Automatic hidden danger identification method and device for optical cable external force construction site

Abstract

The invention relates to an automatic hidden danger identification method and device for an optical cable external force construction site, and belongs to the technical field of communication construction safety. Performing data standardization processing on the image data to obtain a target size image; constructing a hidden danger recognition model based on the convolutional neural network; and taking the target size image as the input of the hidden danger identification model based on the convolutional neural network, and carrying out hidden danger automatic identification. The system does not depend on manual patrol and a traditional three-staring management mode any more, a large amount of time and manpower resources are saved, real-time monitoring can be achieved, the phenomenon that hidden dangers cannot be found in time due to the fact that working time is not linked is avoided, and safety risks are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of communication construction safety, and in particular relates to a method and a device for automatically identifying hidden dangers at an optical cable external force construction site. Background Art

[0002] With the rapid development of communication technology, optical cables, as the core medium for information transmission, play a vital role in modern communication networks. The safety and reliability of optical cables are directly related to the stability and continuity of communication networks. Therefore, it is particularly important to ensure the safety of optical cables during construction and maintenance, especially to prevent hidden dangers caused by external factors.

[0003] In the existing technology, the management of optical cable external force construction sites mainly relies on manual inspections and the traditional three-monitor management model. Workers use handheld devices to record images and videos of the construction area and rely on manual analysis of this data to identify hidden dangers.

[0004] However, this approach not only requires a lot of time and human resources, but also cannot achieve real-time monitoring. The traditional three-point management model often fails to detect hidden dangers in time due to the lack of connection in working hours, thus increasing safety risks. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the invention is to provide a method and device for automatically identifying hidden dangers at optical cable external force construction sites. By constructing a hidden danger identification model based on a convolutional neural network, and using the target size image obtained after data standardization processing as the input of the hidden danger identification model based on the convolutional neural network, hidden dangers are automatically identified. It no longer relies on manual inspections and the traditional three-watch management mode, saves a lot of time and human resources, can realize real-time monitoring, avoids the phenomenon that hidden dangers are often not discovered in time due to the lack of connection in working hours, and reduces safety risks.

[0006] The first aspect of the present invention provides a method for automatically identifying hidden dangers of external force on optical cable construction sites, comprising:

[0007] S1, obtaining image data of the optical cable external force construction site;

[0008] S2, performing data standardization processing on the image data to obtain an image of a target size;

[0009] S3, building a hidden danger identification model based on convolutional neural network;

[0010] S4, using the target size image as input to the hidden danger identification model based on convolutional neural network to automatically identify hidden dangers.

[0011] Furthermore, the convolutional neural network includes: an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, a Dropout layer and an output layer.

[0012] Further, the convolution layer includes a first convolution layer, a second convolution layer and a third convolution layer, the pooling layer includes a first pooling layer and a second pooling layer, and S4 specifically includes:

[0013] S401, receiving the target size image in the input layer;

[0014] S402, in the first convolutional layer, extracting image features of the target size image, and generating a low-level feature map through a ReLU activation function;

[0015] S403, in the first pooling layer, performing dimensionality reduction processing on the low-level feature map by maximum pooling;

[0016] S404, in the second convolutional layer, extracting image features of the low-level feature map after dimensionality reduction processing, and generating a mid-level feature map through a ReLU activation function;

[0017] S405, in the second pooling layer, performing dimensionality reduction processing on the intermediate feature map by maximum pooling;

[0018] S406, in the third convolutional layer, extracting image features of the intermediate feature map after dimensionality reduction processing, and generating a high-level feature map through a ReLU activation function;

[0019] S407, in the fully connected layer, automatically identifying hidden dangers according to the high-level feature graph through a ReLU activation function;

[0020] S408, in the Dropout layer, a dropout rate of 0.5 is used to reduce overfitting;

[0021] S409: In the output layer, the probability of the hidden danger category is output through the Softmax activation function.

