A cruise ship oil pipeline oil spill monitoring system and method based on machine learning

Abstract

The present invention belongs to the technical field of oil pipeline monitoring, and discloses a cruise ship oil pipeline oil spill monitoring system and method based on machine learning, comprising a data acquisition module, used for acquiring cruise ship oil pipeline status monitoring data; a data processing module, used for preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain an oil pipeline status comprehensive feature data set; an oil spill risk prediction module, used for training and acquiring an oil spill risk assessment model according to the oil pipeline status comprehensive feature data set, and predicting an oil spill risk rate of the oil pipeline based on the oil spill risk assessment model; judging whether the cruise ship oil pipeline has an oil spill risk according to the oil spill risk rate; an oil spill warning module, if the cruise ship oil pipeline has an oil spill risk, performing a graded warning on the oil spill risk and triggering a corresponding alarm instruction; an oil spill source tracking module, used for acquiring the spatial position data of the oil pipeline, and locating the oil spill position in combination with the oil pipeline video monitoring data; and realizing comprehensive monitoring and efficient management of the cruise ship oil pipeline status.

Description

Technical Field

[0001] The present invention relates to the technical field of oil pipeline monitoring, and more specifically, to a cruise ship oil pipeline oil spill monitoring system and method based on machine learning. Background Art

[0002] The patent with patent announcement number CN117944837A discloses an unmanned ship oil spill monitoring system and method, including an information acquisition module for collecting parameter data of various sensors, an oil storage module for storing and transporting the ship's fuel, the oil storage module includes an oil tank and an oil pipe, the oil tank and the oil pipe are connected to the information acquisition module through a liquid level sensor and a flow sensor to monitor the storage and flow of the fuel in real time; a ventilation module for controlling the gas pressure and gas flow in the oil tank, the ventilation module is connected to the information acquisition module through a pressure sensor to monitor the gas pressure changes in the oil tank in real time; a communication module for data communication with a data processing center; the data processing center is used to receive the data collected by the information acquisition module and the signal transmitted by the communication module, and perform real-time fault analysis and oil spill monitoring of the oil storage module and the ventilation module through the onboard algorithm program; an alarm module is used to send an alarm signal when an oil spill is detected.

[0003] The existing cruise ship oil pipeline oil spill monitoring system and method mainly have the following problems:

[0004] Without a mechanism to dynamically adjust the overlap rate, the data in the time window may be insufficient or contain redundant data, affecting the quality and accuracy of feature extraction, and may cause the algorithm to have a reduced ability to identify changes in the state of the oil pipeline, and fail to accurately capture subtle changes in the oil pipeline; if the overlap rate is not flexibly adjusted according to the characteristics of the data, it may result in insufficient information in each window and fail to capture key change trends; or conversely, too much redundant data may affect subsequent analysis and modeling, increase the amount of calculation and time cost, and reduce system efficiency; if the advantages of time-frequency domain analysis are not fully utilized, it may not be possible to effectively extract important features such as frequency components, amplitude and phase from the oil pipeline parameter data, resulting in low accuracy of feature extraction, making it difficult to accurately describe the dynamic changes of the oil pipeline and unable to reflect the potential periodic characteristics or mutations of the oil pipeline; without effective time-frequency analysis, it may be difficult to identify abnormalities or signs of failure of the oil pipeline at different times, which will reduce the early warning capability of the monitoring system and fail to detect potential problems of the oil pipeline in a timely manner, thereby affecting the maintenance and safety management of the equipment; without precise analysis of the time-frequency spectrum, noise may not be effectively removed, thereby affecting the accuracy of subsequent analysis and prediction. The interference of noise will reduce the effect of the model, resulting in inaccurate predictions and inability to make effective decisions;

[0005] Without precise threshold segmentation or adaptive threshold adjustment, the separation of the oil film area and the background may not be accurate enough, which will lead to incorrect identification of the oil film area, and the background area may be misjudged as the oil film area, or the oil film area may be missed, thus affecting the accuracy of subsequent analysis and oil spill location positioning; without introducing the adaptive threshold adjustment formula, changes in image quality will cause the segmentation threshold to be inapplicable, resulting in poor segmentation effect; especially under different environmental conditions, the quality of the image may change, and the existing fixed threshold method is difficult to cope with these changes, reducing the accuracy and reliability of oil film area separation; the diffusion coefficient is not dynamically adjusted through the diffusion coefficient adjustment formula, and the influence of environmental factors such as temperature and pressure is ignored, and the modeling of oil film diffusion may be inaccurate; the model may not be able to truly reflect the actual oil film diffusion situation, resulting in distortion of oil film diffusion prediction; without accurately calculating the diffusion speed and direction of the oil film, it may not be possible to track the propagation trend of the oil film in real time, which will make it difficult to predict the changes in the oil spill location and unable to take timely response measures, resulting in delays in the handling of the oil spill incident; without comprehensive analysis of the spatial distribution, diffusion speed, direction and other characteristics of the oil film, it may not be possible to fully analyze the diffusion trend of the oil film, affecting the precise positioning of the oil spill location.

[0006] In view of this, the present invention proposes a cruise ship oil pipeline oil spill monitoring system and method based on machine learning to solve the above problems. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a cruise ship oil pipeline oil spill monitoring system based on machine learning, comprising:

[0008] Data acquisition module, used to obtain cruise ship oil pipeline status monitoring data;

[0009] A data processing module is used to pre-process the acquired cruise ship oil pipeline status monitoring data to obtain a comprehensive feature data set of the oil pipeline status;

[0010] The oil spill risk prediction module is used to train an oil spill risk assessment model based on the comprehensive feature data set of the oil pipeline status, and predict the oil spill risk rate of the oil pipeline based on the oil spill risk assessment model; based on the oil spill risk rate of the oil pipeline, it is determined whether there is an oil spill risk in the cruise ship oil pipeline;

[0011] Oil spill warning module: if there is an oil spill risk in the cruise ship's oil pipeline, it will issue a graded warning of the oil spill risk and trigger the corresponding alarm command;

[0012] The oil spill source tracking module is used to obtain the spatial location data of the oil pipeline, locate the oil spill location in combination with the oil pipeline video monitoring data, and upload the oil spill location to the oil spill intelligent monitoring terminal to remind relevant personnel to take countermeasures;

[0013] The modules are connected to each other via wired and / or wireless means to achieve data transmission between modules.

