An image recognition-based power optical cable preparation method and system

Abstract

The application provides a power optical cable preparation method and system based on image recognition, and relates to the technical field of optical cable preparation.The method comprises the following steps: acquiring a power optical cable image and a short-term environment video of power optical cable laying; determining a plurality of preliminary extrusion pressures based on the power optical cable image; determining a vibration frequency sequence of a test machine based on the short-term environment video of power optical cable laying; controlling an extruder to extrude plastic melt on a plurality of test power optical cables based on the plurality of preliminary extrusion pressures, and after cooling, controlling the test machine to perform vibration testing on each test power optical cable based on the vibration frequency sequence of the test machine; acquiring an ultrasonic image of each test power optical cable after the vibration testing is completed; and determining a target extrusion pressure based on the ultrasonic image of each test power optical cable after the vibration testing is completed, so that the extrusion pressure suitable for the environmental vibration requirement can be accurately determined to guarantee the quality of the power optical cable coating layer.

Description

Technical Field

[0001] The present invention relates to the technical field of optical cable preparation, and in particular to a method and system for preparing a power optical cable based on image recognition. Background Art

[0002] Power optical cables, as critical infrastructure that carries both power transmission and signal communications, face significant challenges in their manufacturing quality. The extrusion process is a crucial step in ensuring their performance. The extruder precisely coats the cable core with molten plastic, creating a sheath that provides insulation, heat insulation, and protection. The uniformity, density, and bond strength of this sheath with the cable core directly impact the cable's environmental resistance, service life, and operational safety. Currently, numerous challenges exist in the manufacturing of power optical cables. For one thing, extrusion pressure, a key parameter in the extrusion process, determines the quality of the sheath. Traditional methods for determining extrusion pressure often rely on empirical experience or simple parameter matching, making them inadequate for diverse cable structures and appearances. This can lead to uneven sheath thickness, localized voids, or stress concentrations, potentially leading to operational failures. Furthermore, the complex and dynamic operating environments of power optical cables, exposed to harsh conditions such as natural vibration, can easily lead to aging, cracking, or delamination of the sheath and internal structure due to defects in initial manufacturing. In the existing preparation process, the test of the environmental adaptability of optical cables mostly relies on short-term static testing, which makes it difficult to simulate the dynamic stress effects under long-term harsh environments. As a result, some potential quality problems cannot be effectively identified before leaving the factory, which in turn affects the actual service life of the optical cable.

[0003] Therefore, how to accurately determine the extrusion pressure that adapts to the environmental vibration requirements to ensure the quality of the power optical cable coating is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by the present invention is how to accurately determine the extrusion pressure that adapts to the environmental vibration requirements to ensure the quality of the power optical cable coating.

[0005] According to a first aspect, the present invention provides a method for preparing a power optical cable based on image recognition, comprising: acquiring a power optical cable image and a short-term environmental video of the power optical cable laying; determining a plurality of preliminary extrusion pressures based on the power optical cable image; determining a vibration frequency sequence of a test machine based on the short-term environmental video of the power optical cable laying; controlling an extruder to extrude plastic melt to a plurality of test power optical cables respectively based on the plurality of preliminary extrusion pressures and after cooling, controlling a test machine to perform a vibration test on each test power optical cable based on the vibration frequency sequence of the test machine; acquiring an ultrasonic image of each test power optical cable after the vibration test is completed; and determining a target extrusion pressure based on the ultrasonic image of each test power optical cable after the vibration test is completed.

[0006] In one possible implementation, the determining of the vibration frequency sequence of the test machine based on the short-term environmental video of the power optical cable laying includes: determining the short-term environmental vibration frequency sequence of the power optical cable laying based on the short-term environmental video of the power optical cable laying; generating a simulation video of the long-term harsh environment of the power optical cable laying based on the short-term environmental video of the power optical cable laying and the short-term environmental vibration frequency sequence of the power optical cable laying; determining the vibration frequency sequence of the long-term harsh environment of the power optical cable laying based on the simulation video of the long-term harsh environment of the power optical cable laying; and using the vibration frequency sequence of the long-term harsh environment of the power optical cable laying as the vibration frequency sequence of the test machine.

[0007] In one possible implementation, determining the target extrusion pressure based on the ultrasonic image of each test power optical cable after the vibration test is completed includes: dividing the ultrasonic image of each test power optical cable after the vibration test is completed to obtain ultrasonic images of multiple test power optical cable segments of each test power optical cable; processing the ultrasonic images of the multiple test power cable segments of each test power optical cable to obtain ultrasonic information of the multiple test power cable segments of each test power optical cable; clustering the ultrasonic information of the multiple test power cable segments of each test power optical cable using a K-means clustering algorithm to obtain K clusters of each test power cable; and determining the target extrusion pressure based on the K clusters of each test power cable.

[0008] In a possible implementation, the target extrusion pressure is one of a plurality of preliminary extrusion pressures.

