Underwater cable tracking method, electronic equipment and storage medium

Abstract

The invention discloses an underwater cable tracking method, electronic equipment and a storage medium. The method belongs to the field of power systems, and comprises the steps that an underwater video captured by underwater navigation equipment is acquired, and the underwater video is used for representing a video obtained by shooting an initial area where an underwater cable is deployed; the underwater video is recognized based on a cable recognition model, a cable recognition result is obtained, the cable recognition model is deployed in the underwater navigation equipment, and the cable recognition result is used for representing the position of the underwater cable in the initial area; constructing an equipment navigation path based on a cable identification result; and controlling the underwater navigation device to track the underwater cable based on the device navigation path. According to the invention, the technical problem of low recognition accuracy of underwater cables in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and in particular to an underwater cable tracking method, electronic equipment and a storage medium. Background Art

[0002] As key infrastructure connecting land and sea, and between different islands, underwater cables and pipeline systems are vital to modern power, communications, and marine development. They not only undertake the tasks of power transmission and data transmission, but also play an important role in marine scientific research, oil and gas extraction. However, these underwater facilities are exposed to harsh marine environments for a long time and are susceptible to corrosion, wear, biological adhesion, and natural disasters. Therefore, regular inspection and maintenance have become an important part of ensuring their safety and reliability.

[0003] In recent years, with the rapid development of artificial intelligence technology, autonomous underwater vehicles (AUVs) have attracted widespread attention due to their autonomy, flexibility and cost-effectiveness in underwater operations. The visual system and sensors equipped by AUVs can continuously monitor the underwater environment and collect real-time data on cable status, but image processing technology has become a key constraint on their performance. The acquisition of underwater images faces challenges such as insufficient lighting, turbid water, and turbulent interference, which leads to a decrease in image quality, resulting in a low recognition accuracy when using image processing technology to identify underwater cables.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide an underwater cable tracking method, an electronic device, and a storage medium, so as to at least solve the technical problem of low accuracy in identifying underwater cables in the related art.

[0006] According to one aspect of an embodiment of the present invention, there is provided an underwater cable tracking method, comprising: acquiring an underwater video captured by an underwater navigation device, wherein the underwater video is used to represent a video obtained by shooting an initial area where underwater cables are deployed; identifying the underwater video based on a cable recognition model to obtain a cable recognition result, wherein the cable recognition model is deployed in the underwater navigation device, and the cable recognition result is used to represent the position of the underwater cable in the initial area; constructing a device navigation path based on the cable recognition result; and controlling the underwater navigation device to track the underwater cable based on the device navigation path.

[0007] Furthermore, the underwater video is identified based on the cable recognition model to obtain a cable recognition result, including: performing complexity recognition on the underwater video to obtain a complexity index of the initial area, wherein the complexity index is used to characterize the difficulty of identifying the underwater cable from the initial area; based on the complexity index and the video parameters of the underwater video, multiple initial image frames are captured from the underwater video; the multiple initial image frames are standardized to obtain multiple target image frames; and the multiple target image frames are input into the cable recognition model to obtain a cable recognition result.

[0008] Furthermore, complexity identification is performed on the underwater video to obtain a complexity index of the initial area, including: obtaining the area type of the initial area; performing obstacle identification on the underwater video to obtain an obstacle identification result, wherein the obstacle identification result includes the types of obstacles and the number of obstacles existing in the initial area; and determining the complexity index based on the area type, the type of obstacles and the number of obstacles.

[0009] Furthermore, constructing a device navigation path based on the cable identification result includes: determining the integrity of the underwater cables based on the positions of the underwater cables in the initial area, wherein the integrity is used to characterize whether all the underwater cables are in the initial area; in response to the integrity of the underwater cables being that not all the underwater cables are in the initial area, determining the cable direction of the cables based on the cable identification result; determining a target area based on the cable direction, wherein the target area is used to characterize an area outside the initial area that contains the underwater cables; and constructing the device navigation path based on the target area, the initial area and the cable direction.

[0010] Furthermore, the method also includes: obtaining a target video set, wherein the target video set is used to characterize a video set obtained by shooting underwater areas with different complexity indicators, and underwater cables are arranged in the underwater areas; dividing the videos contained in the target video set to obtain training videos, verification videos and test videos, wherein the number of training videos is greater than the number of verification videos, and the number of verification videos is greater than the number of test videos; training an initial recognition model based on the training video to obtain a training recognition model; verifying the training recognition model based on the verification video to obtain a verification recognition model; and testing the verification recognition model based on the test video to obtain a cable recognition model.

[0011] Furthermore, the verification recognition model is tested based on the test video to obtain a cable recognition model, including: inputting the test video into the verification recognition model to obtain a video recognition result; matching the video recognition result with the cable labeling result corresponding to the test video to obtain a video matching result; determining a model evaluation index of the verification recognition model based on the video matching result, wherein the model evaluation index is used to reflect the accuracy of the recognition result of the verification recognition model; in response to the model evaluation index being greater than a preset index threshold, determining that the verification recognition model is a cable recognition model.

[0012] Furthermore, a model evaluation index of the verification recognition model is determined based on the video matching results, including: determining positive samples and negative samples from the video matching results, wherein the positive samples are used to characterize test videos whose matching degree between the video recognition results and the cable labeling results is greater than a first matching degree threshold, and the negative samples are used to characterize test videos whose matching degree between the video recognition results and the cable labeling results is greater than a second matching degree threshold, except for the positive samples, and the first matching degree threshold is greater than the second matching degree threshold; based on the positive samples and the negative samples, determining the precision and recall rate of the verification recognition model; based on the precision and recall rate, determining the model evaluation index of the verification recognition model.