[0014] Further, the cruise ship oil pipeline status monitoring data includes oil pipeline parameter data, cruise ship status data, environmental monitoring data and oil pipeline video monitoring data;

[0015] The oil pipeline parameter data includes oil pipeline pressure data, oil pipeline flow data, oil pipeline temperature data, oil pipeline vibration data and oil pipeline wear data; the cruise ship status data includes cruise ship speed data, cruise ship heading data, cruise ship position data and cruise ship load data; the environmental monitoring data includes temperature data, humidity data, water pollution data, wind speed data and oil and gas concentration change data; the oil pipeline video monitoring data includes oil pipeline external video monitoring data and oil pipeline internal video monitoring data.

[0016] Furthermore, the method of preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain an oil pipeline status comprehensive feature data set includes:

[0017] The improved local Fourier transform is used to extract the features of the oil pipeline parameter data to obtain the oil pipeline parameter feature data set; the K-Means clustering algorithm is used to identify and eliminate the abnormal values ​​in the cruise ship status data and environmental monitoring data included in the cruise ship oil pipeline status monitoring data to obtain the cruise ship status feature data set and the environmental monitoring feature data set; the feature extraction of the oil pipeline video monitoring data is performed to obtain the oil film diffusion image feature data set;

[0018] The oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset are normalized by standard deviation and converted into standard normal distribution with mean 0 and standard deviation 1 to obtain the normalized oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset; the normalized oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset are weighted fused to obtain the comprehensive feature dataset of oil pipeline status.

[0019] Furthermore, the method of extracting features from the oil pipe parameter data using the improved local Fourier transform to obtain the oil pipe parameter feature data set includes: obtaining Continuous tubing parameter data over time ;in, are equally spaced timestamps; is the timestamp; is the total number of timestamps; the preset overlapping time window length is , defining the overlap rate of overlapping time windows as , the tubing parameter data Divide into overlapping time windows, the starting time interval of each time window is ;in, is the starting time interval of each time window; the overlap rate of overlapping time windows is adjusted by the overlap adjustment formula Perform dynamic design adjustments, and the overlap adjustment formula is: ;in, is the dimension of the tubing parameter data; is the length of the overlapping time window; To control the dynamic range of the overlap rate, the STFT is applied to perform a local Fourier transform on the tubing parameter data in each overlapping time window, thereby obtaining the spectrum information corresponding to each overlapping time window. ;in, is the complex number representation of the tubing parameter data in the time-frequency domain; is the window function; is the offset within the overlapping time window; is the complex exponential function; The frequency is the frequency, and the spectrum information corresponding to each overlapping time window is spliced ​​and combined into an STFT time spectrum; the frequency component, amplitude and phase of the oil pipeline parameter data are extracted by analyzing the STFT time spectrum, and then the oil pipeline parameter feature data set is obtained. Furthermore, the method for extracting features from the oil pipeline video monitoring data to obtain the oil film diffusion image feature data set includes:

[0020] S51, use the library function in the video processing library OpenCV to read the images in the YouTube video monitoring data frame by frame; it is assumed that the YouTube video monitoring data has a total of Frame image, preset Frame image in spatial coordinates The pixel value at ;in, and are the spatial horizontal and vertical coordinates of the image respectively; The index of the image frame, ranging from 1 to ;

[0021] S52, preset a segmentation threshold, and segment the image using a threshold segmentation function to separate the oil film area from the background area. The threshold segmentation function is: ;in, Indicates the oil film area; Represents the background area; is the segmentation threshold; the segmentation threshold is adjusted by the adaptive threshold adjustment formula Dynamic adjustment is performed, and the adaptive threshold adjustment formula is: ;in, is the segmentation threshold after dynamic adjustment; is the total number of pixels of the image; is the total number of frames of the image; To control the influence of the total number of pixels of the image on the segmentation threshold; To control the influence of the total number of frames of the image on the segmentation threshold;

[0022] S53, further extract the edge of the oil film area through Canny edge detection ,in, For the The result of Canny edge detection for each pixel in the frame image; the segmentation effect is improved through morphological operations, noise is removed and the oil film area is connected to obtain the oil film area image ;

[0023] S54, the oil film diffusion process in the preset oil film area follows the diffusion equation, which is: ;in, For Time, space coordinates Oil film concentration at is the diffusion coefficient; is the Laplace operator, which indicates the spatial variation of oil film concentration; is the index of the moment;

[0024] S55, Diffusion coefficient adjustment formula Dynamic adjustment is performed, and the diffusion coefficient adjustment formula is: ;in, is the diffusion coefficient after dynamic adjustment; is the influence factor of temperature on the diffusion coefficient; is the current temperature of the oil pipe; is the oil pipe reference temperature; is the influence factor of pressure on the diffusion coefficient; is the current pressure of the oil pipe; is the oil pipe base pressure;

[0025] S56. According to the oil film diffusion process in the oil film area, the diffusion velocity of the oil film is calculated by using a diffusion velocity calculation formula; the diffusion velocity calculation formula is: ;in, The oil film concentration is Time has come The rate of change between moments; For Time, space coordinates Oil film concentration at for The next moment of the moment; For Time has come the time difference between moments;

[0026] S57. Calculate the spatial gradient of the oil film concentration by using the spatial gradient calculation formula to estimate the diffusion direction of the oil film; the spatial gradient calculation formula is: ;in, is the oil film concentration in spatial coordinates The gradient at indicates the diffusion direction of the oil film; is the spatial coordinate The concentration of oil film at is the oil film concentration Partial derivatives about the axis; is the oil film concentration Partial derivatives about the axis;

[0027] S58. Through the steps of S51 to S57, the oil film area, the diffusion speed of the oil film and the diffusion direction characteristics of the oil film are extracted to form an oil film diffusion image feature data set.

[0028] Furthermore, the training method of the oil spill risk assessment model includes:

[0029] The data set is divided into a training set, a validation set and a test set, and an oil spill risk assessment model is constructed. The oil spill risk assessment model includes input data and output labels. The input data of the model is a historical oil pipeline state comprehensive feature data set, and the output label of the model is the oil pipeline oil spill risk rate. RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized. The hyperparameters of the model include a regularization parameter and a γ value of the RBF kernel function. The oil spill risk assessment model is a support vector machine model.

[0030] Use the ℇ-insensitive loss function as the loss function of the model to measure the difference between the model's predicted value and the true value; train the model on the data set, use the SMO algorithm to solve the optimization problem of ℇ-SVR, and adjust the hyperparameter combination through cross-validation based on the validation set data to find the hyperparameter that minimizes the loss function on the validation set;

[0031] The model is evaluated using the test data set. The model performance is evaluated by observing the performance indicators of the model in the prediction task. When the model performance no longer improves, the test is stopped to obtain a trained oil spill risk assessment model. The trained oil spill risk assessment model is used to predict the current oil pipeline status comprehensive feature data set to obtain the pipeline oil spill risk rate.

