Ship operation cable damage degree identification and early warning method and system

Abstract

The invention provides a ship operation cable damage degree identification and early warning method and system, and relates to the technical field of traffic safety and environment, and the method comprises the steps: obtaining cable state data corresponding to a to-be-identified ship operation cable, and calling a cable damage degree identification model to carry out the damage identification of the cable state data, the cable damage degree identification model is a backbone network including a YOLOv8 model of a large-scale selection kernel network, and a coordinate attention network is embedded in the backbone network; and determining an early warning level of ship operation cable damage based on the state type, and performing early warning according to an early warning strategy corresponding to the early warning level. The method is used for solving the technical problems that direct identification of a large special ship cable damage body is lacked, and the identification precision and the early warning time efficiency are difficult to balance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of traffic safety and environmental technology, and particularly relates to a ship operation cable damage degree identification and early warning method and system. BACKGROUND

[0002] In the modern port transportation system, ship cable as the "lifeline" to maintain the stable connection between ship and shore facilities, its performance state directly determines the safety of ship anchoring and the efficiency of port operation. From the berthing of ocean liners to the towing operation of engineering ships, the cable always bears the core function of resisting wind and wave impact and restraining ship displacement, and is the key barrier to ensure water transportation safety. However, the complex and changeable marine / inland / port and other water environment poses a serious challenge to the durability of the cable: the continuous dynamic load causes structural fatigue, the salt water environment accelerates material corrosion, and mechanical friction causes surface wear, all of which lead to structural damage such as tearing, breaking, and local breaking of the cable. If the damage is not discovered in time, it may cause ship displacement, collision and other accidents, or even cause disastrous consequences such as ship overturning, which seriously threatens waterway transportation safety and water ecological environment. Therefore, building an efficient and accurate ship cable damage identification and intelligent early warning technology system has become a key issue to be tackled in the field of waterway transportation and port safety.

[0003] Traditional ship cable operation state detection methods have significant limitations in complex working conditions. Manual visual inspection relies on the experience and subjective judgment of the detection personnel, and in dangerous operating environments such as high altitude and underwater, it is not only inefficient, but also difficult to capture hidden damage. Although contact detection methods such as static tension testing can evaluate the load-carrying capacity of the cable, they inevitably cause secondary damage to the cable and cannot perform real-time monitoring and identification. Sensor detection technologies based on ultrasonic and electromagnetic principles are severely affected by salt spray corrosion, water mist interference, and ship body vibration in water environments, making it difficult to meet the demand for high-precision and continuous monitoring. With the rapid development of artificial intelligence and machine vision technology, non-contact and intelligent detection solutions have gradually become a research hotspot. Among them, target detection algorithms that integrate deep learning have brought new technical paths for ship cable damage degree identification and safety warning due to their strong feature extraction capability and real-time processing advantage. However, existing methods still have technical bottlenecks: most solutions focus on the indirect analysis of the stress state of ordinary cargo ship cables, and lack direct identification of the damage ontology of large special ships (such as transport ships and ocean-going container ships); traditional machine vision algorithms have insufficient recognition accuracy for subtle damage in extreme scenarios of special ships (such as cable material hardening caused by -160 degrees Celsius low temperature of transport ships and rapid damage spread caused by high load of container ships), and do not adapt to image grayscale distortion in low temperature environments; at the same time, existing algorithms do not consider the characteristics of special ship cables, such as high cost of single damage repair and large loss of downtime, and cannot balance the recognition accuracy and warning timeliness. SUMMARY

[0004] Therefore, it is necessary to provide a ship operation cable damage degree identification and warning method and system to solve the technical problems of lacking direct identification of the damage ontology of large special ship cables and difficulty in balancing recognition accuracy and warning timeliness.

[0005] To solve the above problems, the present application provides a ship operation cable damage degree identification and warning method, comprising: Obtaining cable state data corresponding to the ship operation cable to be identified, the cable state data including cable images and ship environment data; Calling a cable damage degree identification model to identify the damage of the cable state data, and obtaining the state type of the damage degree of the ship operation cable to be identified, the cable damage degree identification model being a YOLOv8 model including a large selection kernel network, and the backbone network being embedded with a coordinate attention network; Determining the warning level of the ship operation cable damage based on the state type, and warning according to the corresponding warning strategy of the warning level.

[0006] In a possible implementation, the cable damage degree identification model is constructed by the following method: The C2f module with a set number of layers in the backbone network of the YOLOv8 model is updated to a large selection kernel network, and a coordinate attention network is embedded before a spatial pyramid pooling module in the backbone network to obtain a cable damage degree recognition model.

[0007] In a possible implementation, the training process of the cable damage degree recognition model comprises: An image sample of a ship operation cable is constructed, and the image sample is labeled with a state label of a cable health state and a damage area label box, the state label corresponding one-to-one to a state type of a to-be-identified ship operation cable damage degree; The image sample is input into the cable damage degree recognition model for forward propagation to obtain a damage area prediction box corresponding to the ship operation cable; A WIoU v3 loss function is constructed according to the damage area label box and the damage area label box; The WIoU v3 loss function is used for backward propagation in the cable damage degree recognition model to update model parameters of the cable damage degree recognition model.

[0008] In a possible implementation, the state label comprises: intact, slight wear, severe wear, and breakage, and the state label is determined in the following manner: When the ship operation cable in the image sample has no local wear and no local fiber breakage, the state label of the image sample is determined to be intact; When the ship operation cable in the image sample has local wear and the area proportion of the local wear does not exceed a preset proportion threshold, the state label of the image sample is determined to be slight wear; When the ship operation cable in the image sample has local wear and the area proportion of the local wear has exceeded the proportion threshold or the cable has local fiber breakage but the number of breakage roots does not exceed a preset number threshold, the state label of the image sample is determined to be severe wear; When the number of breakage roots of the local fiber breakage cable of the ship operation cable in the image sample has exceeded the number threshold or there is breakage, the state label of the image sample is determined to be breakage.

