0009] Step S40: Obtain current external environment data, apply a random forest algorithm to perform a comprehensive assessment based on the current operation status and the current external environment data, and output a crew operation risk value. When the crew operation risk value exceeds a preset crew operation risk value threshold, a danger warning message is generated;

[0010] Step S50: Pushing danger warning information to the monitoring center in real time, and providing corresponding safety handling suggestions based on the danger warning information.

[0011] Preferably, in step S10, the steps of acquiring real-time image data of the crew operating area based on a camera on the ship, removing noise from the real-time image data and enhancing the image brightness to obtain a preliminary clear image specifically include:

[0012] Step S101: obtaining real-time image data of the crew's operating area based on a camera on the ship to extract real-time image noise caused by sea surface fluctuations and hull shaking;

[0013] Step S102: applying bilateral filtering to the real-time image noise to obtain a denoised real-time image;

[0014] Step S103: Adaptive histogram equalization is used to perform brightness enhancement and contrast enhancement processing on the denoised real-time image to obtain a preliminary clear image.

[0015] Preferably, in step S20, the step of performing occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image specifically includes:

[0016] Step S201: Detect occluders based on the Mask R-CNN neural network, identify occluded areas, including crew members' arms, tools, and ship equipment, and generate occluded calibration areas for the occluders;

[0017] Step S202: Constructing a generative adversarial network (cGAN), including:

[0018] Build a generator in a generative adversarial network to generate missing image parts based on information about the area surrounding the occluded calibration area;

[0019] Construct a discriminator in a generative adversarial network to evaluate the naturalness of the generated missing image parts;

[0020] Step S203: Obtain historical sample data, construct a loss function L, train a generative adversarial network (cGAN) based on the historical sample data and the loss function L, input the preliminary clear image into the trained generative adversarial network (cGAN), and output the completed real-time image;

[0021] Step S204: Calculate the structural similarity index SSIM between the preliminary clear image and the supplemented real-time image, and preset a structural similarity index threshold. If the structural similarity index SSIM is greater than the preset structural similarity index threshold, proceed to step S205; otherwise, return to step S203;

[0022] Step S205: performing image pyramid fusion on the preliminary clear image and the completed real-time image to obtain a completed clear image.

[0023] Preferably, in step S20, the calculation formula of the structural similarity index SSIM is:

[0024]

[0025] Among them, μ x and μ y is the mean pixel value of the initial clear image and the completed real-time image, and is the variance of the initial clear image and the completed real-time image, σ xy is the covariance of the initial clear image and the completed real-time image, and C1 and C2 are the structural similarity index adjustment coefficients.

[0026] Preferably, in step S30, a posture estimation model is introduced, and the static working state and dynamic working state of the crew member are identified based on the supplemented clear image by applying the posture estimation model, and the step of generating the current working state according to the static working state and the dynamic working state specifically includes:

[0027] Step S301: The completed clear image is used as input data and input into a preset pose estimation model OpenPose. The pose estimation model outputs body point data in the crew member image, including head point data, shoulder point data, elbow point data, and knee point data.

[0028] Step S302: constructing a skeleton model of the crew member based on the body point data, and calculating the body joint angles using the triangle rule or vector operation, and inferring the crew member's static working state based on the body joint angles;

[0029] Step S303: By tracking and calculating the joint angles and body point motion trajectories, body motion data is generated, including the body's movement direction and limb activity frequency. Based on the body motion data, the crew's dynamic working status is analyzed;

[0030] Step S304: Generate the current job status according to the static job status and the dynamic job status.

[0031] Preferably, in step S40, the current external environment data includes meteorological data, sea condition data and real-time equipment status data of the ship around the ship; the meteorological data around the ship includes wind speed, air pressure, temperature and humidity; the sea condition data includes wave height and current speed; the real-time equipment status data of the ship includes engine temperature, fuel consumption and power status.