[0009] According to a second aspect, the present invention provides an image recognition-based power optical cable preparation system, comprising: an acquisition module for acquiring a power optical cable image and a short-term environmental video of the power optical cable laying; a preliminary pressure determination module for determining a plurality of preliminary extrusion pressures based on the power optical cable image; a vibration frequency determination module for determining a vibration frequency sequence of a test machine based on the short-term environmental video of the power optical cable laying; a testing module for controlling an extruder based on the plurality of preliminary extrusion pressures to extrude plastic melt from a plurality of test power optical cables respectively and after cooling, controlling a test machine to perform a vibration test on each test power optical cable based on the vibration frequency sequence of the test machine; an image acquisition module for acquiring an ultrasonic image of each test power optical cable after the vibration test is completed; and a target pressure determination module for determining a target extrusion pressure based on the ultrasonic image of each test power optical cable after the vibration test is completed.

[0010] In a possible implementation, the vibration frequency determination module is also used to: determine the short-term environmental vibration frequency sequence of the power optical cable laying based on the short-term environmental video of the power optical cable laying; generate a simulation video of the power optical cable laying in a long-term harsh environment based on the short-term environmental video of the power optical cable laying and the short-term environmental vibration frequency sequence of the power optical cable laying; determine the vibration frequency sequence of the power optical cable laying in a long-term harsh environment based on the simulation video of the power optical cable laying in a long-term harsh environment; and use the vibration frequency sequence of the power optical cable laying in a long-term harsh environment as the vibration frequency sequence of the test machine.

[0011] In one possible implementation, the target pressure determination module is further used to: divide the ultrasonic image of each test power optical cable after the vibration test is completed to obtain ultrasonic images of multiple test power optical cable segments of each test power optical cable; process the ultrasonic images of the multiple test power cable segments of each test power optical cable to obtain ultrasonic information of the multiple test power cable segments of each test power optical cable; cluster the ultrasonic information of the multiple test power cable segments of each test power optical cable using a K-means clustering algorithm to obtain K clusters of each test power cable; and determine the target extrusion pressure based on the K clusters of each test power cable.

[0012] In a possible implementation, the target extrusion pressure is one of a plurality of preliminary extrusion pressures.

[0013] According to a third aspect, an embodiment of the present invention provides an electronic device, comprising: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: acquiring an image of a power optical cable and a short-term environmental video of the laying of the power optical cable; determining a plurality of preliminary extrusion pressures based on the image of the power optical cable; determining a vibration frequency sequence of a test machine based on the short-term environmental video of the laying of the power optical cable; controlling an extruder to extrude plastic melt from a plurality of test power optical cables respectively based on the plurality of preliminary extrusion pressures and after cooling, controlling the test machine to perform a vibration test on each test power optical cable based on the vibration frequency sequence of the test machine; acquiring an ultrasonic image of each test power optical cable after the vibration test is completed; and determining a target extrusion pressure based on the ultrasonic image of each test power optical cable after the vibration test is completed.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned image recognition-based power optical cable preparation method, the method comprising: acquiring a power optical cable image and a short-term environmental video of the power optical cable laying; determining a plurality of preliminary extrusion pressures based on the power optical cable image; determining a vibration frequency sequence of a test machine based on the short-term environmental video of the power optical cable laying; controlling an extruder to extrude plastic melt to a plurality of test power optical cables respectively based on the plurality of preliminary extrusion pressures and after cooling, controlling the test machine to perform a vibration test on each test power optical cable based on the vibration frequency sequence of the test machine; acquiring an ultrasonic image of each test power optical cable after the vibration test is completed; and determining a target extrusion pressure based on the ultrasonic image of each test power optical cable after the vibration test is completed.

[0015] The present invention provides a method and system for preparing an electric optical cable based on image recognition, which includes obtaining an image of the electric optical cable and a short-term environmental video of the laying of the electric optical cable; determining multiple preliminary extrusion pressures based on the image of the electric optical cable; determining a vibration frequency sequence of a testing machine based on the short-term environmental video of the laying of the electric optical cable; controlling an extruder based on the multiple preliminary extrusion pressures to extrude plastic melt from multiple test electric optical cables respectively and after cooling, controlling a testing machine to perform a vibration test on each test electric optical cable based on the vibration frequency sequence of the testing machine; obtaining an ultrasonic image of each test electric optical cable after the vibration test is completed; and determining a target extrusion pressure based on the ultrasonic image of each test electric optical cable after the vibration test is completed. This method can accurately determine the extrusion pressure that meets the environmental vibration requirements to ensure the quality of the electric optical cable sheath. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a process for preparing a power optical cable based on image recognition provided by an embodiment of the present invention;

[0017] Figure 2 is a schematic diagram of a power optical cable according to an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of a flow chart for determining a vibration frequency sequence of a test machine according to an embodiment of the present invention;

[0019] Figure 4 A schematic diagram of a process for determining a target extrusion pressure according to an embodiment of the present invention;

[0020] Figure 5 A schematic diagram of a power optical cable preparation system based on image recognition provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0021] The present invention will be further described in detail below by means of specific embodiments in conjunction with the accompanying drawings. Similar elements in different embodiments are numbered with associated similar elements. In the following embodiments, many detailed descriptions are provided to enable the present invention to be better understood. However, those skilled in the art will readily appreciate that some of the features may be omitted under different circumstances, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the present invention are not shown or described in the specification. This is to avoid the core of the present invention being overwhelmed by excessive descriptions, and for those skilled in the art, it is not necessary to describe these related operations in detail. They can fully understand the related operations based on the description in the specification and the general technical knowledge in the art.