[0009] In a possible implementation, the calling of the cable damage degree recognition model to perform damage identification on the cable state data to obtain the state type of the to-be-identified ship operation cable damage degree comprises: Image features and environment features of the cable state data are extracted, and the image features and the environment features are fused to obtain fused features; The cable damage degree recognition model is called to recognize damage of the fused features, so as to obtain a damage area frame corresponding to the ship operation cable; An injury degree score is calculated for the damage area regression frame, and a state type of the damage degree of the ship operation cable to be recognized is determined based on the injury degree score.

[0010] In a possible implementation, the method further includes: When the state type is the intact type, the method further includes determining that the warning level of the ship operation cable damage is no warning. When the state type is the slight wear type, the method further includes determining that the warning level of the ship operation cable damage is a first-level warning, and the first-level warning is used to trigger manual review. When the state type is the serious wear type, the method further includes determining that the warning level of the ship operation cable damage is a second-level warning, and the second-level warning is used to trigger suspension of the ship operation. When the state type is the breakage type, the method further includes determining that the warning level of the ship operation cable damage is a third-level warning, and the third-level warning is used to trigger ship emergency braking.

[0011] In a possible implementation, the method further includes: When the warning level is the first-level warning, the method further includes triggering recording of abnormal information of the ship operation cable to be recognized and sending manual review warning information. When the warning level is the second-level warning, the method further includes triggering the ship alarm to alarm and sending a ship operation suspension instruction. When the warning level is the third-level warning, the method further includes triggering the ship alarm to alarm and sending the ship operation suspension instruction, and triggering a ship emergency braking signal.

[0012] The application further provides a ship operation cable damage degree recognition and warning system, including: A data acquisition module is configured to acquire cable state data corresponding to a ship operation cable to be recognized, the cable state data including a cable image and ship environment data. A damage recognition module is configured to call a cable damage degree recognition model to recognize damage of the cable state data, so as to obtain a state type of the damage degree of the ship operation cable to be recognized, the cable damage degree recognition model being a YOLOv8 model including a backbone network of a large selection kernel network, and the backbone network being embedded with a coordinate attention network. A hierarchical warning module is configured to determine a warning level of the ship operation cable damage based on the state type, and perform warning according to a preset warning strategy corresponding to the warning level.

[0013] The application further provides an electronic device, comprising a memory and a processor, wherein the memory is configured to store a program; and the processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the ship operation cable damage degree identification and early warning method.

[0014] The application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the ship operation cable damage degree identification and early warning method.

[0015] The beneficial effects of the above implementation manner are as follows: the ship operation cable damage degree identification and early warning method and system provided by the application realizes the direct identification of the ship operation cable body based on the cable image itself in combination with the ship environment data and by using an improved YOLOv8 model. The improved YOLOv8 model uses a large selection kernel network, can adaptively adjust the convolution kernel scale according to the spatial distribution of the cable in the image, can capture the overall structural features of the long-distance cable, and can focus on the local small damage area. The coordinate attention network can accurately locate the position of the cable in the complex background, and can concentrate the attention of the loss identification on the key features such as the wear marks and broken fibers on the surface of the cable, thereby significantly improving the identification sensitivity of the small target damage. The above improvements reduce the interference of the environmental state on the damage identification, significantly improve the damage identification accuracy of the ship operation cable body, and trigger the hierarchical early warning according to the identification result after the loss identification, thereby ensuring the timeliness of the early warning response. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Figure 1 The flowchart of the ship operation cable damage degree identification and early warning method provided by the application is shown in the figure. Figure 2 The example diagram of part of the image samples provided by the application is shown in the figure. Figure 3 The principle diagram of the ship operation cable damage degree identification provided by the application is shown in the figure. Figure 4 The recognition effect diagram of the cable damage degree identification model provided by the application is shown in the figure. Figure 5 The structure diagram of the ship operation cable damage degree identification and early warning system provided by the application is shown in the figure. Figure 6An embodiment structure schematic diagram of the electronic device provided by the present application is shown. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the protection scope of the present application.

[0019] In the description of the embodiments of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified.

[0020] In the embodiments of the present application, the terms "comprising" and "having" and any variations thereof are intended to cover the inclusions that are not exclusive, for example, the processes, methods, systems, products or devices comprising a series of steps or modules do not have to be limited to the clearly listed steps or modules, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] The naming or numbering of the steps appearing in the embodiments of the present application does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The flow steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0022] Reference to "an embodiment" in this document means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily independent or alternative embodiments to other embodiments. It is explicitly and implicitly understood by a person skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] The ship operation cable damage degree identification and early warning method provided by the present application can be applied in a ship navigation operation scene, and can be deployed in a ship alarm system or a ship internal communication system. Through various sensors deployed on the ship body to collect cable state data, and then upload the cable state data back to the ship alarm system or the ship internal communication system, then call the ship operation cable damage degree identification and early warning method provided by the present application, realize the damage degree identification of the ship operation cable and trigger the early warning, the object of the early warning can be the ship alarm system or the ship internal communication system, so as to trigger the early warning to prompt the ship operation personnel or the shore station, and take corresponding response measures.

[0024] The ship operation cable damage degree recognition and early warning method provided by the application is specifically introduced as follows. Figure 1 The ship operation cable damage degree recognition and early warning method provided by the application is specifically introduced as follows. Figure 1 The ship operation cable damage degree recognition and early warning method provided by the application is specifically introduced as follows.

[0025] Step 101, acquiring cable state data corresponding to the ship operation cable to be identified.