[0032] Preferably, in step S40, the step of applying the random forest algorithm to perform a comprehensive assessment based on the current operating status and the current external environment data, and outputting the crew operation risk value, specifically includes: performing standardization processing on the current operating status and the current external environment data to obtain standardized operating status data and standardized environment data; performing feature extraction on the standardized operating status data and the standardized environment data, and constructing an input feature vector; and applying the random forest algorithm for training to output the crew operation risk value.

[0033] The present invention also provides a high-seas vessel personnel danger warning system based on image recognition, comprising:

[0034] The image acquisition and preprocessing module is used to obtain real-time image data of the crew's working area based on the camera on the ship, remove noise from the real-time image data and enhance the image brightness to obtain a preliminary clear image;

[0035] The occlusion detection and image completion module is used to perform occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image;

[0036] The posture estimation and operation status recognition module is used to introduce the posture estimation model, apply the posture estimation model based on the completed clear image to identify the static operation status and dynamic operation status of the crew member, and generate the current operation status based on the static operation status and dynamic operation status;

[0037] The risk assessment and alarm module is used to obtain the current external environment data, apply the random forest algorithm to conduct a comprehensive assessment based on the current operation status and the current external environment data, and output the crew operation risk value. When the crew operation risk value exceeds the preset crew operation risk value threshold, a danger alarm message is generated;

[0038] The alarm push and processing suggestion module is used to push danger alarm information to the monitoring center in real time and provide corresponding safety processing suggestions based on the danger alarm information.

[0039] The present invention also provides a computer program product, including a distant sea vessel personnel danger warning program based on image recognition, which implements the distant sea vessel personnel danger warning method based on image recognition when the distant sea vessel personnel danger warning program based on image recognition is executed by a processor.

[0040] The beneficial effects of the present invention are that: by combining image visual recognition, posture estimation and deep learning technology, the present invention can identify the crew's operating status in real time, including static and dynamic operating status, promptly discover the crew's dangerous operating behaviors (such as improper posture, fatigue or high-risk actions), and combine environmental data to conduct accurate operating risk assessment, thereby significantly improving the safety of crew operations.

[0041] The present invention solves the problem of occlusion caused by factors such as equipment, environment or sea conditions during crew operations through image completion technology, ensuring that the crew's operating status can still be accurately identified in harsh sea conditions, low light or equipment obstruction, thereby avoiding the technical problem that traditional monitoring methods cannot achieve accurate identification and timely alarm in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 The figure is a flow chart of a first embodiment of a method for warning personnel danger on a high-sea vessel based on image recognition according to the present invention.

[0044] Figure 2 The present invention is a schematic diagram of a device for a method for warning personnel of danger on a high-sea vessel based on image recognition. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] Example 1: Figure 1 1 is a flow chart of a first embodiment of a method for warning personnel of a high-sea vessel in danger based on image recognition according to the present invention, which provides a first embodiment of a method for warning personnel of a high-sea vessel in danger based on image recognition according to the present invention.

[0047] In a first embodiment, the method for warning personnel of danger on a high-sea vessel based on image recognition includes:

[0048] Step S10: acquiring real-time image data of the crew's operating area based on a camera on the ship, removing noise from the real-time image data and enhancing the image brightness to obtain a preliminary clear image;

[0049] It's important to note that after capturing real-time image data of the crew's work area via the ship's cameras, the image data is first subjected to denoising and brightness enhancement to ensure image quality suitable for subsequent processing. This step aims to remove noise and enhance image detail, making the crew's work status clearer and providing high-quality image data for subsequent image completion, pose estimation, and other processes.

[0050] It's understandable that image denoising and brightness enhancement not only improve image quality but also mitigate the impact of external environmental factors (such as sea surface reflections, ship sway, and insufficient lighting). By using adaptive filtering techniques (such as median or bilateral filtering) for denoising, combined with adaptive histogram equalization and other technologies, image quality can be maintained even in dynamic and complex environments, ensuring accurate capture of the crew's postures and movements.