[0013] According to another aspect of an embodiment of the present invention, there is also provided an underwater cable tracking device, including: a video acquisition module, used to acquire underwater video captured by underwater navigation equipment, wherein the underwater video is used to characterize a video obtained by shooting an initial area where underwater cables are deployed; a video recognition module, used to recognize the underwater video based on a cable recognition model to obtain a cable recognition result, wherein the cable recognition model is deployed in the underwater navigation equipment, and the cable recognition result is used to characterize the position of the underwater cable in the initial area; a path construction module, used to construct an equipment navigation path based on the cable recognition result; and an equipment control module, used to control the underwater navigation equipment to track the underwater cable based on the equipment navigation path.

[0014] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0015] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0017] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0018] According to another aspect of the embodiments of the present invention, a computer program is further provided. When the computer program is executed by a processor, the methods in the embodiments of the present invention are implemented.

[0019] In an embodiment of the present invention, the underwater video captured by the underwater navigation equipment is acquired; the underwater video is identified based on a cable recognition model to obtain a cable recognition result; the equipment navigation path is constructed based on the cable recognition result; and the underwater navigation equipment is controlled to track the underwater cable based on the equipment navigation path. The underwater video is identified by the cable recognition model to accurately locate the specific position of the underwater cable, and the tracking path of the underwater navigation equipment is set based on the position to avoid the influence of the complex underwater environment on the underwater cable tracking, so that the underwater navigation equipment can quickly and accurately classify and track the underwater cables, thereby achieving the purpose of improving the underwater cable image classification accuracy and robustness, thereby realizing the technical effect of improving the underwater cable recognition accuracy, and further solving the technical problem of low underwater cable recognition accuracy in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0021] Figure 1 is a flow chart of an underwater cable tracking method according to an embodiment of the present invention;

[0022] Figure 2 is a flow chart of training an optional underwater cable identification model according to an embodiment of the present invention;

[0023] Figure 3 is a flowchart of an optional data collection and data preprocessing process according to an embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of an optional confusion matrix according to an embodiment of the present invention;

[0025] Figure 5 is a schematic diagram of an underwater cable tracking device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] According to an embodiment of the present invention, an embodiment of an underwater cable tracking method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0029] Figure 1 is a flow chart of an underwater cable tracking method according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0030] Step S102, obtaining an underwater video captured by an underwater navigation device, wherein the underwater video is used to represent a video obtained by shooting an initial area where the underwater cables are deployed.

[0031] The underwater navigation equipment may be any equipment and tool that can be operated underwater, for example, an underwater drone, an underwater vehicle, etc., but is not limited thereto. The initial area may be an area where underwater cables are deployed.

[0032] In an optional embodiment, considering that the underwater environment may be more complex, and underwater video can help operators intuitively see the position and status of the cables, which is crucial for determining the exact location of the cables and identifying the types of cables, the underwater cable tracking system (hereinafter referred to as the tracking system) can use underwater navigation equipment to shoot the initial area where the underwater cables are deployed, and return the captured video to the tracking system, so that the tracking system can identify and locate the underwater cables based on the above video.

[0033] For example, in order to accurately obtain video information of the initial area where the underwater cables are deployed, the underwater navigation equipment may be an autonomous underwater vehicle, which may be equipped with a camera and lighting equipment and have sufficient power and video storage space. The tracking system may pre-set the underwater activity area of ​​the autonomous underwater vehicle based on the initial area where the underwater cables are deployed, so that the autonomous underwater vehicle can shoot videos in the above area, and the tracking system may obtain the underwater video captured by the autonomous underwater vehicle in real time through wireless communication or physical connection.

[0034] Step S104, identifying the underwater video based on the cable recognition model to obtain a cable recognition result, wherein the cable recognition model is deployed in the underwater navigation equipment, and the cable recognition result is used to characterize the position of the underwater cable in the initial area.

[0035] The cable identification model may be a machine learning or deep learning model for identifying and classifying different types of cables. For example, the cable identification model may be a deep convolutional neural network model, a support vector machine, etc., but is not limited thereto. The cable identification result may be a result for indicating the specific location of the underwater cable.

[0036] In an optional embodiment, considering that the underwater environment shown in the underwater video may be relatively complex, the use of manual identification to identify underwater cables is not only time-consuming and labor-intensive, but also prone to errors. In order to improve the efficiency and accuracy of underwater cable identification, the tracking system can use the above-mentioned cable identification model to identify the underwater video, thereby obtaining a cable identification result to reduce errors caused by human factors. The above-mentioned cable identification result can indicate the specific location of the underwater cable in the above-mentioned initial area, providing a basis for subsequent tracking work.

[0037] For example, the cable recognition model can be a deep convolutional neural network model. Assuming that the tracking system has completed the training of the deep convolutional neural network model in advance, in order to accurately locate the specific position of the underwater cable, after the tracking system obtains the underwater video taken by the underwater navigation equipment, it can first perform pre-processing operations on the underwater video, including denoising, contrast enhancement, brightness and contrast adjustment, etc., which can improve the visibility of the cable. Then, the tracking system can input the pre-processed underwater video into the trained deep convolutional neural network model, and the model can perform cable recognition on each frame of the underwater video, thereby obtaining a cable recognition result that characterizes the position of the underwater cable in the initial area.

[0038] Step S106: constructing a device navigation path based on the cable identification result.

[0039] The above-mentioned equipment navigation path may be the movement path of the above-mentioned underwater navigation equipment when tracking the underwater cable.

[0040] In an optional embodiment, considering that the underwater environment may interfere with the tracking of underwater cables, for example, there may be obstacles such as rocks and vegetation in the underwater environment. If the underwater navigation equipment does not have a clear navigation path, it may collide with these obstacles, causing damage to the underwater navigation equipment or failure of the underwater cable tracking task, thereby reducing the efficiency of underwater cable tracking. Therefore, the tracking system can accurately plan the navigation path of the underwater navigation equipment based on the cable identification results to help the underwater navigation equipment collect cable data more quickly and accurately, thereby improving the tracking efficiency of underwater cables and reducing underwater operation costs.