[0022] In an embodiment of the present invention, there is provided Figure 1 A method for preparing a power optical cable based on image recognition is shown, and the method for preparing a power optical cable based on image recognition includes steps S1 to S6:

[0023] Step S1: Acquire an image of a power optical cable and a short-term environment video of the power optical cable laying.

[0024] Power optical cables are specialized cables made from a conductive fiber optic core, precisely coated with a plastic melt for insulation and heat protection through an extrusion process, followed by molding and other processes. The plastic melt provides an electrically insulating environment for power transmission and a stable physical channel for signal transmission, ensuring long-term and reliable power and communication functions. Figure 2 Schematic diagram of a power optical cable in an embodiment of the present invention.

[0025] The power optical cable image is a static image of the appearance and structure of the power optical cable before overmolding, captured by an industrial high-resolution camera.

[0026] The short-term environmental video of the power cable laying site is a three-hour dynamic video recorded by a camera installed near the laying area. It contains dynamic changes caused by natural and man-made vibration sources in the laying environment, such as wind, traffic, and construction machinery, including information such as vibration frequency, amplitude, and duration.

[0027] Step S2: determining a plurality of preliminary extrusion pressures based on the power optical cable image.

[0028] In some embodiments, a preliminary pressure determination model may be used to determine multiple preliminary extrusion pressures, wherein the preliminary pressure determination model is a convolutional neural network model, the input of the preliminary pressure determination model is the power optical cable image, and the output of the preliminary pressure determination model is multiple preliminary extrusion pressures.

[0029] Convolutional neural network models include convolutional neural networks (CNNs), which are deep learning models that extract image features through multiple layers of convolution and pooling operations. Convolutional neural networks use convolutional layers to extract local features of an image, pooling layers to compress feature dimensions while retaining key information, and fully connected layers to integrate features and output predictions.

[0030] Extrusion pressure is used to control the flow rate of plastic melt in the extruder.

[0031] The multiple preliminary extrusion pressures are multiple candidate extrusion pressure parameter values ​​obtained after analyzing the power optical cable image through the preliminary pressure determination model.

[0032] The image of the power optical cable contains visual information such as the appearance structure, dimensional features, and surface state of the power optical cable core before overmolding. This information directly reflects the pressure conditions that may be required during the preparation of the optical cable and can be extracted by the convolutional neural network and converted into features related to extrusion pressure. The convolutional neural network can perform standardization and noise reduction preprocessing on the power optical cable image, and then use the convolution layer to extract local features in the power optical cable image, such as the diameter uniformity and surface smoothness of the optical cable. The pooling layer of the convolutional neural network can simplify the features to retain key patterns, and the fully connected layer can integrate high-dimensional features. By learning the correlation between image features and extrusion pressure, multiple preliminary extrusion pressures can be output. Multiple preliminary extrusion pressures are generated based on the different dimensions of the image features to ensure that possible optimization directions are covered.

[0033] In some embodiments, determining a plurality of preliminary extrusion pressures based on the power optical cable image includes steps S21 to S23:

[0034] Step S21 : determining appearance details of the power optical cable core, dimension feature data of the power optical cable core, and surface status records of the power optical cable core based on the power optical cable image.

[0035] In some embodiments, the appearance details of the power optical cable core, the size characteristic data of the power optical cable core, and the surface state record of the power optical cable core can be determined through a deep neural network.

[0036] A deep neural network (DNN) is an artificial intelligence model composed of multiple layers of neurons. It is one of the core architectures of deep learning. By simulating the connections between neurons in the human brain, DNNs can extract and process features from input data such as images, text, and numerical values ​​through multiple layers of nonlinear transformations, ultimately outputting the desired result.

[0037] The appearance details of the power optical cable core are the overall appearance feature information of the power optical cable core output by the deep neural network. The appearance details of the power optical cable core include the color uniformity of the surface of the power optical cable core, whether there are obvious wrinkles or deformations, and other intuitively visible appearance status information.

[0038] The dimensional characteristic data of the power optical cable core is the data reflecting the size specifications of the optical cable core output by the deep neural network, including the overall diameter value and the diameter difference at different positions.

[0039] The surface condition record of the power optical cable core is a record of the smoothness of the surface of the power optical cable core output by the deep neural network, including the location of scratches, protrusions, depressions and other defects on the surface.

[0040] Deep neural networks can process power cable images through the collaborative action of multiple layers of neurons. After the input layer of a deep neural network receives the pixel matrix of the power cable image, the shallow hidden layers sequentially perform nonlinear transformations on the data. Shallow neurons can capture basic visual features such as edges and color gradients, while deep neurons can integrate these features to identify the overall shape, dimensional patterns, and surface defects of the cable. By mapping the image features and appearance attributes learned during the training phase, deep neural networks can ultimately transform abstract visual signals into structured appearance details, dimensional data, and surface state records, thus achieving the transformation from image to quantitative features.