[0026] Here, the cable state data includes cable images and ship environment data, wherein the ship environment data includes illumination data and temperature and humidity data. The cable state data is collected by a high-definition camera, an ambient light sensor and a temperature and humidity detector arranged on the ship body. An image acquisition system is constructed by configuring the high-definition camera, the ambient light sensor and the temperature and humidity detector, images of the ship operation cable to be identified are shot at a rate of more than 30 frames per second, cable changes are dynamically recorded, and illumination intensity and temperature and humidity data are recorded in real time. The image acquisition system has a local cache function to avoid data loss when the network is interrupted, and the collected data is uploaded in real time through the ship internal network or satellite link. In addition, the system also establishes a device state monitoring mechanism to monitor the camera online, storage and network in real time, regularly clean the lens and check the shell sealing to ensure that the equipment can operate stably in the water environment for a long time.

[0027] In addition, in order to ensure data quality, the collected data needs to be preprocessed before uploading, for example, the cable images are filtered (Gaussian or median filtering) to remove random noise, and then the dynamic range is optimized by local contrast enhancement or histogram equalization to strengthen the cable texture and damage features. Canny edge detection algorithm is used to extract the cable contour, Harris corner point detection is used to locate the joint and wear and other key feature points, and high-purity boundary and feature input data are provided for subsequent model loss recognition.

[0028] Step 102, calling a cable damage degree recognition model to recognize the damage of the cable state data, and obtaining the state type of the damage degree of the ship operation cable to be identified.

[0029] Here, first of all, a model specially for cable damage identification needs to be customized, and in the embodiment of the application, a YOLOv8 model is taken as a basic model, and the basic model is improved to obtain a cable damage degree identification model. The cable damage degree identification model is a YOLOv8 model with a large selection kernel network in the backbone network, and the backbone network is embedded with a coordinate attention network. Here, the C2f module in the backbone network of the YOLOv8 model is replaced with a large selection kernel network (LSKNet), and the attention mechanism module in the backbone network is improved, and a coordinate attention network (Coordinate Attention) is embedded as a new attention mechanism module, so that the improved YOLOv8 model is taken as the cable damage degree identification model.

[0030] In a possible implementation, the cable damage degree identification model is constructed by updating the C2f module with a set number of layers in the backbone network of the YOLOv8 model to a large selection kernel network, and embedding a coordinate attention network before a spatial pyramid pooling module in the backbone network to obtain the cable damage degree identification model.

[0031] Specifically, the set number of layers in the backbone network can be the 6th layer and the 8th layer, that is, the C2f module in the 6th layer and the 8th layer of the backbone network of the YOLOv8 model is replaced with a large selection kernel network, and the large selection kernel network convolution kernel scale is designed as a multi-scale combination of “3×3+5×5+7×7”, and the dynamic receptive field range is set to 128×128 pixels, so as to ensure that the overall structure context of the cable and the damage features distributed in different positions are effectively captured without excessive increase in computational burden. The large selection kernel network can realize a dynamic receptive field adjustment mechanism, which can adaptively adjust the convolution kernel scale according to the spatial distribution of the cable in the image, so as to capture the overall structure features of the long-distance cable and focus on the local small damage area. Embedding the coordinate attention network before the spatial pyramid pooling module in the backbone network can convert the image spatial coordinate information into a four-dimensional tensor mapping, and combined with the channel attention weighting strategy, the position of the cable in the complex background can be accurately located, and the feature weight of the damage area can be dynamically improved.

[0032] YOLOv8 model backbone network, can take into account the overall structure features and local micro features in the cable image, while reducing the parameter quantity and calculation quantity of the deep network, realizing the balance of model lightweight and feature extraction capability, and laying a high-efficiency feature foundation for subsequent damage identification. Through embedding the coordinate attention network, the background noise can be effectively suppressed, so that the model focuses on the key area in the cable image, and the attention is concentrated on the key features such as wear marks and broken fibers on the cable surface, which significantly improves the identification sensitivity of small target damage and avoids the problems of missed detection and false detection caused by background interference.

[0033] In a possible implementation, the training process of the cable damage degree recognition model can be implemented in the following manner, which is described in detail below.

[0034] An image sample of the ship operation cable is constructed, and the image sample is labeled with a state label of a cable health state and a damage area bounding box. The state label corresponds one-to-one to a state type of the ship operation cable damage degree to be identified.

[0035] The image sample of the ship operation cable can be collected by a high-definition camera arranged on the ship body. The exposure time and white balance are dynamically adjusted according to the environmental light during collection. In a low-light scene, the brightness is improved through precise gain compensation and overexposure is avoided to ensure that the image is clear and the color is true under different light.

[0036] The collected cable image still needs to be preprocessed, for example, random noise is removed through filtering processing (Gaussian or median filtering), and then the dynamic range is optimized through local contrast enhancement or histogram equalization to strengthen the cable texture and damage features. Combined with the Canny edge detection algorithm, the cable contour and key feature points such as the joint and wear are extracted. The preprocessed cable image can be used as the image sample of the ship operation cable, and some image samples are shown as follows. Figure 2

[0037] In addition, the image sample is labeled with a state label of a cable health state and a damage area bounding box. The damage area bounding box is used to label the area where the cable damage exists in the image sample, and the state label corresponds one-to-one to the state type of the ship operation cable damage degree to be identified, and is used to divide the state type of the ship operation cable damage.

[0038] In a possible implementation, the state label can be determined in the following manner, which is described in detail below.

[0039] ​Firstly, the state label is divided according to the typical damage form of the ship cable in the water operation environment and the industry safety evaluation standard, including four types of intact, slight wear, serious wear and breakage. Here, the local wear condition and the local fiber broken strand condition of the ship operation cable in the image sample are detected to determine the state label of the image sample. The specific determination method can be manually annotated and uploaded, which can be directly obtained, or it can be determined by calling image algorithm to compare and determine.