[0041] For example, in order to reasonably plan the navigation path of the above-mentioned underwater navigation equipment, thereby improving the efficiency and accuracy of underwater cable tracking, the tracking system can analyze the underwater environment based on the above-mentioned underwater video, so as to determine the factors affecting the navigation path including underwater obstacles, and then set the navigation path of the underwater navigation equipment based on the above-mentioned cable identification results and the factors affecting the navigation path using the path planning algorithm.

[0042] Step S108, controlling the underwater navigation device to track the underwater cable based on the device navigation path.

[0043] In an optional embodiment, considering that the navigation path of the above-mentioned equipment is formulated based on the specific position of the underwater cable, by precisely controlling the navigation path of the underwater navigation equipment, the interference of the underwater environment on the tracking of the underwater cable can be avoided, thereby improving the efficiency and accuracy of tracking the underwater cable. Therefore, the tracking system can control the underwater navigation equipment to track the underwater cable based on the navigation path of the equipment.

[0044] For example, the staff can equip the above-mentioned underwater navigation equipment with precise positioning control devices in advance, so that during underwater operations, the tracking system can monitor the position of the underwater navigation equipment in real time. If the heading trajectory of the underwater navigation equipment is found to deviate, the tracking system can sense it in time and correct the heading of the underwater navigation equipment, thereby improving the efficiency and accuracy of underwater cable tracking.

[0045] In an embodiment of the present invention, the underwater video captured by the underwater navigation equipment is acquired; the underwater video is identified based on a cable recognition model to obtain a cable recognition result; the equipment navigation path is constructed based on the cable recognition result; and the underwater navigation equipment is controlled to track the underwater cable based on the equipment navigation path. The underwater video is identified by the cable recognition model to accurately locate the specific position of the underwater cable, and the tracking path of the underwater navigation equipment is set based on the position to avoid the influence of the complex underwater environment on the underwater cable tracking, so that the underwater navigation equipment can quickly and accurately classify and track the underwater cables, thereby achieving the purpose of improving the underwater cable image classification accuracy and robustness, thereby realizing the technical effect of improving the underwater cable recognition accuracy, and further solving the technical problem of low underwater cable recognition accuracy in the related technology.

[0046] Furthermore, the underwater video is identified based on the cable recognition model to obtain a cable recognition result, including: performing complexity recognition on the underwater video to obtain a complexity index of the initial area, wherein the complexity index is used to characterize the difficulty of identifying the underwater cable from the initial area; based on the complexity index and the video parameters of the underwater video, multiple initial image frames are captured from the underwater video; the multiple initial image frames are standardized to obtain multiple target image frames; and the multiple target image frames are input into the cable recognition model to obtain a cable recognition result.

[0047] The complexity index may be an index indicating the difficulty of identifying underwater cables from the initial area. For example, the complexity index may include at least one or more of the following: lighting conditions, water turbidity, target occlusion, etc., but not limited thereto. The video parameters may be parameters used to describe underwater video characteristics. For example, the video parameters may include at least one or more of the following: video duration, video encoding format, color space, etc., but not limited thereto. The initial image frame may be a batch of image frames captured from an underwater video. The standardization process may be converting the initial image frame to a uniform scale to improve the efficiency and accuracy of subsequent processing. For example, the size may be adjusted to 75×75 pixels. The target image frame may refer to an image frame that has been standardized, and the target image frame has been converted into a format suitable for input into a cable recognition model.

[0048] In an optional embodiment, the complexity of the underwater environment may seriously affect the identification of cables. For example, underwater light conditions, water flow disturbances, suspended objects, obstacles and other factors can increase the difficulty of cable identification. The tracking system can obtain the complexity index of the above-mentioned initial area by performing complexity identification on the underwater video. The index can quantify the difficulty of underwater cable identification, and the tracking system can perform subsequent processing according to the above-mentioned complexity index. Further considering that only some image frames in the underwater video may have captured underwater cables, or only some image frames have environmental conditions that are more suitable for underwater cable identification. The tracking system can intercept multiple initial image frames based on the above-mentioned complexity index and video parameters, thereby ensuring that the initial image frames used for identification have good cable identification conditions, while avoiding processing a large amount of useless or repeated information, thereby effectively reducing the amount of calculation and improving the processing speed. In order to further improve the accuracy of underwater cable identification, the tracking system can also perform standardization processing on the above-mentioned multiple initial image frames to obtain the above-mentioned multiple target image frames to ensure that the image data input into the cable identification model has consistency in parameters, thereby reducing the uncertainty of the model when processing images. At this time, the tracking system can input the above-mentioned multiple target image frames into the cable recognition model. Since the above-mentioned multiple target image frames are preprocessed and standardized images, factors that may interfere with underwater cable recognition are removed, so the cable recognition model can obtain the above-mentioned cable recognition results more accurately.

[0049] For example, the cable recognition model can be a deep convolutional neural network model. Assuming that the underwater environment is a complex and changeable underwater environment, in order to determine the difficulty of underwater cable recognition in the complex underwater environment, the tracking system can perform complexity recognition on the underwater video based on a pre-constructed deep learning algorithm. The algorithm can consider factors such as lighting conditions, water turbidity, environmental complexity (such as the distribution of underwater vegetation, sediments, and rocks), and possible obstructions (such as underwater organisms and underwater structures), and calculate the complexity index of the initial area. The index can be a value between 0 and 1. The larger the value, the higher the complexity and the more difficult it is to identify underwater cables. Since there are many obstacles in the underwater environment, the value of the index is large, and the difficulty of identifying underwater cables is high. After obtaining the complexity index, the device system can divide the underwater video duration according to the fixed number of image frames according to the complexity index and parameters such as the video duration of the underwater video, so as to intercept multiple initial image frames from the image frames of the underwater video, thereby ensuring the accuracy of underwater cable recognition in a complex underwater environment. Then, the tracking system can adjust the multiple initial image frames to a size of 75×75 pixels and set the image frames to RGB color mode, thereby obtaining the multiple target image frames to enhance the visibility of underwater cables in the image. After completing the above standardization process, the tracking system can input the target image frames into the above deep convolutional neural network model, which will identify the cables in each frame and output the cable identification results.