[0032] Furthermore, the method for determining whether there is an oil spill risk in the cruise ship oil pipeline according to the oil spill risk rate of the oil pipeline includes:

[0033] A threshold value of the oil pipeline spill risk rate is preset, and the predicted oil pipeline spill risk rate is compared with the preset oil pipeline spill risk rate threshold value;

[0034] If the predicted oil pipeline spill risk rate is less than the preset oil pipeline spill risk rate threshold, it is determined that there is no oil spill risk in the cruise ship oil pipeline;

[0035] If the predicted oil pipeline spill risk rate is greater than or equal to the preset oil pipeline spill risk rate threshold, it is determined that the cruise ship oil pipeline has an oil spill risk.

[0036] Furthermore, if there is an oil spill risk in the cruise ship oil pipeline, the method of providing a graded warning for the oil spill risk and triggering a corresponding alarm instruction includes: further dividing the preset oil pipeline oil spill risk rate threshold into a first oil pipeline oil spill risk rate threshold The second threshold of pipeline oil spill risk rate ; The oil pipeline spill risk rate The first threshold of pipeline spill risk rate The second threshold of pipeline oil spill risk rate To compare; if , the oil spill risk level is determined to be low risk level, and for low risk level oil spill risk, a yellow alarm instruction is triggered; if , the oil spill risk level is determined to be medium risk level, and for medium risk level oil spill risk, an orange alarm instruction is triggered; if , the oil spill risk level is determined to be a high risk level, and a red alarm instruction is triggered for the high risk level of oil spill risk.

[0037] Furthermore, the method of obtaining the spatial position data of the oil pipeline and locating the oil spill position in combination with the oil pipeline video monitoring data includes:

[0038] The three-dimensional spatial position coordinates of the oil pipe are obtained using GPS positioning, and the oil film diffusion image feature data set is reconstructed into a point cloud using three-dimensional reconstruction technology to generate a three-dimensional point cloud model of the oil film diffusion. The three-dimensional spatial position coordinates of the oil pipe are paired with the three-dimensional point cloud model of the oil film diffusion using a three-dimensional reconstruction algorithm to locate the oil spill.

[0039] A cruise ship oil pipeline spill monitoring method based on machine learning, comprising:

[0040] S1. Obtain cruise ship oil pipeline status monitoring data;

[0041] S2, preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain an oil pipeline status comprehensive feature data set;

[0042] S3. Train an oil spill risk assessment model based on the comprehensive feature data set of the oil pipeline status, and predict the risk rate of oil spill from the oil pipeline based on the oil spill risk assessment model; and determine whether there is an oil spill risk in the cruise ship oil pipeline based on the risk rate of oil spill from the oil pipeline;

[0043] S4. If there is an oil spill risk in the cruise ship's oil pipeline, a graded warning of the oil spill risk will be issued and the corresponding alarm command will be triggered;

[0044] S5. Obtain the spatial location data of the oil pipeline, locate the oil spill location in combination with the oil pipeline video monitoring data, and upload the oil spill location to the oil spill intelligent monitoring terminal to remind relevant personnel to take countermeasures.

[0045] Technical effects and advantages of a cruise ship oil pipeline oil spill monitoring system and method based on machine learning in the present invention:

[0046] The present invention dynamically designs and adjusts the overlap rate through the overlap adjustment formula, and can flexibly adjust it according to the characteristics of the oil pipeline parameter data and the time scale to be analyzed, which can ensure that each time window contains enough information while avoiding too much redundant data, which is helpful to improve the accuracy and efficiency of feature extraction; the dynamic adjustment of the overlap rate helps to capture the subtle changes of the oil pipeline data at different time scales, making the analysis more refined and improving the monitoring ability of the health status of the oil pipeline; STFT can convert the oil pipeline parameter data from the time domain to the time-frequency domain, capture the frequency components, amplitudes and phases of different time periods, and the local spectrum analysis method can help identify potential periodic features at different time scales. The time-frequency analysis of STFT can effectively extract the law of such changes, and then identify the abnormality or fault signs of the oil pipeline, and warn the problem in advance. Through STFT, the amplitude and phase information of different frequency components can be extracted from the oil pipeline parameter data. These frequency components can be used to describe the working mode of the oil pipeline. The amplitude reflects the intensity change of the signal, and the phase can provide the time delay information of the signal. Through the analysis of the time-frequency spectrum, meaningful features can be extracted from the data more accurately, reducing the influence of noise, thereby improving the effect of subsequent analysis and prediction models.

[0047] Using OpenCV to read the video frame by frame and perform threshold segmentation can accurately separate the oil film area from the background, ensuring more accurate identification of the oil film area. The introduction of the adaptive threshold adjustment formula allows the segmentation threshold to be dynamically adjusted according to the total number of pixels and total number of frames of the image, thereby adapting to the quality and characteristics of different images and avoiding incorrect segmentation due to changes in image quality; it can effectively cope with changes in image quality under different environmental conditions, improve the accuracy of oil film area separation, and thus provide more reliable input for subsequent processing; extract the edge of the oil film area through the Canny edge detection algorithm, and further use morphological operations to remove noise and connect the oil film area to obtain a clearer oil film area image, which can effectively improve the image processing effect, making the outline of the oil film more obvious and the details clearer; removing noise and strengthening the coherence of the oil film area will help improve the accuracy and stability of oil film area identification. , avoid false detection or missed detection due to interference factors; dynamically adjust the diffusion coefficient through the diffusion coefficient adjustment formula, combined with the influence of environmental factors such as temperature and pressure, the diffusion coefficient can be adjusted in real time to ensure that the modeling of the oil film diffusion process is more accurate, and can reflect the oil film diffusion in the actual environment of the oil pipe. The diffusion speed and diffusion direction of the oil film are calculated respectively through the diffusion speed calculation formula and the spatial gradient calculation formula, which can help the system track the propagation speed and direction of the oil film in real time, and then infer the changing trend of the oil spill location; by comprehensively analyzing the spatial distribution, diffusion speed and direction of the oil film, the diffusion trend of the oil film can be comprehensively analyzed, providing a strong basis for subsequent monitoring and prediction, which can significantly improve the accuracy and timeliness of oil spill detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a structural schematic diagram of a cruise ship oil pipeline oil spill monitoring system based on machine learning of the present invention;