[0041] Step S22 , determining the location of the uneven thickness area, the distribution of surface roughness points, the required coating thickness, and the extrusion pressure range based on the appearance details of the power optical cable core, the dimensional characteristic data of the power optical cable core, and the surface condition record of the power optical cable core.

[0042] In some embodiments, the location of the uneven thickness area, the distribution of surface roughness points, the required coating thickness, and the extrusion pressure range can be determined through a deep neural network.

[0043] The location of the uneven thickness area is the specific location where the surface diameter of the power optical cable core has obvious deviations, determined by a deep neural network, such as the location mark where the diameter of a certain area suddenly becomes thicker or thinner.

[0044] The surface roughness point distribution is the distribution of rough areas and defect points on the surface of the power optical cable core output by the deep neural network.

[0045] The required coating thickness value is the thickness value that the plastic melt coating layer needs to reach, determined by a deep neural network.

[0046] The extrusion pressure range is the interval between the maximum and minimum extrusion pressure of the extruder preliminarily delineated by the deep neural network.

[0047] Deep neural networks can perform multi-layer feature fusion on the input data of the appearance details, dimensional characteristics, and surface condition records of the power optical cable core. The fully connected layers of the deep neural network can correlate different features, and the activation function can strengthen the influence of key features. Based on the correspondence between the basic features learned during training and the process requirements, the deep neural network can convert dimensional characteristics data into specific location markers of uneven thickness, convert surface conditions into the distribution pattern of roughness points, and infer the appropriate coating thickness based on the appearance details. Based on the combined influence of these features, a preliminary extrusion pressure range can be defined.

[0048] Step S23 , determining a plurality of preliminary extrusion pressures based on the position of the uneven thickness area, the distribution of the surface roughness points, the required coating thickness, and the extrusion pressure range.

[0049] In some embodiments, a deep neural network may be used to determine a plurality of preliminary extrusion pressures.

[0050] The hidden layer of the deep neural network first performs weighted processing on the input data and then captures the correlation between parameters through complex feature interactions. Based on the mapping between process parameters and pressure values ​​learned during the training phase, the deep neural network can convert the multi-dimensional input features into specific pressure values, and then output multiple preliminary extrusion pressure values ​​based on the feature differences in different regions.

[0051] Step S3: determining the vibration frequency sequence of the test machine based on the short-term environmental video of the power optical cable laying.

[0052] In some embodiments, Figure 3 A flow chart of determining a vibration frequency sequence of a test machine according to an embodiment of the present invention is provided. Determining the vibration frequency sequence of the test machine includes steps S31 to S34:

[0053] Step S31 : determining a short-term environmental vibration frequency sequence of the power optical cable laying based on the short-term environmental video of the power optical cable laying.

[0054] In some embodiments, a first sequence analysis model can be used to determine a short-term environmental vibration frequency sequence for power cable installation. The first sequence analysis model is a recurrent neural network, the input of the first sequence analysis model is a short-term environmental video of the power cable installation, and the output of the first sequence analysis model is a short-term environmental vibration frequency sequence for the power cable installation.

[0055] Recurrent neural networks (RNNs) are deep learning models that excel at processing time series data. Through their cyclic structure, RNNs make neuron outputs dependent on both the current input and historical hidden states, thereby capturing dependencies in data over time. RNNs can effectively extract dynamic patterns in time series data, such as vibration periodicity and frequency trends.

[0056] The short-term vibration frequency series for power cable installations is derived from short-term environmental video analysis of power cable installations using the first sequence analysis model. It reflects the temporal changes in vibration frequency within the power cable installation environment. This series records the frequency values ​​and temporal changes in the vibration environment within the power cable installation environment over a short three-hour period, such as fluctuations in vibration frequency caused by strong winds.

[0057] Short-term environmental videos of power cable installations contain dynamic temporal information about environmental vibrations, such as the time, intensity, and period of the vibrations. This information is fundamental for analyzing environmental vibration patterns and can be converted into a quantized frequency sequence by a recurrent neural network. Recurrent neural networks can split short-term environmental videos into a continuous sequence of frames based on time. The hidden state is then passed through a recurrent structure, allowing the model to capture correlations between vibration characteristics at different time points. The model can learn periodic patterns of vibration in short-term environmental videos of power cable installations, such as changes in vibration frequency caused by varying wind speeds. This allows the model to convert visual dynamic information into corresponding frequency values. After multiple iterations, the final output is a short-term environmental vibration frequency sequence that accurately reflects the temporal patterns of vibrations over a short period of time.

[0058] Step S32, generating a simulation video of the long-term harsh environment of the power optical cable laying based on the short-term environment video of the power optical cable laying and the short-term environment vibration frequency sequence of the power optical cable laying;

[0059] In some embodiments, a variational autoencoder can be used to generate a simulated video of a power cable installation in a long-term harsh environment. The variational autoencoder inputs a short-term environmental video of the power cable installation and a short-term environmental vibration frequency sequence of the power cable installation, and the variational autoencoder outputs a simulated video of the power cable installation in a long-term harsh environment.

[0060] A variational autoencoder (VAE) is a generative deep learning model consisting of an encoder and a decoder. The encoder maps input data into a latent probability space, obtaining the distribution of latent variables. The decoder reconstructs or generates new data based on the latent variables. By learning the underlying distribution of the data, the variational autoencoder can generate logically consistent long-term or extreme simulation data from short-term data, preserving core features while expanding the data context.