[0040] When the ship operation cable in the image sample does not exist local wear and does not exist local fiber broken strand, the state label of the image sample is determined as intact type. Here, the normal cable image can be taken as a reference, and then compared with the damage area annotation box part in the image sample, such as judging similar to the normal cable image, wherein there is no local wear and no local fiber broken strand, the state label of the image sample can be determined as intact.

[0041] When the ship operation cable in the image sample exists local wear and the area proportion of local wear does not exceed the preset proportion threshold, the state label of the image sample is determined as slight wear type. Here, by comparing the damage area annotation box part with the normal cable image, if local wear is found, the area of local wear can be calculated in the damage area annotation box of the image sample. When the area proportion does not exceed the proportion threshold of the area of the damage area annotation box, it is indicated that the ship operation cable has no obvious broken slight damage, and the state label of the image sample can be determined as slight wear.

[0042] For example, if the area of local wear is calculated to be less than 10% of the area of the damage area annotation box in the image sample, the state label can be determined as slight wear.

[0043] When the ship operation cable in the image sample exists local wear and the area proportion of local wear has exceeded the proportion threshold or exists local fiber broken strand but the number of broken strands does not exceed the preset number threshold, the state label of the image sample is determined as serious wear.

[0044] Here, when calculating the area proportion of local wear, the number of broken strands of the local fiber broken strand cable can also be counted. If the area proportion of local wear has exceeded the proportion threshold or the number of broken strands does not exceed the preset number threshold, it is indicated that the ship operation cable may have appeared the moderate damage of surface separation, and there may be the risk of breakage, and the state label of the image sample can be determined as serious wear.

[0045] For example, if the area of local wear is calculated to be more than 10% of the area of the damage area annotation box in the image sample, or the number of local fiber broken strand cables is counted to be less than 3, the state label can be determined as serious wear.

[0046] When the number of broken fibers of the local fiber broken rope in the image sample exceeds the number threshold or a breakage exists, the state label of the image sample is determined as broken.

[0047] Here, when counting the number of broken fibers of the local fiber broken rope, if the number of broken fibers exceeds the number threshold or a breakage exists, that is, the number of broken fibers exceeds 3 or a breakage has occurred, it indicates that the rope is currently in a serious failure state of single-strand complete breakage, multi-strand breakage or overall breakage, and the state label is determined as broken.

[0048] In the embodiment of the application, for each image sample of the ship operation rope, the state label of the rope health state is determined by calculating the local wear area and counting the number of local fiber broken ropes, which can provide clear quantitative judgment basis for the state type of the to-be-identified ship operation rope damage degree, and meets the actual engineering actual needs.

[0049] Further, the image sample is input into the rope damage degree recognition model for forward propagation to obtain a damage area prediction box corresponding to the ship operation rope.

[0050] The rope damage degree recognition model can regress and predict the damage area of the rope in the image sample, so as to label the corresponding damage area prediction box.

[0051] Next, a WIoU v3 loss function is constructed according to the damage area labeling box and the damage area labeling box. Here, the WIoU v3 loss function replaces the CIoU loss function of the original YOLOv8 model. Compared with the CIoU loss function which only considers the optimization logic of the boundary box overlap degree, the center point distance and the width-height ratio, the WIoU v3 loss function adds a curvature compensation factor and a topological structure constraint term. On the one hand, the curvature compensation factor adapts to the long strip shape of the rope, avoiding damage area misjudgment caused by insufficient boundary box fitting degree. On the other hand, the topological structure constraint term strengthens the fitting accuracy of the damage area edge contour, especially suitable for boundary box regression of irregular shape damage such as tearing and wear.

[0052] Finally, based on the WIoUv3 loss function, the rope damage degree recognition model is back propagated to update the model parameters of the rope damage degree recognition model.

[0053] In each iteration of the back propagation process, the gradient of the model is calculated by the WIoUv3 loss function, and the gradient value is optimized by calling the gradient optimization algorithm, thereby updating the model parameters of the rope damage degree recognition model.

[0054] Specifically, the network is trained using the AdamW optimizer during backpropagation, the model parameter category is 4, the iteration number is 300, the batch size is 32, the image size of the input ship operation cable image sample is 840*840, the initial momentum is matched with the weight decay, and the initial learning rate is 0.01.

[0055] During the training process, the randomized strategy is used to divide the data set of the ship operation cable image sample, and the expanded image data is allocated according to the proportion of 70% for the training set, 15% for the validation set, and 15% for the test set, so as to ensure the consistent distribution of each subset data and avoid artificial bias, so as to meet the needs of model training, parameter optimization and performance evaluation.

[0056] Before formal model training, pre-training is performed to provide optimal parameters for the self-made data set for formal training, including training batch size, initial learning rate Ir0, and final learning rate Irf. In order to exclude the influence of each improved module, the original YOLOv8 model without improvement is selected as the experimental model, the self-made data set is selected as the data set, and the training round is 25 rounds.

[0057] The principle of controlling variables is adopted in the experiment. Considering the mutual influence relationship among the training batch size, the initial learning rate and the final learning rate, the training batch size is used as the only variable in the first stage, the initial learning rate and the final learning rate and other key hyperparameters are fixed, and through multiple comparison experiments (such as testing the convergence speed and training stability of the model under different batch sizes), the optimal training batch size suitable for the cable damage detection task is determined. In the second stage, based on the optimal batch size determined in the first stage, the initial learning rate is set as the variable, the final learning rate and other parameters are kept unchanged, the model loss function decay trend and the validation set accuracy change are monitored, and the initial learning rate that makes the model quickly enter the effective training interval is selected. In the third stage, the optimal batch size and the initial learning rate determined in the first two stages are fixed, only the final learning rate is used as the variable, the influence of different final learning rates on the accuracy improvement (such as mAP@50 and recall rate) of the model in the later stage is verified, and the optimal combination of the three is finally determined, which provides a scientific hyperparameter basis for subsequent formal model training.