[0051] For example, by using adaptive filtering technology to process real-time images of the crew's operating area, high-frequency noise in the image can be significantly reduced, and the crew's visual features can be enhanced in low-light environments. In experiments, applying this denoising and brightness enhancement process improved the image's signal-to-noise ratio (SNR) by over 30%, increasing the accuracy of the subsequent pose estimation model by approximately 15%, further verifying the positive impact of improved image quality on the accuracy of crew operating status recognition.

[0052] Step S20: performing occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image;

[0053] It should be noted that during crew operations, some body parts may be obscured by ship equipment, tools, or other objects. To ensure accurate identification of crew operating status, this step uses deep learning methods (such as Mask R-CNN or YOLO) to detect occluded areas in the image and uses generative adversarial networks (GANs) or conditional generative adversarial networks (cGANs) to complete these occluded areas.

[0054] It's understandable that occlusion detection and image completion are automated processes. After using a deep learning model to identify occluded areas in an image, a generative adversarial network intelligently fills these areas based on the surrounding environment and background information. This means that even if an image is occluded or partially missing, a natural and consistent completed image can still be generated, avoiding the inaccuracies and inefficiencies of traditional manual image restoration methods.

[0055] It should be understood that the deep learning-based occlusion detection and image completion method can efficiently process various complex environments (such as those with obstructions in the crew work area and uneven lighting). This differs from traditional methods that rely on simple image interpolation and threshold setting. The present invention automatically identifies and repairs occluded areas by training a deep neural network, improving robustness and adaptability, and significantly improving completion accuracy in dynamic scenes.

[0056] Step S30: introducing a posture estimation model, applying the posture estimation model based on the completed clear image to identify the static working state and dynamic working state of the crew member, and generating the current working state according to the static working state and the dynamic working state;

[0057] It's important to note that the posture estimation model accurately identifies the crew's movements and postures by analyzing the crew's skeletal structure in the completed image. Static operating states typically involve the crew maintaining a certain posture while operating or standing, while dynamic operating states involve frequent movement or actions (such as carrying, bending, and turning). This technology not only helps distinguish between static and dynamic states but also identifies dangerous operating behaviors by assessing the degree of posture abnormality, such as fatigue or excessive bending.

[0058] It should be understood that by introducing advanced posture estimation models, every detail of the crew's movements in complex operating environments can be captured in real time. Regardless of conditions such as dim light, choppy seas, or equipment obstruction, the crew's posture and movements can still be effectively detected and analyzed. Compared with traditional rule-based operating status recognition methods, this method has higher accuracy and adaptability, especially when crew members are performing complex tasks.

[0059] For example, in one set of experiments, when a crew member performed dynamic movements such as carrying objects, the posture estimation model accurately identified the movement trajectories of their arms and legs and demarcated them as dynamic work states. If a crew member excessively bent over during the process, this dangerous posture could be detected in real time and, combined with environmental data, assessed for potential high-risk operational behaviors. In this case, a timely danger alert could be generated, prompting the crew member and the monitoring center to take necessary safety measures. These test results demonstrate that the posture estimation model can improve the accuracy of identifying crew member work states and the efficiency of early warnings.

[0060] Step S40: Obtain current external environment data, apply a random forest algorithm to perform a comprehensive assessment based on the current operation status and the current external environment data, and output a crew operation risk value. When the crew operation risk value exceeds a preset crew operation risk value threshold, a danger warning message is generated;

[0061] It's important to note that external environmental data includes information such as wind speed, temperature, wave height, and the status of ship equipment, all of which directly impact the safety of crew operations. Furthermore, operational status data reflects crew behavior patterns, such as whether they are bent over or working in the same position for extended periods. The random forest algorithm, combined with this data, can conduct a multi-dimensional assessment of the risks faced by crew members during operations, providing a more accurate and comprehensive operational risk assessment.