[0049] Figure 2 The present invention is a flow chart of a cruise ship oil pipeline oil spill monitoring method based on machine learning. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention. Embodiment 1

[0051] See also Figure 1 As shown, this embodiment is a cruise ship oil pipeline oil spill monitoring system based on machine learning, including:

[0052] Data acquisition module, used to obtain cruise ship oil pipeline status monitoring data;

[0053] A data processing module is used to pre-process the acquired cruise ship oil pipeline status monitoring data to obtain a comprehensive feature data set of the oil pipeline status;

[0054] The oil spill risk prediction module is used to train an oil spill risk assessment model based on the comprehensive feature data set of the oil pipeline status, and predict the oil spill risk rate of the oil pipeline based on the oil spill risk assessment model; based on the oil spill risk rate of the oil pipeline, it is determined whether there is an oil spill risk in the cruise ship oil pipeline;

[0055] Oil spill warning module: if there is an oil spill risk in the cruise ship's oil pipeline, it will issue a graded warning of the oil spill risk and trigger the corresponding alarm command;

[0056] The oil spill source tracking module is used to obtain the spatial location data of the oil pipeline, locate the oil spill location in combination with the oil pipeline video monitoring data, and upload the oil spill location to the oil spill intelligent monitoring terminal to remind relevant personnel to take countermeasures;

[0057] The modules are connected to each other via wired and / or wireless means to achieve data transmission between modules.

[0058] Cruise ship oil pipeline status monitoring data includes oil pipeline parameter data, cruise ship status data, environmental monitoring data and oil pipeline video monitoring data;

[0059] The oil pipeline parameter data includes oil pipeline pressure data, oil pipeline flow data, oil pipeline temperature data, oil pipeline vibration data and oil pipeline wear data; the cruise ship status data includes cruise ship speed data, cruise ship heading data, cruise ship position data and cruise ship load data; the environmental monitoring data includes temperature data, humidity data, water pollution data, wind speed data and oil and gas concentration change data; the oil pipeline video monitoring data includes oil pipeline external video monitoring data and oil pipeline internal video monitoring data.

[0060] The method for preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain the oil pipeline status comprehensive feature data set includes:

[0061] The improved local Fourier transform is used to extract the features of the oil pipeline parameter data to obtain the oil pipeline parameter feature data set; the K-Means clustering algorithm is used to identify and eliminate the abnormal values ​​in the cruise ship status data and environmental monitoring data included in the cruise ship oil pipeline status monitoring data to obtain the cruise ship status feature data set and the environmental monitoring feature data set; the feature extraction of the oil pipeline video monitoring data is performed to obtain the oil film diffusion image feature data set;

[0062] The oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset are normalized by standard deviation and converted into standard normal distribution with mean 0 and standard deviation 1 to obtain the normalized oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset; the normalized oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset are weighted fused to obtain the comprehensive feature dataset of oil pipeline status.

[0063] Methods for extracting features of tubing parameter data using improved local Fourier transform to obtain tubing parameter feature data sets include:

[0064] Get Continuous tubing parameter data over time ;in, are equally spaced timestamps; is the timestamp; is the total number of timestamps; the preset overlapping time window length is The length of the overlapping time window is selected according to the characteristics of the signal and the analysis requirements. For example, if the signal changes slowly, a longer window can be selected; if the signal changes quickly, a shorter window should be selected. The overlap rate of the overlapping time windows is defined as , the tubing parameter data Divide into overlapping time windows, then the starting time interval of each overlapping time window is ;in, is the starting time interval for each overlapping time window; the overlap ratio of the overlapping time windows is adjusted by the overlap adjustment formula Perform dynamic design adjustments, and the overlap adjustment formula is: ;in, is the dimension of the oil pipeline parameter data (i.e., the number of types); is the length of the overlapping time window; is the adjustment factor for controlling the dynamic range of the overlap ratio;

[0065] The overlap adjustment formula can be adjusted according to the dimensions of different tubing parameter data. and time window length Automatically adjust the overlap rate to adapt to the characteristics of different signals. If the dimension of the pipeline parameter data is large, it usually means that the data diversity is high, and the overlap rate will be correspondingly small. If the time window is small, the overlap rate is high, which helps to capture more signal changes in a short time. If the time window is large, the overlap rate gradually decreases, thereby avoiding redundant information caused by excessive overlap. For example, the dimension of the pipeline parameter data 5, the length of overlapping time windows =10; adjustment factor for controlling the dynamic range of overlap rate is 0.1; then the overlap rate of overlapping time windows .

[0066] In each overlapping time window, the local Fourier transform of the oil pipeline parameter data is performed by applying STFT to obtain the spectrum information corresponding to each overlapping time window. ;in, is the complex number representation of the tubing parameter data in the time-frequency domain, that is, the tubing parameter data in time and frequency The amplitude and phase of the following; is a window function, which is used to limit the overlapping time windows of the local Fourier transform and determines the range of the tubing parameter data within each overlapping time window. Commonly used window functions include rectangular window, Hamming window, Hanning window, etc. is the offset within the overlapping time window, and the window function is The location weights the tubing parameter data; is a complex exponential function, representing the frequency component in the local Fourier transform; is the frequency, which represents the frequency domain component after local Fourier transform; the spectrum information corresponding to each overlapping time window is spliced ​​and combined into the STFT time spectrum; the STFT time spectrum is a two-dimensional matrix containing time and frequency; the frequency component, amplitude and phase of the oil pipeline parameter data are extracted by analyzing the STFT time spectrum, and then the oil pipeline parameter feature data set is obtained;

[0067] By analyzing the frequency components, periodic fluctuations can be identified, such as periodic changes in pipeline pressure or flow signals. Amplitude is a key indicator in frequency domain analysis, indicating the strength of each frequency component in a specific overlapping time window. It can help identify the periodic characteristics and frequency characteristics in the pipeline parameter data. The larger the amplitude, the stronger the contribution of the frequency component to the pipeline parameter data. Phase information provides the timing information of the pipeline parameter data in the time window, reflecting the time structure and periodic changes of the pipeline parameter data, which is very important for timing analysis. For example, a sudden change in phase may indicate that the pipeline parameter data has changed rapidly.