[0058] Further, according to the pre-training result, the parameter settings of the cable damage degree recognition model during training are as follows: training batch size batch size=32, initial learning rate Ir0=0.0006, and final learning rate Irf=0.01.

[0059] After 300 rounds of training, the recall rate (precision) and average precision (mAP@50) of the cable damage degree recognition model for ship cable health state detection reached 93.3% and 94.9%, respectively, which increased by 12.4% and 5.2% compared with the original YOLOv8 baseline model. The improved model reduced the calculation parameter amount and GFLOPs by 6.08% and 9.88% compared with the original YOLOv8 baseline model, effectively reducing the calculation cost.

[0060] In addition, the ablation study of the cable damage degree recognition model is also carried out in the embodiment of the application. Specifically, an iterative enhancement strategy is adopted. First, a CA attention mechanism is added before the backbone network SPPF module, then the deep C2f module (6th and 8th layers) of the backbone network is replaced with an LSKNet module, and finally a Wiou v3 loss function is used to calculate the loss of boundary box regression. Under the condition of constant parameters in the control group, comparative experiments are carried out, and the experimental result data are as follows: Table 1:

[0061] According to the experimental results in Table 1 above, under the synergistic effect of the three improvement strategies, the model mAP@50 accumulatively improves by 5.2%, the precision rate increases by 3.9%, and the recall rate increases by 2.1%, and the calculation amount decreases from 8.1 GFLOPS to 7.0 GFLOPS, a decrease of 13.6%. The recognition ability of the model for micro-damage features is significantly enhanced, and the synergistic optimization of the three-stage core module produces a cumulative gain effect, which also confirms the effectiveness of the cable damage degree recognition model architecture design in the embodiment of the application. The model has made breakthrough progress in three dimensions of detection sensitivity, algorithm robustness and operation efficiency in cable damage detection.

[0062] In the embodiment of the application, the Wiou v3 loss function is constructed to replace the original YOLOv8 model damage function, thereby constructing the cable damage degree recognition model for recognizing cable damage. In the recognition process, the regression precision of the labeled boundary box can be effectively improved, providing support for accurate classification of damage type and degree.

[0063] After the model training is completed, the trained cable damage degree recognition model can be called to recognize the cable state data, and the state type of the damage degree of the to-be-recognized ship operation cable is obtained.

[0064] In one possible implementation, the cable damage degree recognition model is called to recognize the cable state data, and the state type of the damage degree of the to-be-identified ship operation cable is obtained, which can be realized in the following manner, which will be described in detail below.

[0065] First, image features and environmental features of the cable status data are extracted, and then the image features and environmental features are fused to obtain fused features.

[0066] like Figure 3 As shown, when identifying the degree of damage to ship operation cables, multi-source data feature extraction is first performed on the cable status data. Here, features are extracted for different types of data. Image feature extraction is performed on the pre-processed cable images, while for the ship environment data, including illumination and temperature / humidity data, corresponding illumination features and temperature / humidity sequence features are extracted after appropriate preprocessing. Furthermore, the extracted features are fused. Here, an attention mechanism can be used to fuse image feature extraction, illumination features, and temperature / humidity sequence features for intelligent analysis.

[0067] Next, the cable damage identification model is called to identify damage from the fused features, resulting in the damage area bounding box corresponding to the ship's operating cable.

[0068] During intelligent analysis, the fused features are input into the cable damage identification model for regression prediction. The model outputs bounding boxes of the damaged areas corresponding to the ship's operational cables, labeled in each cable image. See the example below. Figure 4 , Figure 4 The results of the cable damage identification model on the damage level identification of some cable images are shown.

[0069] Finally, a damage severity score is calculated for the regression box of the damaged area, and the state type of the damage severity of the ship's operating cables to be identified is determined based on the damage severity score.

[0070] The calculation of the damage severity score here requires combining illumination features and temperature and humidity sequence features. When the model identifies the bounding box of the damaged area, it will simultaneously output a probability value for image recognition. Here, a multilayer perceptron or a fully connected neural network can be used to map the illumination features and temperature and humidity sequence features into a probability value respectively. Then, according to the importance or reliability of each data, the weight coefficients corresponding to the image, illumination, temperature and humidity are set. The probability values ​​of the three features are weighted by the weight coefficients respectively, and the weighted results are summed to obtain the damage severity score, which is used to comprehensively assess the health status of the cable.

[0071] Further, a preset score section can be obtained, each score section corresponding to a state type of the damage degree of the working cable of the to-be-identified ship. Then, the calculated damage degree score is compared with the score section, so as to determine in which score section the damage degree score falls, thereby judging the state type of the damage degree of the working cable of the to-be-identified ship. The state type corresponds to the state label in the training of the cable damage degree identification model, and therefore the state type includes: an intact type, a slight wear type, a severe wear type, and a broken type.

[0072] In step 103, a warning level of the damage of the working cable of the ship is determined based on the state type, and a warning is performed according to a warning strategy corresponding to the warning level.

[0073] As shown in FIG. 10, when the intelligent analysis determines the state type of the damage degree of the working cable of the to-be-identified ship, it can be judged whether it is abnormal according to the state type. If it is normal, no warning is selected and monitoring is continued. If it is abnormal, a corresponding warning mechanism is triggered, and then a corresponding warning strategy is executed to formulate a response measure. Figure 3

[0074] Specifically, first, the warning level of the damage of the working cable of the ship is determined based on the state type. Based on the difference of the state type, the warning level that needs to be warned can be determined. Each warning level is set with a corresponding warning strategy, and finally the warning is triggered and implemented according to the warning strategy.