[0062] It should be understood that the application of the random forest algorithm enables a real-time, comprehensive assessment based on multiple environmental factors and operational status, automatically deriving the crew's current operational risk value based on historical patterns in the training data. Compared to traditional single-data-source risk assessments, this approach considers more complex influencing factors, resulting in greater accuracy and adaptability, especially when faced with dynamically changing sea conditions, weather, or crew operational status.

[0063] For example, suppose it is obtained that the crew is currently in a dynamic operation state (for example, the crew is carrying items), and the external environmental data shows that the wave height is high and the wind speed is beyond the normal range. The random forest model will combine these data to calculate an operation risk value. If the risk value exceeds the preset threshold, a danger warning message will be generated to remind the crew that there is a high safety risk in the current operation, and it is recommended to suspend the operation or take appropriate safety protection measures. In the experiment, based on the comprehensive evaluation of multi-source data, the random forest algorithm can accurately identify higher-risk operation states, which improves the accuracy of the early warning compared to the traditional single threshold judgment method.

[0064] Step S50: Pushing danger warning information to the monitoring center in real time, and providing corresponding safety handling suggestions based on the danger warning information.

[0065] It's important to note that safety action recommendations are generated based on intelligent analysis of warning information, taking into account multiple factors, including the crew's operational status, the external environment, and equipment conditions. If a crew member's operation is at a high risk (e.g., improper posture, severe sea conditions), specific action recommendations are automatically provided to monitoring personnel, such as suspending operations or notifying the crew to take protective measures. This differs from traditional approaches that rely solely on manual intervention and provides more intelligent, real-time safety warnings and action plans.

[0066] For example, suppose the system detects that a crew member is carrying items in high sea conditions, and their posture indicates that they are excessively bent over, and external environmental data indicates strong winds and high waves. At this point, the system automatically generates a high-risk alarm and pushes the alarm information to the monitoring center through the communication system, while providing handling suggestions. For example, it recommends suspending the current operation, notifying the crew to adjust their working posture and take safety measures, or relocating the crew's working area to a safer location. In experimental tests, this real-time alarm and handling suggestion push method enabled the monitoring center to make effective intervention within 5 minutes.

[0067] Embodiment 2: In addition, the present invention provides an image recognition-based danger warning system for seagoing vessels, which utilizes the image recognition-based danger warning method for seagoing vessels described in the above embodiment to solve the technical problem of image recognition-based danger warning for seagoing vessels. Compared with the prior art, the image recognition-based danger warning system for seagoing vessels provided by the present invention has the same beneficial effects as the image recognition-based danger warning method for seagoing vessels provided in the above embodiment. Other technical features of the image recognition-based danger warning system for seagoing vessels are the same as those disclosed in the above embodiment and are not further elaborated here.

[0068] Example 3: The present invention provides a seagoing vessel personnel danger warning device based on image recognition, please refer to Figure 2A device for warning personnel of danger to a distant sea vessel based on image recognition includes: at least one processor; and a memory in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for warning personnel of danger to a distant sea vessel based on image recognition in the first embodiment described above. The device for warning personnel of danger to a distant sea vessel based on image recognition in the embodiment of the present invention may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (e.g., vehicle-mounted navigation terminals), and fixed terminals such as digital TVs and desktop computers. The device for warning personnel of danger to a distant sea vessel based on image recognition is merely an example and should not limit the functions and scope of use of the embodiment of the present invention. A device for warning personnel of danger to distant sea vessels based on image recognition may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the device for warning personnel of danger to distant sea vessels based on image recognition. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input device 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage device 1003 including, for example, a magnetic tape, hard disk, etc.; and communication device 1009. Communication device 1009 can allow an image recognition-based offshore vessel personnel danger warning device to communicate wirelessly or wired with other devices to exchange data. Although the figure shows an image recognition-based offshore vessel personnel danger warning device with various systems, it should be understood that implementation or presence of all the illustrated systems is not required. More or fewer systems may alternatively be implemented or present.