[0061] This video, generated using a variational autoencoder, simulates the vibration conditions of a power fiber optic cable installation under harsh long-term conditions over a week. This video demonstrates extreme vibration conditions, such as higher frequencies and more intense fluctuations.

[0062] Short-term environmental videos of power cable installations can extract basic visual features and dynamic patterns of the environment, while short-term vibration frequency sequences provide the core frequency patterns of the vibrations. Together, these provide the data distribution foundation for the variational autoencoder, enabling it to learn the characteristics of environmental vibrations and scale to harsh scenarios.

[0063] The variational autoencoder's encoder can jointly encode short-term environmental videos of power cable installations and short-term environmental vibration frequency sequences. By mapping the visual and vibration frequency characteristics of the installation environment into a latent space, the probability distribution parameters are obtained. The decoder can sample and reconstruct data from the latent space. Based on the vibration feature evolution patterns learned from the short-term installation environment data, such as the intrinsic correlation pattern between vibration intensity and frequency, it can be extended to generate simulated videos of power cable installations under long-term, harsh conditions. The variational autoencoder optimizes the reconstruction loss and distribution loss to ensure that the simulated video conforms to the data patterns of the short-term installation environment while reflecting the extreme characteristics of the harsh conditions in the installation environment.

[0064] Step S33, determining a vibration frequency sequence of the power optical cable when the power optical cable is laid in a long-term harsh environment based on a simulation video of the power optical cable being laid in a long-term harsh environment;

[0065] In some embodiments, the vibration frequency sequence of the power optical cable when the long-term environment is harsh can be determined by a second sequence analysis model. The second sequence analysis model is a recurrent neural network. The input of the second sequence analysis model is a simulation video of the power optical cable when the long-term environment is harsh, and the output of the second sequence analysis model is the vibration frequency sequence of the power optical cable when the long-term environment is harsh.

[0066] The vibration frequency sequence of power fiber optic cables installed in a long-term harsh environment is output by the second sequence analysis model, reflecting the time-varying vibration frequency in such a harsh environment. This frequency sequence records the frequency variations in such harsh environments.

[0067] A simulation video of a long-term harsh environment contains dynamic time series information about extreme vibrations in a power fiber optic cable installation. This dynamic time series information can be extracted and converted into a quantized frequency sequence using a recurrent neural network to reflect the vibration characteristics of the harsh environment. The recurrent structure of the recurrent neural network captures the temporal dependencies of long-term vibrations, including persistent frequency fluctuations and peak patterns. The model learns the characteristic patterns of extreme vibrations in the video, converting the visual dynamic information into corresponding frequency values. Through multiple iterations, the model outputs the vibration frequency sequence of the long-term harsh environment.

[0068] Step S34: using the vibration frequency sequence of the power optical cable when it is laid in a long-term harsh environment as the vibration frequency sequence of the test machine.

[0069] The vibration frequency sequence of the test machine is a frequency standard used to control the test machine to perform vibration testing on the test power optical cable.

[0070] The vibration frequency sequence of the test machine can enable the test machine to simulate the vibration state of a long-term harsh environment, ensuring the effectiveness and environmental adaptability of the test.

[0071] Step S4: controlling the extruder based on the multiple preliminary extrusion pressures to extrude plastic melt from the multiple test power optical cables respectively. After cooling, controlling the test machine to perform a vibration test on each test power optical cable based on the vibration frequency sequence of the test machine.

[0072] After the vibration frequency sequence of the test machine is determined, the extruder is controlled to extrude plastic melt into the multiple test power optical cables based on the multiple preliminary extrusion pressures. After cooling is completed, the test machine is controlled to perform a vibration test on the test power optical cables corresponding to each preliminary extrusion pressure based on the vibration frequency sequence of the test machine.

[0073] Step S5: Acquire an ultrasonic image of each tested power optical cable after the vibration test is completed.

[0074] Ultrasonic images of each tested power cable after vibration testing are captured by ultrasonic testing equipment. These images penetrate the cable's surface structure and clearly reveal its internal physical condition, including the distribution of defects, the uniformity of the sheath thickness, and the presence of voids and areas of stress concentration.

[0075] Step S6: determining a target extrusion pressure based on the ultrasonic image of each tested power optical cable after the vibration test is completed.

[0076] In some embodiments, Figure 4A schematic diagram of a process for determining a target extrusion pressure according to an embodiment of the present invention is provided. The process for determining the target extrusion pressure includes steps S61 to S64:

[0077] Step S61 : dividing the ultrasonic image of each tested power optical cable after the vibration test is completed to obtain ultrasonic images of a plurality of tested power optical cable segments of each tested power optical cable.

[0078] Ultrasonic images of multiple test cable segments are obtained by segmenting the vibration-tested ultrasonic image of the cable, resulting in a collection of images that can be used to accurately analyze the quality of each part of the cable. The ultrasonic images of each test cable segment contain the internal structural characteristics of the cable at the location of the segment, including details such as defects, coating thickness, uniformity, voids, and stress distribution within the segment.