[0068] The method for extracting features from the oil pipeline video monitoring data to obtain an oil film diffusion image feature data set includes:

[0069] S51, use the library function in the video processing library OpenCV to read the images in the oil pipeline video monitoring data frame by frame, and these frame images contain the dynamic information of oil film diffusion; the preset oil pipeline video monitoring data has a total of Frame image, preset Frame image in spatial coordinates The pixel value at ;in, and are the spatial horizontal and vertical coordinates of the image respectively; The index of the image frame, ranging from 1 to ;

[0070] S52, preset a segmentation threshold, and segment the image using a threshold segmentation function to separate the oil film area from the background area. The threshold segmentation function is: ;in, Indicates the oil film area. Frame image in spatial coordinates The pixel value at When it is greater than the segmentation threshold, the pixel is considered to belong to the oil film area, so it is assigned ; Represents the background area. Frame image in spatial coordinates The pixel value at When it is less than or equal to the segmentation threshold, the pixel is considered to belong to the background area, so it is assigned ; is the segmentation threshold; the segmentation threshold is adjusted by the adaptive threshold adjustment formula Dynamic adjustment is performed, and the adaptive threshold adjustment formula is: ;in, is the segmentation threshold after dynamic adjustment; is the total number of pixels of the image; is the total number of frames of the image; To control the influence of the total number of pixels of the image on the segmentation threshold; To control the influence of the total number of frames of the image on the segmentation threshold; as time goes by, the changes in the image in the video will lead to dynamic adjustment of the threshold. According to the adaptive threshold adjustment formula, The increase in the segmentation threshold The smaller the threshold, the more detailed the segmentation. The smaller the threshold, which means that as the image changes dynamically, a smaller threshold is needed to process the details, thereby achieving a finer segmentation. The image segmentation threshold is usually not fixed, especially in dynamic scenes (for example, the oil film diffusion process). As the image content changes, the segmentation threshold needs to be dynamically adjusted. and total number of pixels The combined use of can realize the flexible adjustment of different image data sets to meet the needs of image segmentation in different scenarios; for example, the segmentation threshold is 100, the total number of pixels in the image 250000; the total number of frames of the image is 500; controls the influence factor of the total number of pixels of the image on the segmentation threshold 0.1; controls the influence of the total number of image frames on the segmentation threshold is 0.2; then the dynamically adjusted segmentation threshold .

[0071] S53, further extract the edge of the oil film area through Canny edge detection ,in, For the The result of Canny edge detection for each pixel in the frame image; morphological operations (such as corrosion and expansion) are used to improve the segmentation effect, remove noise and connect the oil film area to obtain the oil film area image ;

[0072] S54, the oil film diffusion process in the preset oil film area follows the diffusion equation, which is: ;in, For Time, space coordinates Oil film concentration at is the diffusion coefficient, which indicates the rate of oil film diffusion and describes the ability of oil film diffusion, applicable to all time and space coordinates; is the Laplace operator, which indicates the spatial variation of oil film concentration; is the index of the moment;

[0073] S55, Diffusion coefficient adjustment formula Dynamic adjustment is performed, and the diffusion coefficient adjustment formula is: ;in, is the diffusion coefficient after dynamic adjustment; is the influence factor of temperature on the diffusion coefficient; is the current temperature of the oil pipe; is the oil pipe reference temperature; is the influence factor of pressure on the diffusion coefficient; is the current pressure of the oil pipe; is the base pressure of the oil pipe; molecular dynamics theory shows that as temperature rises, the average kinetic energy of molecules increases, causing the molecular motion to accelerate, thereby increasing the diffusion rate, and the diffusion coefficient increases with increasing temperature; an increase in pressure usually compresses the fluid and reduces the space between molecules, which may slow down the movement of molecules and reduce the diffusion rate; when temperature rises, the molecular kinetic energy increases and the diffusion coefficient increases; and when pressure increases, the distance between molecules decreases and the diffusion coefficient decreases. For example, the diffusion coefficient is 0.5, the influence factor of temperature on the diffusion coefficient 0.1, the current temperature of the oil pipe 30 degrees; oil pipe reference temperature is 25 degrees; the influence factor of pressure on the diffusion coefficient 0.6; current pressure of the oil pipe 12; oil pipe base pressure is 10; then the dynamically adjusted diffusion coefficient .

[0074] S56. According to the oil film diffusion process in the oil film area, the diffusion velocity of the oil film is calculated by using a diffusion velocity calculation formula; the diffusion velocity calculation formula is: ;in, The oil film concentration is Time has come The rate of change between moments, that is, the diffusion rate of the oil film, is different from the diffusion coefficient , is the diffusion velocity of the actual oil film, reflecting the rate at which the oil film concentration changes with time; For Time, space coordinates Oil film concentration at for The next moment of the moment; For Time has come the time difference between moments;

[0075] S57. Calculate the spatial gradient of the oil film concentration by using the spatial gradient calculation formula to estimate the diffusion direction of the oil film; the spatial gradient calculation formula is: ;in, is the oil film concentration in spatial coordinates The gradient at indicates the diffusion direction of the oil film; is the spatial coordinate The concentration of oil film at is the oil film concentration The partial derivative in the axial direction describes the rate of change of the oil film concentration in the horizontal direction; is the oil film concentration The partial derivative in the axial direction describes the rate of change of the oil film concentration in the vertical direction;

[0076] S58. Through the steps of S51 to S57, the oil film area, the diffusion speed of the oil film and the diffusion direction characteristics of the oil film are extracted to form an oil film diffusion image feature data set.

[0077] The training methods of the oil spill risk assessment model include:

[0078] The data set is divided into a training set, a validation set and a test set, and an oil spill risk assessment model is constructed. The oil spill risk assessment model includes input data and output labels. The input data of the model is a historical oil pipeline state comprehensive feature data set, and the output label of the model is the oil pipeline oil spill risk rate. RBF is selected as the kernel function of the model, and the hyperparameters of the model are initialized. The hyperparameters of the model include a regularization parameter and a γ value of the RBF kernel function. The oil spill risk assessment model is a support vector machine model.

[0079] Use the ℇ-insensitive loss function as the loss function of the model to measure the difference between the model's predicted value and the true value; train the model on the data set, use the SMO algorithm to solve the optimization problem of ℇ-SVR, and adjust the hyperparameter combination through cross-validation based on the validation set data to find the hyperparameter that minimizes the loss function on the validation set;

[0080] The model is evaluated using the test data set. The model performance is evaluated by observing the performance indicators of the model in the prediction task. When the model performance no longer improves, the test is stopped to obtain a trained oil spill risk assessment model. The trained oil spill risk assessment model is used to predict the current oil pipeline status comprehensive feature data set to obtain the pipeline oil spill risk rate.