[0075] In one possible implementation, the warning level of the damage of the working cable of the ship based on the state type can be determined by the following method, which is described in detail below.

[0076] When the state type is the intact type, the warning level of the damage of the working cable of the ship is determined as no warning.

[0077] Here, the state type is the intact type, which means that the health status of the working cable of the ship is relatively good and no obvious damage phenomenon occurs, so no warning is needed and monitoring is continued.

[0078] When the state type is the slight wear type, the warning level of the damage of the working cable of the ship is determined as a first-level warning, and the first-level warning is used to trigger manual review.

[0079] The state type is the slight wear type, which means that the working cable of the ship is in a relatively safe service stage at this time, and only has potential risk signs such as slight surface wear (such as wear area ratio and depth not reaching the critical value affecting the structural safety), which does not affect the normal mooring function. However, it still needs to be reviewed and confirmed, so the warning level is set as the first-level warning. The first-level warning is mainly used to alert manual re-inspection of the cable.

[0080] ​When the state type is the serious wear type, the early warning level of the ship operation cable damage is determined as a second-level early warning, and the second-level early warning is used to trigger the suspension of the ship operation.

[0081] Here, if the state type is identified as the serious wear type, it indicates that the ship operation cable has a serious wear defect at this time and may have local fiber breakage, surface shedding and other damages, which have affected the normal mooring function, and the ship operation needs to be suspended. At this time, the early warning level is set to the second-level early warning, and the second-level early warning is mainly used to trigger the suspension of the ship operation.

[0082] When the state type is the breakage type, the early warning level of the ship operation cable damage is determined as a third-level early warning, and the third-level early warning is used to trigger the ship emergency braking.

[0083] When the state type is identified as the breakage type, it indicates that the cable is in a "broken" state, including partial breakage or complete breakage, at this time the ship is already in a dangerous environment, and not only the ship operation needs to be suspended, but also the ship movement needs to be suspended. At this time, the early warning level is set to the third-level early warning, and the third-level early warning belongs to the highest level of early warning, and is mainly used to trigger the ship emergency braking.

[0084] In the embodiment of the application, the early warning level of the ship operation cable damage is determined through different state types, the ship operation cable damage degree identification is strongly bound with the early warning response mechanism, the closed-loop processing from abnormal identification to early warning research and judgment is realized, the real-time nature of the ship operation cable damage degree identification is also ensured, the timeliness of the early warning is ensured, and the safety of the ship operation is improved.

[0085] In a possible implementation manner, the early warning can be performed according to the early warning strategy corresponding to the early warning level, which can be implemented in the following manner, which will be specifically described below.

[0086] When the early warning level is a first-level early warning, the abnormal information of the to-be-identified ship operation cable is triggered to be recorded and an artificial review warning information is sent.

[0087] The first-level early warning indicates that the current ship operation danger level is low, although the abnormality occurs, but the normal mooring function is not affected, but the abnormal information of the to-be-identified ship operation cable needs to be recorded, specifically, the key information involved in this abnormality, including damage degree identification time, cable position, wear area image, environment data and the like, and an artificial review warning information is sent through a corresponding visual interface or an external device, an artificial review prompt is pushed to the ship personnel, and the personnel are guided to further confirm the wear trend in combination with the on-site inspection, so that the potential risk accumulation is avoided.

[0088] When the early warning level is a second-level early warning, the ship alarm is triggered to alarm, and a ship operation suspension instruction is sent.

[0089] When the early warning level is the second level, it means that the current ship operation is at a high risk, and the normal mooring function has been affected, and the ship operation needs to be stopped immediately. Here, the corresponding ship alarm can be directly triggered to realize the alarm, and the instruction to suspend the ship operation is sent. The ship alarm here can be a terminal alarm of the wharf monitoring center, a warning light in the ship's cabin, etc., and the instruction to suspend the ship operation can be directly pushed to the ship personnel, for example, through a walkie-talkie or a mobile terminal of the ship personnel, the high-risk operation related to the cable (such as cargo loading and unloading, ship displacement, etc.) can be suspended to prevent the damage from being aggravated and causing a safety accident.

[0090] When the early warning level is the third level, the ship alarm is triggered to realize the alarm, the instruction to suspend the ship operation is sent, and the ship emergency braking signal is triggered.

[0091] When the early warning level is the third level, it means that the risk coefficient of the current ship operation has reached the highest level, and not only the ship operation needs to be stopped immediately, but also the ship emergency braking needs to be executed to limit the displacement of the ship, in addition, the ship collision avoidance system needs to be activated to reduce the risk of the ship out of control due to the cable breakage. Here, on the basis of triggering the corresponding ship alarm to realize the alarm and sending the instruction to suspend the ship operation, the ship emergency braking signal needs to be triggered to limit the displacement of the ship, and the ship collision avoidance system needs to be activated to gain time for emergency rescue.

[0092] Of course, in the specific implementation, a dynamic threshold adjustment algorithm can be used, and when determining the early warning level, the alarm strategy is not immediately executed, but the triggering condition is updated adaptively according to the real-time environmental factors (such as wind and wave) and the aging degree of the cable, the early warning false alarm rate is strictly controlled below 3% through multi-modal data cross-validation, and the accuracy and timeliness of the early warning are ensured.