[0050] For another example, suppose that the underwater environment is a swimming pool without obstacles. In this environment, the underwater navigation equipment can clearly capture the underwater cables. The tracking system can perform complexity recognition on the underwater video based on the pre-constructed deep learning algorithm, so as to calculate the complexity index. Since the underwater environment of the swimming pool is relatively simple and the visibility of the cables is relatively high, after obtaining the complexity index, the device system can intercept fewer initial image frames from the image frames of the underwater video at a pre-set time interval according to the complexity index and parameters such as the video length of the underwater video, so as to improve the efficiency of underwater cable recognition in the swimming pool environment. Then, the tracking system can adjust the initial image frame to a size of 75×75 pixels and set the image frame to RGB color mode, so as to obtain the target image frame. After completing the standardization process, the tracking system can input the target image frame into the deep convolutional neural network model, which will identify the cables in each frame and output the cable recognition result.

[0051] It should be noted that the specific values ​​such as the pixel size after adjustment of the above-mentioned initial image frame are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0052] Furthermore, complexity identification is performed on the underwater video to obtain a complexity index of the initial area, including: obtaining the area type of the initial area; performing obstacle identification on the underwater video to obtain an obstacle identification result, wherein the obstacle identification result includes the types of obstacles and the number of obstacles existing in the initial area; and determining the complexity index based on the area type, the type of obstacles and the number of obstacles.

[0053] The above-mentioned area type may be an underwater environment type corresponding to the above-mentioned initial area. The above-mentioned obstacle recognition result may be the type and number of obstacles existing in the initial area obtained by recognizing the above-mentioned underwater video.

[0054] In an optional embodiment, considering that different underwater environments may affect the underwater cable identification process, the tracking system may first obtain the area type of the initial area, and adjust the processing strategy according to the above area type to improve the underwater cable identification accuracy under different area types. Further considering that underwater obstacles may block underwater cables and cause visual interference, the tracking system may further perform obstacle identification on underwater videos to evaluate the impact of different types and numbers of underwater obstacles on the underwater cable identification task. The tracking system may comprehensively consider the above area types, obstacle types and number of obstacles to obtain a complexity index that reflects the overall difficulty of underwater cable identification, thereby providing a reference for subsequent underwater cable identification.

[0055] For example, the underwater area types included in the underwater video may include open waters, underwater vegetation areas, underwater sediment areas, and underwater structure areas. Specifically, for open waters, the color distribution in the underwater video image may be relatively uniform. For underwater vegetation areas, underwater video images may contain a large number of green or brown areas. For underwater sediment areas, underwater video images may show large areas of gray or brown. For underwater structure areas, underwater video images may contain more edges and textures. The tracking system can identify the area type of the initial area based on feature extraction and clustering algorithms, and by analyzing and comparing the features of the above different area types, it can be determined that the type of the above initial area is an underwater sediment area. In order to accurately obtain the type and number of underwater obstacles, the tracking system can use an obstacle recognition model based on deep learning, such as the Mask R-CNN model (Mask Region-based Convolutional Neural Network, region-based mask convolutional neural network model), to analyze each frame in the video, thereby outputting the type and number of obstacles existing in the above initial area. Finally, the tracking system can analyze the above complexity indicators based on the above area types, obstacle types and number of obstacles.

[0056] Furthermore, constructing a device navigation path based on the cable identification result includes: determining the integrity of the underwater cables based on the positions of the underwater cables in the initial area, wherein the integrity is used to characterize whether all the underwater cables are in the initial area; in response to the integrity of the underwater cables being that not all the underwater cables are in the initial area, determining the cable direction of the cables based on the cable identification result; determining a target area based on the cable direction, wherein the target area is used to characterize an area outside the initial area that contains the underwater cables; and constructing the device navigation path based on the target area, the initial area and the cable direction.

[0057] The integrity may be an indicator for evaluating whether the underwater cable is in the current "initial area". The cable direction may refer to the direction in which the cable extends in the underwater environment. The target area may refer to the area that the underwater navigation equipment needs to explore to detect the intact cable.

[0058] In an optional embodiment, considering that the camera or sensor of the underwater navigation equipment has a certain field of view, if one end or part of the cable exceeds this range, the complete cable information cannot be obtained. Therefore, the tracking system can first determine the integrity of the underwater cable based on the position of the underwater cable in the initial area, so as to help identify whether the position or direction of the underwater navigation equipment needs to be adjusted to capture the complete part of the underwater cable. When the above integrity shows that the underwater cable does not appear completely in the initial area, the tracking system can determine the extension direction of the underwater cable based on the cable identification result to guide the next action of the underwater navigation equipment, that is, in which direction the underwater navigation equipment moves to track the other parts of the cable. After determining the extension direction of the underwater cable, the tracking system can calculate the area where the underwater cable may extend, that is, the above target area, to ensure that the underwater cable can be completely tracked. Finally, the tracking system can comprehensively consider the target area, the initial area and the direction of the underwater cable, and construct the navigation path of the underwater navigation equipment to continue tracking the cable. At the same time, it can avoid unnecessary detours, save time and energy, and improve the tracking efficiency of the underwater cable.

[0059] For example, it can be assumed that the tracking system has acquired a set of underwater videos captured in real time by the underwater navigation equipment. In order to track the underwater cable more completely, the tracking system can use the underwater video to determine the integrity of the underwater cable based on the position of the underwater cable in the initial area. The integrity indicates that the underwater cable is interrupted at the right boundary of the underwater video image and does not appear completely in the initial area. At this time, the tracking system can determine that the underwater cable will extend to the upper right of the current initial area based on the current position of the cable and the total length of the cable, and determine the area located at the upper right of the current initial area as the target area. Then, the tracking system can plan the navigation path of the underwater navigation equipment in combination with the target area, the initial area and the cable direction. The path can guide the underwater navigation equipment to track along the cable direction until the entire length of the cable is detected.