[0079] For example, an ultrasonic image of a test power optical cable after the vibration test is completed is evenly divided according to the length of the cable. For example, an ultrasonic image of a test power optical cable of 100 meters is divided into 100 ultrasonic images of test power optical cable segments of 1 meter in length.

[0080] Step S62: Processing the ultrasonic images of the multiple test power optical cable segments of each test power optical cable to obtain ultrasonic information of the multiple test power optical cable segments of each test power optical cable.

[0081] In some embodiments, ultrasonic information of the multiple test power cable segments of each test power cable can be obtained by processing the ultrasonic images of the multiple test power cable segments of each test power cable using a cable information determination model. The cable information determination model is a convolutional neural network, the input of the cable information determination model is the ultrasonic images of the multiple test power cable segments of each test power cable, and the output of the cable information determination model is the ultrasonic information of the multiple test power cable segments of each test power cable.

[0082] Ultrasonic information from a test cable segment is a specific quantitative parameter output by a cable information determination model to describe the internal quality of the tested cable segment. This information includes the number of defects within the segment and the three-dimensional coordinates of each defect, the coating thickness, structural uniformity, void morphology and location parameters, and stress concentration areas and values.

[0083] The gap shape and position parameters include the gap length, width, depth (unit: mm) and the gap center coordinates.

[0084] The ultrasonic images of multiple test power cable segments of each test power cable contain the internal structural visual features of each segment, which can be extracted by convolutional neural networks and converted into quantitative quality indicators.

[0085] The first convolutional layer of the convolutional neural network uses a convolution kernel of a preset size to perform a sliding scan of the ultrasonic image of the test power optical cable segment. It initially captures the grayscale changes, edge contours (such as the coating boundary, the boundary between defects and normal structures), and regional texture features (such as the texture difference between uniformly coated areas and gap areas) in different areas of the image, and then generates a feature map containing basic visual features. Subsequently, the convolutional neural network's multiple convolutional layers can be processed sequentially and progressively. Shallow convolutional layers can focus on extracting local detailed features, such as the edge shape of defects and areas with sudden changes in coating thickness. Deep convolutional layers aggregate shallow features to identify more complex global correlation features, such as the spatial relationship between gap areas and surrounding structures and the overall distribution trend of stress concentration areas. Activation functions are used to enhance the response strength of effective features while suppressing irrelevant noise interference. During the convolution process, the pooling layer simultaneously reduces the dimensionality of the feature maps output by each layer. Through max pooling or average pooling, the pooling layer retains the most representative key information in the feature map, such as the maximum edge intensity and average grayscale difference. This reduces data redundancy while enhancing feature robustness, ensuring that subsequent processing focuses on core quality characteristics. After multiple rounds of convolution and pooling, the convolutional neural network converts the extracted high-dimensional feature maps into one-dimensional feature vectors through flattening, which are then input to the fully connected layer. The fully connected layer nonlinearly integrates the feature vectors using a weight matrix to establish a mapping between visual features and quantitative parameters. Specifically, based on visual signals captured in the image, such as defect distribution, coating morphology, void characteristics, and stress areas, the corresponding quantitative parameters can be accurately calculated and output, ultimately generating ultrasonic information for the tested power cable segment. Throughout this process, the feature recognition rules learned during the convolutional neural network training phase are required to accurately translate the visual features of the ultrasonic image into quantitative internal quality parameters.

[0086] Step S63: clustering is performed using a K-means clustering algorithm based on the ultrasonic information of the multiple test power optical cable segments of each test power optical cable to obtain K clusters of each test power optical cable.

[0087] The K-means clustering algorithm is an unsupervised learning algorithm. It can be used to partition data into K clusters based on similarity. The algorithm initializes K cluster centers, calculates the distance between each sample and the center, and assigns the sample to the closest cluster. It then iteratively updates the centers until they stabilize, allowing samples with similar characteristics to be grouped together. The K-means clustering algorithm is suitable for classification analysis of quality features.

[0088] In some embodiments, the K value can be obtained by artificially setting in advance.

[0089] In some embodiments, a K-means clustering algorithm is used to cluster K clusters for a test power optical cable as follows: First, from a dataset of ultrasonic information from multiple test power optical cable segments, K segments of ultrasonic information are randomly selected as initial cluster centers. The cluster centers contain core feature dimensions such as the number of defects, coating thickness, void morphology and location parameters, and stress concentration characteristic parameters. For each ultrasonic information segment in the dataset, the feature similarity between it and the K initial cluster centers is calculated using Euclidean distance, and the segment is then divided into the corresponding cluster based on the principle of closest distance. After all segments have been initially divided, the average value of each feature in the ultrasonic information of all segments within each cluster is recalculated to update the cluster center of each cluster. This step is repeated until the difference in the feature value between the two cluster centers is less than a preset threshold, and the clustering process converges, ultimately obtaining K clusters for the test power optical cable.

[0090] Each cluster represents a collection of segments with similar internal quality characteristics within the tested power optical cable. Multiple segments within the same cluster have a high degree of consistency in their ultrasonic information, including the number and distribution of defects, coating thickness uniformity, void morphology and location characteristics, and stress concentration. For example, a cluster may contain multiple segments with ≤2 defects, coating thickness deviation ≤0.1mm, and no obvious voids or stress concentration, indicating that the internal quality of these segments is stable. Another cluster may contain segments with ≥5 defects, coating thickness deviation >0.3mm, and significant voids, indicating that these segments have significant quality risks.