[0081] Based on the oil pipeline spill risk rate, the methods for determining whether there is an oil spill risk in the cruise ship oil pipeline include:

[0082] A threshold value of the oil pipeline spill risk rate is preset, and the predicted oil pipeline spill risk rate is compared with the preset oil pipeline spill risk rate threshold value;

[0083] If the predicted oil pipeline spill risk rate is less than the preset oil pipeline spill risk rate threshold, it is determined that there is no oil spill risk in the cruise ship oil pipeline;

[0084] If the predicted oil pipeline spill risk rate is greater than or equal to the preset oil pipeline spill risk rate threshold, it is determined that the cruise ship oil pipeline has an oil spill risk.

[0085] If there is an oil spill risk in the cruise ship oil pipeline, the method of providing a graded warning for the oil spill risk and triggering a corresponding alarm instruction includes: further dividing the preset oil pipeline oil spill risk rate threshold into the first oil pipeline oil spill risk rate threshold The second threshold of pipeline oil spill risk rate ; The oil pipeline spill risk rate The first threshold of pipeline spill risk rate The second threshold of pipeline oil spill risk rate To compare; if , the oil spill risk level is determined to be low risk level, and for low risk level oil spill risk, a yellow alarm instruction is triggered; if , the oil spill risk level is determined to be medium risk level, and for medium risk level oil spill risk, an orange alarm instruction is triggered; if , the oil spill risk level is determined to be a high risk level, and a red alarm instruction is triggered for the high risk level of oil spill risk.

[0086] The method of obtaining the spatial position data of the oil pipeline and locating the oil spill location in combination with the oil pipeline video monitoring data includes:

[0087] Use GPS positioning to obtain the three-dimensional spatial position coordinates of the oil pipe, and use the three-dimensional reconstruction technology to reconstruct the point cloud of the oil film diffusion image feature data set to generate a three-dimensional point cloud model of the oil film diffusion; use the three-dimensional reconstruction algorithm to pair the three-dimensional spatial position coordinates of the oil pipe with the three-dimensional point cloud model of the oil film diffusion. First, it is necessary to convert the three-dimensional spatial position coordinate system of the oil pipe into the same coordinate system as the three-dimensional model of the oil film diffusion, and then align the point cloud model of the oil film diffusion to the position coordinates of the oil pipe through geometric transformations such as rotation and translation. The ICP algorithm can be used, which can find the optimal pairing between point clouds and optimize the matching between coordinate systems, thereby locating the oil spill location.

[0088] The preset oil pipeline oil spill risk rate threshold is set by the staff, and different oil pipeline oil spill risk rates are collected through the oil spill intelligent monitoring terminal, and the average value of multiple oil pipeline oil spill risk rates is taken as the preset oil pipeline oil spill risk rate threshold; similarly, the first threshold of the oil pipeline oil spill risk rate is set The second threshold of pipeline oil spill risk rate .

[0089] In this embodiment, the overlap rate is dynamically designed and adjusted through the overlap adjustment formula. It can be flexibly adjusted according to the characteristics of the oil pipeline parameter data and the time scale to be analyzed, which can ensure that each time window contains sufficient information while avoiding excessive redundant data, which helps to improve the accuracy and efficiency of feature extraction; the dynamic adjustment of the overlap rate helps to capture subtle changes in the oil pipeline data at different time scales, making the analysis more refined and improving the monitoring ability of the health status of the oil pipeline; STFT can convert the oil pipeline parameter data from the time domain to the time-frequency domain, capture the frequency components, amplitudes and phases of different time periods, and the local spectrum analysis method can help identify potential periodicity at different time scales. Characteristics or mutations reflect the dynamic changes of oil pipeline parameters; the working status of the oil pipeline may change over time, and the time-frequency analysis of STFT can effectively extract the law of such changes, and then identify abnormalities or signs of failure of the oil pipeline, and warn of problems in advance; through STFT, the amplitude and phase information of different frequency components can be extracted from the oil pipeline parameter data. These frequency components can be used to describe the working mode of the oil pipeline. The amplitude reflects the intensity change of the signal, and the phase can provide the time delay information of the signal; through the analysis of the time-frequency spectrum, meaningful features can be extracted from the data more accurately, reducing the impact of noise, thereby improving the effect of subsequent analysis and prediction models;

[0090] Using OpenCV to read the video frame by frame and perform threshold segmentation can accurately separate the oil film area from the background, ensuring more accurate identification of the oil film area. The introduction of the adaptive threshold adjustment formula allows the segmentation threshold to be dynamically adjusted according to the total number of pixels and total number of frames of the image, thereby adapting to the quality and characteristics of different images and avoiding incorrect segmentation due to changes in image quality; it can effectively cope with changes in image quality under different environmental conditions, improve the accuracy of oil film area separation, and thus provide more reliable input for subsequent processing; extract the edge of the oil film area through the Canny edge detection algorithm, and further use morphological operations to remove noise and connect the oil film area to obtain a clearer oil film area image, which can effectively improve the image processing effect, making the outline of the oil film more obvious and the details clearer; removing noise and strengthening the coherence of the oil film area will help improve the accuracy and stability of oil film area identification. , avoid false detection or missed detection due to interference factors; dynamically adjust the diffusion coefficient through the diffusion coefficient adjustment formula, combined with the influence of environmental factors such as temperature and pressure, the diffusion coefficient can be adjusted in real time to ensure that the modeling of the oil film diffusion process is more accurate, and can reflect the oil film diffusion in the actual environment of the oil pipe. The diffusion speed and diffusion direction of the oil film are calculated respectively through the diffusion speed calculation formula and the spatial gradient calculation formula, which can help the system track the propagation speed and direction of the oil film in real time, and then infer the changing trend of the oil spill location; by comprehensively analyzing the spatial distribution, diffusion speed and direction of the oil film, the diffusion trend of the oil film can be comprehensively analyzed, providing a strong basis for subsequent monitoring and prediction, which can significantly improve the accuracy and timeliness of oil spill detection. Embodiment 2

[0091] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a cruise ship oil pipeline oil spill monitoring method based on machine learning, including:

[0092] S1. Obtain cruise ship oil pipeline status monitoring data;

[0093] S2, preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain an oil pipeline status comprehensive feature data set;

[0094] S3. Train an oil spill risk assessment model based on the comprehensive feature data set of the oil pipeline status, and predict the risk rate of oil spill from the oil pipeline based on the oil spill risk assessment model; and determine whether there is an oil spill risk in the cruise ship oil pipeline based on the risk rate of oil spill from the oil pipeline;

[0095] S4. If there is an oil spill risk in the cruise ship's oil pipeline, a graded warning of the oil spill risk will be issued and the corresponding alarm command will be triggered;

[0096] S5. Obtain the spatial location data of the oil pipeline, locate the oil spill location in combination with the oil pipeline video monitoring data, and upload the oil spill location to the oil spill intelligent monitoring terminal to remind relevant personnel to take countermeasures.