[0069] Example 4: The present invention also provides a computer program product, comprising a computer program. When executed by a processor, the computer program implements the steps of the above-described method for warning personnel of danger on a high-seas vessel based on image recognition. The computer program product provided by the present invention can solve the technical problem of warning personnel of danger on a high-seas vessel based on image recognition. Compared with the prior art, the beneficial effects of the computer program product provided by the present invention are the same as those of the method for warning personnel of danger on a high-seas vessel based on image recognition provided by the above-described embodiment, and are not further elaborated here.

[0070] In particular, according to the embodiments disclosed in the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present invention are performed.

[0071] It should be understood that the various parts disclosed in the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any appropriate manner in any one or more embodiments or examples.

[0072] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for warning personnel danger on a high-sea vessel based on image recognition, characterized in that: Methods include: Step S10: acquiring real-time image data of the crew's operating area based on a camera on the ship, removing noise from the real-time image data and enhancing the image brightness to obtain a preliminary clear image; Step S20: performing occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image; Step S30: introducing a posture estimation model, applying the posture estimation model based on the completed clear image to identify the static working state and dynamic working state of the crew member, and generating the current working state according to the static working state and the dynamic working state; Step S40: Obtain current external environment data, apply a random forest algorithm to perform a comprehensive assessment based on the current operation status and the current external environment data, and output a crew operation risk value. When the crew operation risk value exceeds a preset crew operation risk value threshold, a danger warning message is generated; Step S50: Pushing danger warning information to the monitoring center in real time, and providing corresponding safety handling suggestions based on the danger warning information.

2. The method for warning personnel danger on a high-sea vessel based on image recognition according to claim 1, characterized in that: In step S10, the steps of obtaining real-time image data of the crew operating area based on the camera on the ship, removing noise in the real-time image data and enhancing the image brightness to obtain a preliminary clear image specifically include: Step S101: obtaining real-time image data of the crew's operating area based on a camera on the ship to extract real-time image noise caused by sea surface fluctuations and hull shaking; Step S102: applying bilateral filtering to the real-time image noise to obtain a denoised real-time image; Step S103: Adaptive histogram equalization is used to perform brightness enhancement and contrast enhancement processing on the denoised real-time image to obtain a preliminary clear image.

3. The method for warning personnel danger on a high-sea vessel based on image recognition according to claim 1, characterized in that: In step S20, the steps of performing occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image specifically include: Step S201: Detect occluders based on the Mask R-CNN neural network, identify occluded areas, including crew members' arms, tools, and ship equipment, and generate occluded calibration areas for the occluders; Step S202: Constructing a generative adversarial network (cGAN), including: Build a generator in a generative adversarial network to generate missing image parts based on information about the area surrounding the occluded calibration area; Construct a discriminator in a generative adversarial network to evaluate the naturalness of the generated missing image parts; Step S203: Obtain historical sample data, construct a loss function L, train a generative adversarial network (cGAN) based on the historical sample data and the loss function L, input the preliminary clear image into the trained generative adversarial network (cGAN), and output the completed real-time image; Step S204: Calculate the structural similarity index SSIM between the preliminary clear image and the supplemented real-time image, and preset a structural similarity index threshold. If the structural similarity index SSIM is greater than the preset structural similarity index threshold, proceed to step S205; otherwise, return to step S203; Step S205: performing image pyramid fusion on the preliminary clear image and the completed real-time image to obtain a completed clear image.

4. The method for warning personnel danger on a high-sea vessel based on image recognition according to claim 3, characterized in that: In step S20, the calculation formula of the structural similarity index SSIM is: Among them, μ x and μ y is the mean pixel value of the initial clear image and the completed real-time image, and is the variance of the initial clear image and the completed real-time image, σ xy is the covariance of the initial clear image and the completed real-time image, and C1 and C2 are the structural similarity index adjustment coefficients.