[0093] In addition, the early warning condition for executing the early warning strategy can be defined, and the common cable defect types need to be defined, such as typical damage such as wear, breakage or corrosion. According to the severity of each defect, the corresponding triggering condition is set, for example, the state type is continuously identified as a serious wear type for a plurality of times, and the number of times of identification exceeds a preset threshold, and then the early warning level evaluation is triggered. With the extension of the service time of the cable and the change of the use state, the early warning threshold needs to be reviewed and calibrated regularly to ensure that it reflects the latest cable state and environmental influence.

[0094] The embodiment of the application automatically triggers different early warning strategies to realize ship early warning for different early warning levels, improves the reliability of early warning decision, ensures the reliable operation of the grading early warning mechanism, and provides practical technical support for ship mooring safety monitoring.

[0095] The embodiment of the application realizes direct identification of the ship operation cable body by combining the ship environment data with the cable image itself and using an improved YOLOv8 model. The improved YOLOv8 model uses a large selection kernel network, which can adaptively adjust the convolution kernel scale according to the spatial distribution of the cable in the image, so as to capture the overall structural features of long-distance cables and focus on local small damage areas. The coordinate attention network can accurately locate the position of the cable in the complex background, and focus the attention of loss identification on the key features such as wear marks and broken fibers on the surface of the cable, thereby significantly improving the recognition sensitivity of small target damage. The above improvements reduce the interference of the environmental state on damage identification, and significantly improve the damage identification accuracy of the ship operation cable body. After loss identification, a hierarchical warning is triggered according to the identification result, ensuring the timeliness of the warning response.

[0096] The ship operation cable damage degree identification and warning system provided by the application will be specifically introduced below.

[0097] Figure 5 is a structural schematic diagram of the ship operation cable damage degree identification and warning system provided by the application, as Figure 5 shown, the ship operation cable damage degree identification and warning system specifically comprises: a data acquisition module 501, a damage identification module 502, and a hierarchical warning module 503.

[0098] Specifically, the data acquisition module 501 is configured to acquire cable state data corresponding to a ship operation cable to be identified, wherein the cable state data comprises a cable image and ship environment data; the damage identification module 502 is configured to call a cable damage degree identification model to perform damage identification on the cable state data, and obtain a state type of the damage degree of the ship operation cable to be identified, wherein the cable damage degree identification model is a YOLOv8 model comprising a backbone network including a large selection kernel network, and the backbone network is embedded with a coordinate attention network; and the hierarchical warning module 503 is configured to determine a warning level of the damage of the ship operation cable based on the state type, and perform warning according to a preset warning strategy corresponding to the warning level.

[0099] The damage identification module 502 is further configured to construct the cable damage degree identification model by updating a C2f module with a set number of layers in the backbone network of the YOLOv8 model to a large selection kernel network, and embedding a coordinate attention network before a spatial pyramid pooling module in the backbone network.

[0100] The ship operation cable damage degree identification and early warning system provided by the above embodiments can implement the technical solutions described in the ship operation cable damage degree identification and early warning method embodiments, and the principles of implementation of the above modules or units can be referred to the corresponding content in the ship operation cable damage degree identification and early warning method embodiments. The technical effects can also be mutually corresponding, and will not be repeated here.

[0101] As shown in Figure 6 The present application also provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602, and a display 603. Figure 6 Only part of the components of the electronic device 600 are shown, but it should be understood that all the shown components are not required, and more or less components can be implemented instead.

[0102] The memory 602 can be an internal storage unit of the electronic device 600 in some embodiments, such as a hard disk or memory of the electronic device 600. The memory 602 can also be an external storage device of the electronic device 600 in other embodiments, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.

[0103] Further, the memory 602 can include both an internal storage unit and an external storage device of the electronic device 600. The memory 602 is used to store application software and various data installed in the electronic device 600.

[0104] The processor 601 can be a central processing unit (CPU), a microprocessor or other data processing chip in some embodiments, used to run program codes or process data stored in the memory 602, such as the ship operation cable damage degree identification and early warning method in the present application.

[0105] The display 603 can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. in some embodiments. The display 603 is used to display information of the electronic device 600 and to display visual user interfaces. The components 601-603 of the electronic device 600 communicate with each other through a system bus.

[0106] In some embodiments of the present application, when the processor 601 executes the damage degree identification and early warning program in the memory 602, the following steps can be implemented: obtaining cable state data corresponding to a ship operation cable to be identified, the cable state data including cable images and ship environment data; calling a cable damage degree identification model to identify the damage of the cable state data, and obtaining a state type of the damage degree of the ship operation cable to be identified, the cable damage degree identification model being a YOLOv8 model with a backbone network including a large selection kernel network, the backbone network being embedded with a coordinate attention network; determining an early warning level of the damage of the ship operation cable based on the state type, and performing early warning according to an early warning strategy corresponding to the early warning level.

[0107] It should be understood that, in addition to the above functions, the processor 601 can also implement other functions when executing the damage degree identification and early warning program in the memory 602. For details, please refer to the description of the corresponding method embodiments above.

[0108] Further, the type of the electronic device 600 referred to in the embodiments of the present application is not limited specifically, and the electronic device 600 can be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of the portable electronic device include but are not limited to a portable electronic device running an IOS, android, microsoft or other operating system. The above-mentioned portable electronic device can also be other portable electronic devices, such as a laptop computer having a touch-sensitive surface (e.g. a touch panel), etc. It should also be understood that, in some other embodiments of the present application, the electronic device 600 can also not be a portable electronic device, but a desktop computer having a touch-sensitive surface (e.g. a touch panel).

[0109] In another aspect, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the ship operation cable damage degree identification and early warning method provided by the above-mentioned methods, the method comprising: obtaining cable state data corresponding to a ship operation cable to be identified, the cable state data including cable images and ship environment data; calling a cable damage degree identification model to identify the damage of the cable state data, and obtaining a state type of the damage degree of the ship operation cable to be identified, the cable damage degree identification model being a YOLOv8 model with a backbone network including a large selection kernel network, the backbone network being embedded with a coordinate attention network; determining an early warning level of the damage of the ship operation cable based on the state type, and performing early warning according to an early warning strategy corresponding to the early warning level.