[0060] Furthermore, the method also includes: obtaining a target video set, wherein the target video set is used to characterize a video set obtained by shooting underwater areas with different complexity indicators, and underwater cables are arranged in the underwater areas; dividing the videos contained in the target video set to obtain training videos, verification videos and test videos, wherein the number of training videos is greater than the number of verification videos, and the number of verification videos is greater than the number of test videos; training an initial recognition model based on the training video to obtain a training recognition model; verifying the training recognition model based on the verification video to obtain a verification recognition model; and testing the verification recognition model based on the test video to obtain a cable recognition model.

[0061] The target video set may refer to a series of videos shot in underwater areas with different complexity indicators. The training videos may be the majority of videos divided from the target video set, which are used to train the cable recognition model. The verification videos may be a small portion of videos divided from the target video set, where the number of verification videos is less than the training videos, and are used to verify the performance of the model during the training process. The test videos may be a smaller portion of videos divided from the target video set, where the number of test videos is less than the verification videos, and are used to ultimately evaluate the performance of the cable recognition model. The initial recognition model may be an initially constructed underwater cable recognition model, which has a low recognition accuracy and requires further model training. The training recognition model may be a model obtained by training the initial recognition model based on the training video. The verification recognition model may be a model obtained by verifying the training recognition model based on the verification video.

[0062] In an optional embodiment, considering that the recognition accuracy of the initial recognition model constructed by the tracking system is low, in order to avoid the above initial recognition model from overfitting to a specific environment or condition, so as to show better generalization ability in the actual underwater environment, the tracking system can first obtain the above target video set, which contains videos shot of underwater areas with different complexity indicators, which means that the above initial recognition model will have the opportunity to learn the underwater cable characteristics under various underwater conditions. In order to improve the training effect of the above initial recognition model, the tracking system can divide the above target video set into three categories: training videos, verification videos and test videos. Among them, the number of the above training videos is relatively large, which is mainly used to train the model to learn the characteristics and patterns of cables. A large amount of training data can help the model extract richer information and improve recognition accuracy. The tracking system trains the initial recognition model based on the training video to obtain a training recognition model. The number of the above-mentioned verification videos is less than the number of the above-mentioned training videos, and is used to tune the model parameters during the training process, monitor overfitting problems, and ensure that the model also performs well on unseen data. Through the feedback of the verification set, the complexity of the model can be adjusted and the hyperparameters can be optimized. The tracking system verifies the trained recognition model based on the verification video to obtain a verified recognition model. The number of the above-mentioned test videos is small and less than the number of the above-mentioned verification videos. The test video is a data set used to evaluate the final performance of the above-mentioned verification recognition model after the initial recognition model training and verification process are completely completed. The tracking system tests the verification recognition model based on the test video to obtain a cable recognition model.

[0063] For example, it can be assumed that the tracking system has built an initial recognition model for underwater cable identification based on a deep convolutional neural network. The tracking system can collect 1,000 underwater videos from underwater cable detection projects in various places. Each of the underwater videos is shot in underwater areas with different complexity indicators, and each video contains images of underwater cables. The tracking system can further randomly select 700 of the 1,000 underwater videos as training sets, 200 as validation sets, and the remaining 100 as test sets to ensure that the model is trained on a large amount of data, validated on a smaller but sufficient amount of data, and finally tested on new and unseen data to evaluate its generalization ability. Subsequently, the tracking system can use the videos in the training set as input to train the initial recognition model, thereby obtaining the training recognition model, so that the initial recognition model learns how to identify cables from video frames, including details such as the location, shape, and texture of the cables. In order to enhance the robustness and accuracy of the model, the tracking system can also apply data enhancement techniques, such as rotation, flipping, scaling, etc., to increase the diversity of training data. Then, the tracking system can use the videos in the validation set to verify the training recognition model, thereby obtaining the validation recognition model. The cable locations and attributes in the validation video have been annotated in advance, and these annotated data will serve as a benchmark for evaluating the training recognition model. The performance of the training recognition model on the validation set is used to adjust the parameters of the model, such as reducing overfitting and optimizing the structure of the model to ensure that the model not only performs well on the training data, but also maintains high performance on new data. Finally, the tracking system can use the videos in the test set to perform a final performance test on the validation recognition model, thereby obtaining the cable recognition model. The test set data was not used in the training and validation stages to ensure the fairness and validity of the test results.

[0064] It should be noted that the specific values ​​such as the number of underwater videos, the number of videos included in the training set, the number of videos included in the validation set, and the number of videos included in the test set are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0065] Furthermore, the verification recognition model is tested based on the test video to obtain a cable recognition model, including: inputting the test video into the verification recognition model to obtain a video recognition result; matching the video recognition result with the cable labeling result corresponding to the test video to obtain a video matching result; determining a model evaluation index of the verification recognition model based on the video matching result, wherein the model evaluation index is used to reflect the accuracy of the recognition result of the verification recognition model; in response to the model evaluation index being greater than a preset index threshold, determining that the verification recognition model is a cable recognition model.

[0066] The cable labeling result may refer to the true value information of the cable position on each frame of the test video obtained through manual or automatic labeling. The video matching result may be a comparison result after comparing the recognition result of the model on the test video with the cable labeling result. The model evaluation index may be a set of standards used to quantitatively evaluate the performance of the cable recognition model. For example, the model evaluation index may include at least one or more of the following: accuracy, recall rate, etc., but is not limited thereto. The preset index threshold may be a pre-set value used to determine whether the performance of the verification recognition model can meet the actual application requirements.