[0091] By clustering into multiple clusters, the complex, segment-level ultrasonic information from a single tested power optical cable can be structured and integrated. Grouping segments with similar characteristics avoids the redundant analysis of individual segments, significantly simplifying the data structure and making the distribution of internal cable quality more intuitive. For example, it can be seen which areas have stable quality and which areas have common defects. Clustering also provides feature labels for quality assessment. By analyzing the feature differences between different clusters, it is possible to quickly identify common quality issues that may exist in optical cables after vibration testing.

[0092] Step S64 : determining a target extrusion pressure based on the K clusters of each test power optical cable.

[0093] In some embodiments, determining the target extrusion pressure based on the K clusters of each test power optical cable includes steps S71-S72:

[0094] Step S71, constructing a cable map, the cable map includes multiple cable nodes and multiple edges between the multiple cable nodes, the node feature of each cable node includes K clusters of a test power cable, and the edges between the cable nodes are the differences in extrusion pressure between different test power cables.

[0095] The cable map is a data structure built for analyzing the relationship between quality and pressure. It consists of multiple cable nodes and edges between them. Each node is characterized by K clusters of a tested power cable, reflecting the quality distribution of the cable. The edges represent the differences in initial extrusion pressures for different tested power cables.

[0096] The optical cable atlas can structure the characteristic data of each cluster of scattered test power optical cables and their quality assessment information, and can intuitively present the relationship between different cluster characteristics and the overall quality status of the optical cable, thereby providing a reasonable structural network for subsequent analysis.

[0097] Step S72: Process the optical cable map based on a graph convolutional network to obtain a target extrusion pressure.

[0098] A graph convolutional network (GCN) is a deep learning model capable of processing graph-structured data. It updates node representations by aggregating node features with those of neighboring nodes, capturing topological relationships and inter-node dependencies within the graph. The input of the GCN is the optical cable graph, and its output is the target extrusion pressure.

[0099] The target extrusion pressure is the optimal extrusion pressure parameter obtained by processing the optical cable graph through the graph convolutional network. The target extrusion pressure is one of multiple preliminary extrusion pressures.

[0100] The target extrusion pressure ensures that the optical cable has the best internal quality after vibration testing, such as the least defects and the most uniform structure. The target extrusion pressure can meet the requirements of long-term use in harsh environments.

[0101] The optical cable atlas, through the synergistic effect of nodes, node features, and edge features, provides a systematic data foundation for determining target extrusion pressure. The node features of the optical cable atlas store core information such as defect distribution patterns within cluster segments, coating uniformity parameters, voids, and stress characteristics, deeply linking abstract clustering results with the actual internal quality status of the optical cable. By quantifying associations based on the differences in initial extrusion pressures corresponding to different tested power cables, a gradient network of pressure changes is constructed, making quality differences under different initial extrusion pressures explicit, thus laying the foundation for tracing the path of how extrusion pressure affects quality.

[0102] Graph convolutional networks (GCNs) possess the ability to learn features and reason about relationships that adapt to the structure of optical cable graphs. The GCN uses nodes to store cluster quality features and edges to represent the associations between different initial extrusion pressure differences, forming a complex network that reflects the relationship between extrusion pressure and quality. The GCN uses a message-passing mechanism to aggregate node features of adjacent clusters through edges. During iterative updates, it captures the nonlinear mapping relationship between quality features, pressure differences, and extrusion pressure within a cluster. For example, by learning potential patterns such as the formation of low-defect clusters under specific pressure differences and the emergence of stress concentration clusters when the pressure difference is too large, it transforms dispersed quality features into quantifiable pressure influence patterns. Furthermore, the GCN can optimize the weights of key features such as coating uniformity parameters and void morphology parameters to accurately identify the pressure features corresponding to optimal quality, thereby achieving precise reasoning from the quality feature network to process parameters, ultimately outputting a stable and reliable target extrusion pressure.

[0103] Based on the same inventive concept, Figure 5 A schematic diagram of a power optical cable preparation system based on image recognition provided in an embodiment of the present invention, wherein the power optical cable preparation system based on image recognition includes:

[0104] An acquisition module 71 is used to acquire power optical cable images and short-term environmental videos of power optical cable laying;

[0105] a preliminary pressure determination module 72, configured to determine a plurality of preliminary extrusion pressures based on the power optical cable image;

[0106] a vibration frequency determination module 73 for determining a vibration frequency sequence of a test machine based on a short-term environmental video of the power optical cable laying;

[0107] A testing module 74 is configured to control an extruder based on the multiple preliminary extrusion pressures to extrude plastic melts from the multiple test power optical cables respectively, and after cooling, control a testing machine to perform a vibration test on each test power optical cable based on a vibration frequency sequence of the testing machine;

[0108] An image acquisition module 75 is used to obtain an ultrasonic image of each tested power optical cable after the vibration test is completed;

[0109] The target pressure determination module 76 is configured to determine a target extrusion pressure based on the ultrasonic image of each tested power optical cable after the vibration test is completed.