5. The method for warning personnel danger on a high-sea vessel based on image recognition according to claim 1, characterized in that: In step S30, a posture estimation model is introduced, and the static working state and dynamic working state of the crew member are identified based on the completed clear image using the posture estimation model. The steps of generating the current working state according to the static working state and the dynamic working state specifically include: Step S301: The completed clear image is used as input data and input into a preset pose estimation model OpenPose. The pose estimation model outputs body point data in the crew member image, including head point data, shoulder point data, elbow point data, and knee point data. Step S302: constructing a skeleton model of the crew member based on the body point data, and calculating the body joint angles using the triangle rule or vector operation, and inferring the crew member's static working state based on the body joint angles; Step S303: By tracking and calculating the joint angles and body point motion trajectories, body motion data is generated, including the body's movement direction and limb activity frequency. Based on the body motion data, the crew's dynamic working status is analyzed; Step S304: Generate the current job status according to the static job status and the dynamic job status.

6. The method for warning personnel danger on a high-sea vessel based on image recognition according to claim 1, characterized in that: In step S40, the current external environment data includes meteorological data, sea condition data and real-time equipment status data of the ship around the ship; the meteorological data around the ship includes wind speed, air pressure, temperature and humidity; the sea condition data includes wave height and current speed; the real-time equipment status data of the ship includes engine temperature, fuel consumption and power status.

7. The method for warning personnel danger on a high-sea vessel based on image recognition according to claim 1, characterized in that: In step S40, a random forest algorithm is applied to perform a comprehensive evaluation based on the current operation status and the current external environment data, and the step of outputting the crew operation risk value specifically includes: standardizing the current operation status and the current external environment data to obtain standardized operation status data and standardized environment data; extracting features from the standardized operation status data and the standardized environment data, and constructing an input feature vector; and applying the random forest algorithm for training to output the crew operation risk value.

8. A high-seas vessel personnel danger warning system based on image recognition, applied to a high-seas vessel personnel danger warning method based on image recognition according to any one of claims 1 to 7, characterized in that: The image recognition-based high-seas vessel personnel danger warning system includes: The image acquisition and preprocessing module is used to obtain real-time image data of the crew's working area based on the camera on the ship, remove noise from the real-time image data and enhance the image brightness to obtain a preliminary clear image; The occlusion detection and image completion module is used to perform occlusion detection and image completion processing on the preliminary clear image to obtain a completed clear image; The posture estimation and operation status recognition module is used to introduce the posture estimation model, apply the posture estimation model based on the completed clear image to identify the static operation status and dynamic operation status of the crew member, and generate the current operation status based on the static operation status and dynamic operation status; The risk assessment and alarm module is used to obtain the current external environment data, apply the random forest algorithm to conduct a comprehensive assessment based on the current operation status and the current external environment data, and output the crew operation risk value. When the crew operation risk value exceeds the preset crew operation risk value threshold, a danger alarm message is generated; The alarm push and processing suggestion module is used to push danger alarm information to the monitoring center in real time and provide corresponding safety processing suggestions based on the danger alarm information.

9. A seagoing vessel personnel danger warning device based on image recognition, characterized in that: The image recognition-based deep-sea vessel personnel danger warning device includes: a memory, a processor, and an image recognition-based deep-sea vessel personnel danger warning program stored in the memory and executable on the processor. When the image recognition-based deep-sea vessel personnel danger warning program is executed by the processor, it implements a deep-sea vessel personnel danger warning method based on image recognition according to any one of claims 1 to 7.

10. A computer program product, characterized in that The computer program product includes a distant sea vessel personnel danger warning program based on image recognition, and when the distant sea vessel personnel danger warning program based on image recognition is executed by a processor, it implements a distant sea vessel personnel danger warning method based on image recognition as described in any one of claims 1 to 7.

Images