[0067] In an optional embodiment, considering that the test video is not the data encountered during the training of the above-mentioned verification recognition model, it can be used to evaluate the generalization ability of the above-mentioned verification recognition model, that is, whether the verification recognition model can accurately recognize underwater cable images that have not been seen. Therefore, the tracking system can input the test video into the above-mentioned verification recognition model to obtain the video recognition result. At the same time, the tracking system can match the video recognition result with the cable labeling result corresponding to the test video to obtain the video matching result. Then, the tracking system can match the video recognition result of the verification recognition model with the cable labeling result corresponding to the test video to obtain the video matching result. By comparing the cable position, direction and state recognized by the model with the "true value" information manually labeled, the tracking system can further calculate the model evaluation index of the verification recognition model. Finally, the tracking system can determine whether the verification recognition model has reached the predetermined performance standard based on the model evaluation index determined by the video matching result. If the model evaluation index of the verification recognition model is greater than the preset index threshold, this indicates that the verification recognition model is mature and accurate enough to cope with the actual underwater cable recognition task. At this time, the tracking system can determine the above-mentioned verification recognition model as the cable recognition model, that is, the recognition model that is finally put into practical application.

[0068] For example, it can be assumed that the tracking system has obtained the above-mentioned verification recognition model after the above-mentioned training and verification steps, and the above-mentioned preset index threshold can be 0.8. In order to obtain a cable recognition model with high recognition accuracy, the tracking system can input the pre-divided test video into the above-mentioned verification recognition model to obtain a video recognition result. At the same time, professionals in the field of power systems can perform detailed underwater cable annotation on the above-mentioned test video to obtain a cable annotation result. Then, the tracking system can use a matching algorithm to compare the recognition output of the verification recognition model with the cable annotation result to determine the correctness and accuracy of the recognition of the verification recognition model, that is, the above-mentioned model evaluation index. It can be assumed that the value of the determined model evaluation index is 0.85. Finally, the tracking system can compare the above-mentioned model evaluation index with the above-mentioned preset index threshold. Since the value of the model evaluation index is greater than the preset index threshold, it indicates that the above-mentioned verification recognition model has shown high accuracy and comprehensiveness in identifying underwater cables. Therefore, the tracking system can determine the above-mentioned verification recognition model as a cable recognition model.

[0069] It should be noted that the above-mentioned preset indicator thresholds, model evaluation indicators and other specific values ​​are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0070] Furthermore, a model evaluation index of the verification recognition model is determined based on the video matching results, including: determining positive samples and negative samples from the video matching results, wherein the positive samples are used to characterize test videos whose matching degree between the video recognition results and the cable labeling results is greater than a first matching degree threshold, and the negative samples are used to characterize test videos whose matching degree between the video recognition results and the cable labeling results is greater than a second matching degree threshold, except for the positive samples, and the first matching degree threshold is greater than the second matching degree threshold; based on the positive samples and the negative samples, determining the precision and recall rate of the verification recognition model; based on the precision and recall rate, determining the model evaluation index of the verification recognition model.

[0071] The above-mentioned positive sample may refer to a test video whose matching degree between the video recognition result and the cable labeling result is greater than the above-mentioned first matching degree threshold. The above-mentioned negative sample may be a test video, other than the positive sample, whose matching degree between the video recognition result and the cable labeling result is greater than the above-mentioned second matching degree threshold. The above-mentioned first matching degree threshold may be a threshold used to determine whether the matching degree between the model recognition result and the cable labeling result is high enough. The above-mentioned second matching degree threshold may be a threshold used to distinguish samples with a low matching degree between the model recognition result and the cable labeling result. The above-mentioned precision rate may be the proportion of samples predicted by the above-mentioned verification recognition model as cables that are actually cables. The above-mentioned recall rate may be the proportion of all actual cables correctly identified by the above-mentioned verification recognition model.

[0072] In an optional embodiment, considering that distinguishing the above video matching results helps to more comprehensively understand the performance of the above verification recognition model, the tracking system can first determine positive samples and negative samples from the video matching results. The division of the above positive samples and negative samples is based on the matching degree between the model recognition results and the real annotation results. Specifically, the tracking system can classify the test video whose matching degree between the video recognition results and the cable annotation results is greater than the above first matching degree threshold as a positive sample, and classify the test video whose matching degree between the video recognition results and the cable annotation results is less than or equal to the above first matching degree threshold and greater than the above second matching degree threshold as a negative sample. Then, the tracking system can further determine the precision and recall rate of the verification recognition model based on the positive samples and the negative samples, and can determine the model evaluation index of the verification recognition model based on the precision and recall rate. Through the precision and recall rate, the accuracy and comprehensiveness of the verification recognition model in identifying underwater cables can be quantified, thereby more specifically describing the performance of the verification recognition model under different conditions, and whether the model can effectively discover all existing cables while avoiding unnecessary false alarms.

[0073] For example, it can be assumed that the tracking system has obtained the above-mentioned verification recognition model after the above-mentioned training and verification steps. In order to accurately obtain the model evaluation index of the verification recognition model, so as to more accurately understand the recognition performance of the verification recognition model, the tracking system can set the above-mentioned first matching degree threshold to 0.85 and the above-mentioned first matching degree threshold to 0.4. Then, for each frame in the test video, the tracking system can calculate the intersection over union (IoU) between the cable bounding box identified by the verification recognition model and the cable bounding box annotated by the expert. If the IoU is greater than 0.85, the tracking system can mark this frame as a positive sample, which means that the verification recognition model accurately identifies the underwater cable on this frame. Subsequently, the tracking system can mark the video frames with IoU greater than 0.4 but less than or equal to 0.85 as negative samples. These samples indicate that although the verification recognition model does not meet the standard of "high matching", it can still find the approximate location of the underwater cable. Based on the above positive samples and negative samples, the tracking system can calculate the precision and recall of the verification recognition model. The calculation formula of the precision can be shown as follows:

[0074]

[0075] In the formula, represents the precision, P true It indicates the number of samples that the verification recognition model correctly identifies the cable, that is, the number of positive samples, P falseIt indicates the number of samples where the verification recognition model mistakenly identifies non-cable areas as cables. The calculation formula of recall rate can be shown as follows:

[0076]

[0077] In the formula, γ represents the recall rate, N false It represents the number of samples that the underwater cable actually exists but the verification recognition model fails to recognize. The interpretation of other symbols in the formula is consistent with the above formula and will not be repeated here. Finally, the tracking system can further determine the model evaluation index of the verification recognition model based on the above precision and recall rate.