[0097] Since the electronic device introduced in this embodiment is an electronic device used to implement a cruise ship oil pipeline oil spill monitoring system and method based on machine learning in the embodiment of this application, based on the cruise ship oil pipeline oil spill monitoring system and method based on machine learning introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not introduced in detail here. As long as the technical personnel of this field implement the electronic device used by the cruise ship oil pipeline oil spill monitoring system and method based on machine learning in the embodiment of this application, it belongs to the scope of protection of this application.

[0098] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0099] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A cruise ship oil pipeline spill monitoring system based on machine learning, characterized in that: include: Data acquisition module, used to obtain cruise ship oil pipeline status monitoring data; A data processing module is used to pre-process the acquired cruise ship oil pipeline status monitoring data to obtain a comprehensive feature data set of the oil pipeline status; The preprocessing method includes using improved local Fourier transform to extract features of the oil pipe parameter data to obtain an oil pipe parameter feature data set, specifically including: obtaining continuous oil pipe parameter data within a time period of h x(h) = [x(h1), x(h2), ..., x(h M )];in, h1,h2,...,h M is a time stamp with equal intervals; h is a time stamp; M is the total number of time stamps; the length of the overlapping time window is preset to be K, the overlapping rate of the overlapping time window is defined to be θ, and the oil pipe parameter data x(h) is divided into n overlapping time windows, then the starting time interval of each overlapping time window is K s =K·(1-θ); where K s is the starting time interval of each overlapping time window; The overlap rate θ of the overlapping time windows is dynamically designed and adjusted through the overlap adjustment formula, and the overlap adjustment formula is: Among them, G is the dimension of the tubing parameter data; K is the length of the overlapping time window; σ is the adjustment factor for controlling the dynamic range of the overlap rate; The preprocessing method further includes extracting features from the oil pipeline video monitoring data to obtain an oil film diffusion image feature data set, specifically including: S51, using the library function in the video processing library OpenCV to read the images in the YouTube video monitoring data frame by frame; it is assumed that there are T frames of images in the YouTube video monitoring data, and the pixel value of the t-th frame image at the spatial coordinate (a, b) is I t (a,b); where a and b are the spatial horizontal and vertical coordinates of the image respectively; t is the index of the image frame, ranging from 1 to T; S52, preset a segmentation threshold, and segment the image using a threshold segmentation function to separate the oil film area from the background area. The threshold segmentation function is: Among them, 1 represents the oil film area; 0 represents the background area; B is the segmentation threshold; The segmentation threshold B is dynamically adjusted by the adaptive threshold adjustment formula, and the adaptive threshold adjustment formula is: Among them, B' is the segmentation threshold after dynamic adjustment; N is the total number of pixels in the image; T is the total number of frames in the image; β is the influence factor of the total number of pixels in the control image on the segmentation threshold; η is the influence factor of the total number of frames in the control image on the segmentation threshold; S53, further extract the edge E of the oil film area through Canny edge detection t (a,b), where E t (a, b) are the results of Canny edge detection for each pixel in the t-th frame image; the segmentation effect is improved by morphological operations, noise is removed and the oil film area is connected to obtain the oil film area image M t (a, b); According to the oil film diffusion process in the oil film area, the diffusion velocity of the oil film is calculated by the diffusion velocity calculation formula; the spatial gradient of the oil film concentration is calculated by the spatial gradient calculation formula to estimate the diffusion direction of the oil film; the oil film area, the diffusion velocity of the oil film and the diffusion direction of the oil film are extracted to form an oil film diffusion image feature data set; The oil spill risk prediction module is used to train an oil spill risk assessment model based on the comprehensive feature data set of the oil pipeline status, and predict the oil spill risk rate of the oil pipeline based on the oil spill risk assessment model; based on the oil spill risk rate of the oil pipeline, it is determined whether there is an oil spill risk in the cruise ship oil pipeline; Oil spill warning module: if there is an oil spill risk in the cruise ship's oil pipeline, it will issue a graded warning of the oil spill risk and trigger the corresponding alarm command; The oil spill source tracking module is used to obtain the spatial location data of the oil pipeline, locate the oil spill location in combination with the oil pipeline video monitoring data, and upload the oil spill location to the oil spill intelligent monitoring terminal to remind relevant personnel to take countermeasures; The modules are connected to each other via wired and / or wireless means to achieve data transmission between the modules.

2. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 1, characterized in that: The cruise ship oil pipeline status monitoring data includes oil pipeline parameter data, cruise ship status data, environmental monitoring data and oil pipeline video monitoring data; The oil pipeline parameter data includes oil pipeline pressure data, oil pipeline flow data, oil pipeline temperature data, oil pipeline vibration data and oil pipeline wear data; the cruise ship status data includes cruise ship speed data, cruise ship heading data, cruise ship position data and cruise ship load data; the environmental monitoring data includes temperature data, humidity data, water pollution data, wind speed data and oil and gas concentration change data; the oil pipeline video monitoring data includes oil pipeline external video monitoring data and oil pipeline internal video monitoring data.

3. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 2 is characterized in that: The method of preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain an oil pipeline status comprehensive feature data set includes: The K-Means clustering algorithm is used to identify and remove outliers in the cruise ship status data and environmental monitoring data included in the cruise ship oil pipeline status monitoring data, and the cruise ship status feature data set and the environmental monitoring feature data set are obtained; The oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset are normalized by standard deviation and converted into standard normal distribution with mean 0 and standard deviation 1 to obtain the normalized oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset; the normalized oil pipeline parameter feature dataset, cruise ship status feature dataset, environmental monitoring feature dataset and oil film diffusion image feature dataset are weighted fused to obtain the comprehensive feature dataset of oil pipeline status.

4. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 3 is characterized in that: The method of extracting features from the oil pipe parameter data using the improved local Fourier transform to obtain the oil pipe parameter feature data set also includes: In each overlapping time window, the local Fourier transform of the oil pipeline parameter data is performed by applying STFT to obtain the spectrum information corresponding to each overlapping time window. Where X(h,f) is the complex representation of the tubing parameter data in the time-frequency domain; w(h-τ) is the window function; τ is the offset in the overlapping time window; e -j2πfτ is a complex exponential function; f is the frequency, and the spectrum information corresponding to each overlapping time window is concatenated and combined into the STFT time-spectrum; the frequency component, amplitude and phase of the oil pipeline parameter data are extracted by analyzing the STFT time-spectrum, and then the oil pipeline parameter feature data set is obtained.

5. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 4 is characterized in that: The method for obtaining the oil film diffusion image feature dataset by extracting features from the oil pipe video monitoring data further includes: S54, the oil film diffusion process in the preset oil film area follows the diffusion equation, which is: Where, C(a,b,t′) is the oil film concentration at the spatial coordinate (a,b) at time t′; D is the diffusion coefficient; is the Laplace operator, which indicates the spatial variation of oil film concentration; t′ is the index of the time; S55. Dynamically adjust the diffusion coefficient D through a diffusion coefficient adjustment formula. The diffusion coefficient adjustment formula is: Among them, D' is the diffusion coefficient after dynamic adjustment; α is the influence factor of temperature on the diffusion coefficient; T' is the current temperature of the oil pipe; T'0 is the reference temperature of the oil pipe; δ is the influence factor of pressure on the diffusion coefficient; P is the current pressure of the oil pipe; P0 is the reference pressure of the oil pipe; S56, the diffusion rate calculation formula is: Among them, V t (a, b) is the rate of change of oil film concentration between time t′ and time t′+1; C(a, b, t′+1) is the oil film concentration at the spatial coordinate (a, b) at time t′+1; t′+1 is the next moment after time t′; Δt′ is the time difference from time t′ to time t′+1; S57, spatial gradient calculation formula is: in, is the gradient of the oil film concentration at the spatial coordinate (a, b), indicating the diffusion direction of the oil film; C(a, b) is the concentration of the oil film at the spatial coordinate (a, b); is the partial derivative of the oil film concentration along the a-axis; is the partial derivative of the oil film concentration along the b-axis; S58. Through the steps of S51 - S57, extract the features of the oil film area, the diffusion speed of the oil film, and the diffusion direction of the oil film to form an oil film diffusion image feature dataset.

6. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 5, characterized in that: The training method of the oil spill risk assessment model includes: Divide the dataset into a training set, a validation set, and a test set, and construct an oil spill risk assessment model. The oil spill risk assessment model includes input data and output labels; the input data of the model is the historical comprehensive feature dataset of the oil pipe state, and the output label of the model is the oil pipe oil spill risk rate; select RBF as the kernel function of the model, and initialize the hyperparameters of the model. The hyperparameters of the model include the regularization parameter and the γ value of the RBF kernel function; the oil spill risk assessment model is a support vector machine model; Use the ε-insensitive loss function as the loss function of the model to measure the difference between the predicted value and the true value of the model; perform model training on the dataset, use the SMO algorithm to solve the optimization problem of ε-SVR, and adjust the hyperparameter combination through cross-validation according to the validation set data to find the hyperparameters that minimize the loss function on the validation set; Use the test dataset to evaluate the model, and evaluate the model performance by observing the performance metrics of the model in the prediction task. Stop the test when the model performance no longer improves to obtain a trained oil spill risk assessment model; use the trained oil spill risk assessment model to predict the current comprehensive feature dataset of the oil pipe state to obtain the oil pipe oil spill risk rate.

7. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 6, characterized in that: The method for judging whether there is an oil spill risk in the cruise ship oil pipe according to the oil pipe oil spill risk rate includes: Preset an oil pipe oil spill risk rate threshold, and compare the predicted oil pipe oil spill risk rate with the preset oil pipe oil spill risk rate threshold; If the predicted oil pipe oil spill risk rate is less than the preset oil pipe oil spill risk rate threshold, it is determined that there is no oil spill risk in the cruise ship oil pipe; If the predicted oil pipe oil spill risk rate is greater than or equal to the preset oil pipe oil spill risk rate threshold, it is determined that there is an oil spill risk in the cruise ship oil pipe.

8. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 7, characterized in that: The method for grading and warning the oil spill risk and triggering corresponding alarm instructions if there is an oil spill risk in the cruise ship oil pipe includes: further divide the preset oil pipe oil spill risk rate threshold into the first oil pipe oil spill risk rate threshold E1 and the second oil pipe oil spill risk rate threshold E2; compare the oil pipe oil spill risk rate E with the first oil pipe oil spill risk rate threshold E1 and the second oil pipe oil spill risk rate threshold E2 respectively; If E < E1, it is determined that the oil spill risk level is a low risk level. For the oil spill risk of the low risk level, trigger a yellow alarm instruction; if E1 < E ≤ E2, it is determined that the oil spill risk level is a medium risk level. For the oil spill risk of the medium risk level, trigger an orange alarm instruction; if E > E2, it is determined that the oil spill risk level is a high risk level. For the oil spill risk of the high risk level, trigger a red alarm instruction.

9. The cruise ship oil pipeline oil spill monitoring system based on machine learning according to claim 8, characterized in that: The method for obtaining the oil pipe spatial position data and combining the oil pipe video monitoring data to locate the oil spill position includes: The three-dimensional spatial position coordinates of the oil pipe are obtained using GPS positioning, and the oil film diffusion image feature data set is reconstructed into a point cloud using three-dimensional reconstruction technology to generate a three-dimensional point cloud model of the oil film diffusion. The three-dimensional spatial position coordinates of the oil pipe are paired with the three-dimensional point cloud model of the oil film diffusion using a three-dimensional reconstruction algorithm to locate the oil spill.

10. A cruise ship oil pipeline oil spill monitoring method based on machine learning, used to implement a cruise ship oil pipeline oil spill monitoring system based on machine learning as claimed in any one of claims 1 to 9, characterized in that: include: S1. Obtain cruise ship oil pipeline status monitoring data; S2, preprocessing the acquired cruise ship oil pipeline status monitoring data to obtain a comprehensive feature data set of the oil pipeline status; S3. Train an oil spill risk assessment model based on the comprehensive feature data set of the oil pipeline status, and predict the risk rate of oil spill from the oil pipeline based on the oil spill risk assessment model; and determine whether there is an oil spill risk in the cruise ship oil pipeline based on the risk rate of oil spill from the oil pipeline; S4. If there is an oil spill risk in the cruise ship's oil pipeline, a graded warning of the oil spill risk will be issued and the corresponding alarm command will be triggered; S5. Obtain the spatial location data of the oil pipeline, locate the oil spill location in combination with the oil pipeline video monitoring data, and upload the oil spill location to the oil spill intelligent monitoring terminal to remind relevant personnel to take countermeasures.

Images