[0110] Those skilled in the art can understand that all or part of the processes of the above-mentioned embodiment methods can be completed by instructing the relevant hardware by a computer program, and the program can be stored in a computer readable storage medium. The computer readable storage medium is a magnetic disk, an optical disk, a read-only memory, a random access memory, etc.

[0111] The ship operation cable damage degree identification and early warning method and system provided by the present application are described in detail above. The principles and implementation modes of the present application are described by applying specific examples. The above example is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for identifying and warning of damage to ship operation cables, characterized in that, include: Acquire the cable status data corresponding to the operating cable of the vessel to be identified. The cable status data includes cable images and vessel environment data. The cable damage degree identification model is invoked to identify the damage in the cable state data, and the state type of the damage degree of the ship operation cable to be identified is obtained. The cable damage degree identification model is a YOLOv8 model with a backbone network including a large selection kernel network, and the backbone network is embedded with a coordinate attention network. The warning level for damage to ship operation cables is determined based on the state type, and a warning is issued according to the warning strategy corresponding to the warning level.

2. The method for identifying and warning of damage to ship operation cables according to claim 1, characterized in that, The cable damage identification model is constructed in the following manner: The C2f module with a set number of layers in the backbone network of the YOLOv8 model is updated to a large selection kernel network, and a coordinate attention network is embedded before the spatial pyramid pooling module in the backbone network to obtain a cable damage degree recognition model.

3. The method for identifying and warning of damage to ship operation cables according to claim 2, characterized in that, The training process of the cable damage degree recognition model includes: Construct image samples of ship operation cables, wherein the image samples are labeled with status labels of the cable health status and damage area annotation boxes, and the status labels correspond one-to-one with the status type of the damage degree of the ship operation cable to be identified. The image samples are input into the cable damage degree recognition model for forward propagation to obtain the predicted damage area bounding box corresponding to the ship operation cable. Construct the WIoU v3 loss function based on the damage region bounding box and the damage region bounding box; The WIoUv3 loss function is used to backpropagate the cable damage identification model to update the model parameters.

4. The method for identifying and warning of damage to ship operation cables according to claim 3, characterized in that, The status labels include: intact, slightly worn, severely worn, and broken, and the status labels are determined in the following way: When the ship's operating cable in the image sample does not show local wear and no local fiber breakage, the status label of the image sample is determined to be intact; If the ship's operating cable in the image sample has localized wear and the area of ​​localized wear does not exceed a preset threshold, the status label of the image sample is determined to be slightly worn. When the ship's operating cable in the image sample has local wear and the proportion of the local wear area exceeds the proportion threshold, or there is a cable with local fiber breakage but the number of broken strands does not exceed the preset number threshold, the status label of the image sample is determined to be severe wear. When the number of broken strands in a ship's operating cable in the image sample exceeds the threshold or there is a breakage, the status label of the image sample is determined to be broken.

5. The method for identifying and warning of damage to ship operation cables according to claim 1, characterized in that, The method of calling the cable damage degree recognition model to identify damage to the cable state data yields the state type of the damage degree of the vessel's operating cable to be identified, including: Extract image features and environmental features from the cable state data, and fuse the image features and environmental features to obtain fused features; The cable damage identification model is invoked to identify damage to the fused features, thereby obtaining the damage area bounding box corresponding to the ship's operating cable. A damage severity score is calculated for the regression box of the damaged area, and the state type of the damage severity of the ship's operating cable to be identified is determined based on the damage severity score.

6. The method for identifying and warning of damage to ship operation cables according to claim 1, characterized in that, The method for determining the early warning level of ship operation cable damage based on state type includes: When the status type is intact, the warning level for damage to the ship's operating cables is determined to be no warning. When the state type is mild wear, the warning level for damage to the ship's operating cables is determined to be Level 1 warning, which is used to trigger manual review. When the state type is severe wear, the warning level for damage to the ship's operating cables is determined to be a Level 2 warning, which is used to trigger a suspension of ship operations. When the state type is a breakage type, the warning level for damage to the ship's operating cables is determined to be a Level 3 warning, which is used to trigger the ship's emergency braking.

7. The method for identifying and warning of damage to ship operation cables according to claim 1, characterized in that, The step of issuing an early warning based on the early warning strategy corresponding to the early warning level includes: When the warning level is Level 1, the abnormal information of the vessel's operating cable to be identified is recorded and a manual review warning is sent. When the warning level is Level II, the ship alarm is triggered and a command to suspend ship operations is sent. When the warning level is Level 3, the ship alarm is triggered to sound an alarm and a command to suspend ship operations is sent, and the ship's emergency braking signal is triggered.

8. A system for identifying and warning of damage levels to ship operation cables, characterized in that, include: The data acquisition module is used to acquire the cable status data corresponding to the operating cable of the vessel to be identified. The cable status data includes cable images and vessel environmental data. The damage identification module is used to call the cable damage degree identification model to identify the damage of the cable state data and obtain the state type of the damage degree of the ship operation cable to be identified. The cable damage degree identification model is a YOLOv8 model with a backbone network including a large selection kernel network. The backbone network is embedded with a coordinate attention network. The graded early warning module is used to determine the early warning level of damage to the ship's operating cables based on the state type, and to issue an early warning according to the preset early warning strategy corresponding to the early warning level.

9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps of the method for identifying and warning of damage to ship operation cables as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for identifying and warning of damage to ship operation cables as described in any one of claims 1 to 7.

Images