[0110] It should be noted that, in order to simplify the presentation of this specification and facilitate understanding of one or more embodiments of the invention, the foregoing descriptions of the embodiments of this specification sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this disclosure method does not imply that the subject matter of this specification requires more features than those recited in the claims. In fact, an embodiment may have fewer features than all of the features of a single disclosed embodiment.

[0111] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A method for preparing a power optical cable based on image recognition, characterized in that: include: Acquire power fiber optic cable images and short-term environmental videos of power fiber optic cable laying; determining a plurality of preliminary extrusion pressures based on the power optical cable image; Determining the vibration frequency sequence of the test machine based on the short-term environmental video of the power optical cable laying, wherein determining the vibration frequency sequence of the test machine based on the short-term environmental video of the power optical cable laying comprises: Determining a short-term environmental vibration frequency sequence of the power optical cable laying based on the short-term environmental video of the power optical cable laying; generating a simulation video of the long-term harsh environment of the power optical cable laying based on the short-term environmental video of the power optical cable laying and the short-term environmental vibration frequency sequence of the power optical cable laying; Determining a vibration frequency sequence of the power optical cable when the long-term environment is harsh based on a simulation video of the power optical cable when the long-term environment is harsh; The vibration frequency sequence of the power optical cable when it is laid in a long-term harsh environment is used as the vibration frequency sequence of the test machine; Controlling an extruder based on the multiple preliminary extrusion pressures to extrude plastic melts from the multiple test power optical cables respectively and after cooling, controlling a test machine to perform a vibration test on each test power optical cable based on a vibration frequency sequence of the test machine; Obtain ultrasonic images of each tested power fiber optic cable after the vibration test is completed; Determining the target extrusion pressure based on the ultrasonic image of each tested power optical cable after the vibration test is completed, wherein determining the target extrusion pressure based on the ultrasonic image of each tested power optical cable after the vibration test is completed includes: dividing the ultrasonic image of each tested power optical cable after the vibration test is completed to obtain ultrasonic images of a plurality of tested power optical cable segments of each tested power optical cable; Processing the ultrasonic images of the multiple test power optical cable segments of each test power optical cable to obtain ultrasonic information of the multiple test power optical cable segments of each test power optical cable; Performing clustering using a K-means clustering algorithm based on ultrasonic information of multiple test power optical cable segments of each test power optical cable to obtain K clusters for each test power optical cable; The target extrusion pressure is determined based on the K clusters of each tested power cable.

2. The method for preparing a power optical cable based on image recognition according to claim 1, wherein: The target extrusion pressure is one of a plurality of preliminary extrusion pressures.

3. A power optical cable preparation system based on image recognition, characterized in that: include: An acquisition module, used to acquire power optical cable images and short-term environmental videos of power optical cable laying; a preliminary pressure determination module, configured to determine a plurality of preliminary extrusion pressures based on the power optical cable image; A vibration frequency determination module is used to determine the vibration frequency sequence of the test machine based on the short-term environmental video of the power optical cable laying, and the vibration frequency determination module is further used to: Determining a short-term environmental vibration frequency sequence of the power optical cable laying based on the short-term environmental video of the power optical cable laying; generating a simulation video of the long-term harsh environment of the power optical cable laying based on the short-term environmental video of the power optical cable laying and the short-term environmental vibration frequency sequence of the power optical cable laying; Determining a vibration frequency sequence of the power optical cable when the long-term environment is harsh based on a simulation video of the power optical cable when the long-term environment is harsh; The vibration frequency sequence of the power optical cable when it is laid in a long-term harsh environment is used as the vibration frequency sequence of the test machine; a testing module, configured to control an extruder based on the multiple preliminary extrusion pressures to extrude plastic melts from the multiple test power optical cables respectively, and after cooling, control a testing machine to perform a vibration test on each test power optical cable based on a vibration frequency sequence of the testing machine; An image acquisition module is used to obtain an ultrasonic image of each tested power optical cable after the vibration test is completed; a target pressure determination module, configured to determine a target extrusion pressure based on an ultrasonic image of each tested power optical cable after the vibration test is completed, the target pressure determination module further configured to: dividing the ultrasonic image of each tested power optical cable after the vibration test is completed to obtain ultrasonic images of a plurality of tested power optical cable segments of each tested power optical cable; Processing the ultrasonic images of the multiple test power optical cable segments of each test power optical cable to obtain ultrasonic information of the multiple test power optical cable segments of each test power optical cable; Performing clustering using a K-means clustering algorithm based on ultrasonic information of multiple test power optical cable segments of each test power optical cable to obtain K clusters for each test power optical cable; The target extrusion pressure is determined based on the K clusters of each tested power cable.

4. The power optical cable preparation system based on image recognition according to claim 3, characterized in that: The target extrusion pressure is one of a plurality of preliminary extrusion pressures.

5. An electronic device, characterized in that: include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and is configured to be executed by the processor to implement the power optical cable preparation method based on image recognition as described in any one of claims 1 to 2.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for preparing a power optical cable based on image recognition as described in any one of claims 1 to 2 is implemented.

Images