[0078] It should be noted that the specific values ​​of the first matching degree threshold, the second matching degree threshold, etc. are only for illustrative purposes, and the staff can set them according to actual needs, and are not limited here.

[0079] For ease of understanding, Figure 2 is a flow chart of an optional training of an underwater cable identification model according to an embodiment of the present invention, such as Figure 2 As shown, the tracking system can perform data collection and preprocessing, and build a deep neural network model. Specifically, the tracking system can obtain multiple videos of underwater cable tracking and inspection from underwater cable service companies. These videos cover different types of underwater cables and their status under various environmental conditions. The tracking system can convert the collected videos into a series of static images (video frames) for subsequent image analysis and processing. Subsequently, the tracking system can standardize the above video frames to 75×75 pixels and arrange them accordingly. Then, the tracking system can input the above video frames into the deep convolutional neural network model to train the deep convolutional neural network model, and use transfer learning and data enhancement to optimize the deep convolutional neural network model so that the classification accuracy of the model cable image reaches more than 90%. Finally, the tracking system can optimize the deep convolutional neural network model for underwater cable image classification based on the results of classification and recognition.

[0080] It should be noted that the specific values ​​such as the pixel size after standardization of the above-mentioned video frames are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0081] Figure 3 is a flowchart of an optional data collection and data preprocessing process according to an embodiment of the present invention, such as Figure 3As shown, data collection includes obtaining a video set and extracting video frames, wherein obtaining a video set may refer to obtaining a set of underwater cable-related video clips from multiple underwater cable service companies, and these videos contain actual images of underwater cables in different environments, such as different lighting conditions, water clarity, and cable status. Extracting video frames may refer to segmenting the collected videos into a series of static images, i.e., video frames, for image analysis and training of deep convolutional neural network models. After completing data collection, data set preprocessing may be performed, including constructing positive samples and negative samples, adjusting the video frame size to 3×75×75 pixels, and constructing training data sets, validation data sets, and test data sets, wherein constructing positive samples and negative samples helps to more fully understand the performance of the above-mentioned deep convolutional neural network model. Adjusting the video frame size is to ensure the consistency of the image when input into the above-mentioned deep convolutional neural network model training. Constructing training data sets, validation data sets, and test data sets can improve the training effect of the above-mentioned deep convolutional neural network model, thereby improving the recognition accuracy of the model. After completing data set preprocessing, a deep convolutional neural network model may be constructed, and then the model may be trained based on the above-mentioned data sets.

[0082] It should be noted that the above-mentioned specific values ​​such as the video frame size are only for illustrative purposes. The staff can set them according to actual needs and are not limited here.

[0083] Figure 4 is a schematic diagram of an optional confusion matrix according to an embodiment of the present invention, such as Figure 4 As shown in the figure, TP represents true positive, that is, the model correctly predicts the image containing underwater cables as the positive class, which means that the model not only detects the cable, but also correctly classifies it as "with cable". FP represents false positive, that is, the model mistakenly predicts the image without underwater cables as the positive class, that is, the model mistakenly identifies the image without "cable" as "with cable". FN represents false negative, that is, the model fails to correctly predict the image containing underwater cables as the positive class, that is, the model misses the image with "cable". TN represents true negative, that is, the model correctly predicts the image without underwater cables as the negative class, which means that the model accurately identifies the image without "cable". The confusion matrix provides an intuitive view of the performance of the underwater cable recognition model, helping to understand the classification ability of the model, especially when dealing with positive classes (images with cables) and negative classes (images without cables). By analyzing the confusion matrix, the shortcomings of the model can be identified, and the parameters of the model can be further optimized or the data processing strategy can be improved to improve the overall performance and reliability of the model.

[0084] According to an embodiment of the present invention, an embodiment of an underwater cable tracking device is provided. It should be noted that the device can be used to execute the underwater cable tracking method described above. The specific implementation method and application scenario are the same as those of the above embodiment and will not be described in detail here. Figure 5 is a schematic diagram of an underwater cable tracking device according to an embodiment of the present invention. Figure 5 As shown, the device comprises:

[0085] The video acquisition module 502 is used to acquire underwater videos captured by underwater navigation equipment, wherein the underwater videos are used to represent videos obtained by shooting an initial area where underwater cables are deployed.

[0086] The video recognition module 504 is used to recognize the underwater video based on the cable recognition model to obtain the cable recognition result, wherein the cable recognition model is deployed in the underwater navigation equipment, and the cable recognition result is used to characterize the position of the underwater cable in the initial area.

[0087] The path construction module 506 is used to construct a device navigation path based on the cable identification result.

[0088] The device control module 508 controls the underwater navigation device to track the underwater cable based on the device navigation path.

[0089] Furthermore, the video recognition module is also used to: perform complexity recognition on the underwater video to obtain a complexity index of the initial area, wherein the complexity index is used to characterize the difficulty of identifying the underwater cables from the initial area; based on the complexity index and the video parameters of the underwater video, obtain multiple initial image frames from the underwater video; perform standardization processing on the multiple initial image frames to obtain multiple target image frames; input the multiple target image frames into the cable recognition model to obtain the cable recognition result.

[0090] Furthermore, the video recognition module is also used to: obtain the area type of the initial area; perform obstacle recognition on the underwater video to obtain an obstacle recognition result, wherein the obstacle recognition result includes the type and number of obstacles existing in the initial area; and determine a complexity index based on the area type, the type and the number of obstacles.

[0091] Furthermore, the path construction module is also used to: determine the integrity of the underwater cables based on the positions of the underwater cables in the initial area, wherein the integrity is used to characterize whether all the underwater cables are in the initial area; in response to the integrity of the underwater cables being that not all of the underwater cables are in the initial area, determine the cable direction of the cables based on the cable identification results; determine the target area based on the cable direction, wherein the target area is used to characterize the area outside the initial area that contains the underwater cables; and construct a device navigation path based on the target area, the initial area and the cable direction.

[0092] Furthermore, the device also includes: a first acquisition module, used to acquire a target video set, wherein the target video set is used to characterize a video set obtained by shooting underwater areas with different complexity indicators, and underwater areas are provided with underwater cables; a first division module, used to divide the videos included in the target video set to obtain training videos, verification videos and test videos, wherein the number of training videos is greater than the number of verification videos, and the number of verification videos is greater than the number of test videos; a model training module, used to train an initial recognition model based on the training video to obtain a training recognition model; a model verification module, used to verify the training recognition model based on the verification video to obtain a verification recognition model; and a model testing module, used to test the verification recognition model based on the test video to obtain a cable recognition model.

[0093] Furthermore, the model testing module is also used to: input a test video to the verification and recognition model to obtain a video recognition result; match the video recognition result with the cable labeling result corresponding to the test video to obtain a video matching result; determine a model evaluation index of the verification and recognition model based on the video matching result, wherein the model evaluation index is used to reflect the accuracy of the recognition result of the verification and recognition model; in response to the model evaluation index being greater than a preset index threshold, determine that the verification and recognition model is a cable recognition model.

[0094] Furthermore, the model testing module is also used to: determine positive samples and negative samples from the video matching results, wherein the positive samples are used to characterize the test video whose matching degree between the video recognition results and the cable labeling results is greater than a first matching degree threshold, and the negative samples are used to characterize the test video whose matching degree between the video recognition results and the cable labeling results is greater than a second matching degree threshold except for the positive samples, and the first matching degree threshold is greater than the second matching degree threshold; based on the positive samples and the negative samples, determine the precision and recall rate of the verification recognition model; based on the precision and recall rate, determine the model evaluation index of the verification recognition model.

[0095] An embodiment of the present application further provides an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of the present invention when running.

[0096] An embodiment of the present application further provides a computer-readable storage medium, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0097] An embodiment of the present application further provides a computer program product, including a computer program, which implements the methods in various embodiments of the present invention when executed by a processor.

[0098] An embodiment of the present application further provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0099] The embodiments of the present application further provide a computer program, which implements the methods in the above-mentioned embodiments of the present invention when executed by a processor.

[0100] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0102] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0103] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0104] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0105] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An underwater cable tracking method, characterized in that: include: Acquire an underwater video captured by an underwater navigation device, wherein the underwater video is used to represent a video obtained by shooting an initial area where the underwater cable is deployed; Recognize the underwater video based on a cable recognition model to obtain a cable recognition result, wherein the cable recognition model is deployed in the underwater navigation equipment, and the cable recognition result is used to characterize the position of the underwater cable in the initial area; Constructing a device navigation path based on the cable identification result; The underwater navigation device is controlled to track the underwater cable based on the device navigation path.

2. The method according to claim 1, characterized in that The underwater video is identified based on the cable identification model to obtain a cable identification result, including: Performing complexity recognition on the underwater video to obtain a complexity index of the initial area, wherein the complexity index is used to characterize the difficulty of identifying the underwater cable from the initial area; Based on the complexity index and the video parameters of the underwater video, a plurality of initial image frames are captured from the underwater video; Performing standardization processing on the multiple initial image frames to obtain multiple target image frames; The plurality of target image frames are input into the cable recognition model to obtain the cable recognition result.

3. The method according to claim 1, characterized in that Performing complexity identification on the underwater video to obtain a complexity index of the initial area includes: Obtaining the region type of the initial region; Performing obstacle recognition on the underwater video to obtain an obstacle recognition result, wherein the obstacle recognition result includes the type and number of obstacles existing in the initial area; The complexity index is determined based on the area type, the type of the obstacle and the number of the obstacles.

4. The method according to claim 1, characterized in that Building a device navigation path based on the cable identification result includes: Determining the integrity of the underwater cable based on the position of the underwater cable in the initial area, wherein the integrity is used to indicate whether the underwater cable is completely in the initial area; In response to the integrity of the underwater cable being that the underwater cable is not entirely in the initial area, determining a cable direction of the cable based on the cable identification result; Determining a target area based on the cable direction, wherein the target area is used to represent an area outside the initial area and including the underwater cable; The device navigation path is constructed based on the target area, the initial area and the cable direction.

5. The method according to claim 1, characterized in that The method further comprises: Acquire a target video set, wherein the target video set is used to represent a video set obtained by shooting underwater areas with different complexity indicators, and the underwater area is provided with the underwater cable; Dividing the videos included in the target video set into training videos, verification videos and test videos, wherein the number of the training videos is greater than the number of the verification videos, and the number of the verification videos is greater than the number of the test videos; training an initial recognition model based on the training videos to obtain a training recognition model; Verifying the training recognition model based on the verification video to obtain a verification recognition model; The verification identification model is tested based on the test video to obtain the cable identification model.

6. The method according to claim 5, characterized in that Testing the verification recognition model based on the test video to obtain the cable recognition model includes: Input the test video into the verification recognition model to obtain a video recognition result; Matching the video recognition result with the cable labeling result corresponding to the test video to obtain a video matching result; Determining a model evaluation index of the verification recognition model based on the video matching result, wherein the model evaluation index is used to reflect the accuracy of the recognition result of the verification recognition model; In response to the model evaluation index being greater than a preset index threshold, determining that the verification identification model is the cable identification model.

7. The method according to claim 6, characterized in that Determining a model evaluation index of the verification recognition model based on the video matching result includes: Determine a positive sample and a negative sample from the video matching result, wherein the positive sample is used to characterize a test video whose matching degree between the video recognition result and the cable labeling result is greater than a first matching degree threshold, and the negative sample is used to characterize a test video whose matching degree between the video recognition result and the cable labeling result is greater than a second matching degree threshold except for the positive sample, and the first matching degree threshold is greater than the second matching degree threshold; Based on the positive samples and the negative samples, determining the precision and recall of the verification recognition model; Based on the precision and the recall, the model evaluation index of the verification recognition model is determined.

8. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 7 when running.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images