[0022] Furthermore, the training method of the hidden danger identification model based on convolutional neural network specifically includes:

[0023] Obtain image data samples of the optical cable external force construction site;

[0024] Performing normalization processing on the image data samples;

[0025] Perform data enhancement processing on the normalized image data samples;

[0026] With the goal of minimizing the cross entropy loss function, the Adam optimizer is used to perform batch training on the hidden danger identification model based on the convolutional neural network according to the preset batch size and the preset learning rate decay strategy:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032] m t =β1m t-1 +(1-β1)g t

[0033] Among them, θ t represents the parameters of the hidden danger identification model based on convolutional neural network at time step t, θ t-1 represents the parameters of the hidden danger identification model based on convolutional neural network at the previous time step t-1, α represents the learning rate, represents the second-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, ε represents a small constant for numerical stability, represents the first-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, v t represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β2 represents the second-order moment estimate attenuation coefficient, and v t-1 represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at the previous time step t-1, g t represents the gradient of the hidden danger identification model based on convolutional neural network at time step t, represents the gradient operation, f represents the loss function, represents the gradient operation of the parameters of the hidden danger identification model based on the convolutional neural network at time step t, m t represents the first-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β1 represents the first-order moment estimate attenuation coefficient, and m t-1 Represents the first-order moment estimate used for the hidden danger identification model based on convolutional neural network at the previous time step t-1.

[0034] Furthermore, after S4, the method further includes:

[0035] S5: When the hidden danger identification model based on the convolutional neural network identifies that the target size image has hidden dangers, push warning information to the person in charge of the current level of warning;

[0036] S6: When the hidden danger is not handled within the preset time, the warning information is pushed to the next level warning person in charge through the MQ delayed message mechanism.

[0037] The second aspect of the present invention provides a device for automatically identifying hidden dangers of external force on optical cable construction sites, comprising:

[0038] An acquisition module, used to acquire image data of the optical cable external force construction site;

[0039] A processing module, used for performing data standardization processing on the image data to obtain an image of a target size;

[0040] A building module for building a hidden danger identification model based on convolutional neural network;

[0041] The recognition module is used to use the target size image as the input of the hidden danger recognition model based on the convolutional neural network to automatically identify hidden dangers.

[0042] Furthermore, the convolutional neural network includes: an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, a Dropout layer and an output layer.

[0043] Furthermore, the convolution layer includes a first convolution layer, a second convolution layer and a third convolution layer, the pooling layer includes a first pooling layer and a second pooling layer, and the recognition module is specifically used for:

[0044] In the input layer, receiving the target size image;

[0045] In the first convolutional layer, image features of the target size image are extracted, and a low-level feature map is generated through a ReLU activation function;

[0046] In the first pooling layer, the low-level feature map is subjected to dimensionality reduction processing by maximum pooling;

[0047] In the second convolutional layer, the image features of the low-level feature map after dimensionality reduction are extracted, and the mid-level feature map is generated through the ReLU activation function;

[0048] In the second pooling layer, the intermediate feature map is subjected to dimensionality reduction processing by maximum pooling;

[0049] In the third convolutional layer, image features of the intermediate feature map after dimensionality reduction are extracted, and a high-level feature map is generated through a ReLU activation function;

[0050] In the fully connected layer, hidden dangers are automatically identified according to the high-level feature map through the ReLU activation function;

[0051] In the Dropout layer, a dropout rate of 0.5 is used to reduce overfitting;

[0052] In the output layer, the probability of the hidden danger category is output through the Softmax activation function.

[0053] Furthermore, the training method of the hidden danger identification model based on convolutional neural network specifically includes:

[0054] Obtain image data samples of the optical cable external force construction site;

[0055] Performing normalization processing on the image data samples;

[0056] Perform data enhancement processing on the normalized image data samples;

[0057] With the goal of minimizing the cross entropy loss function, the Adam optimizer is used to perform batch training on the hidden danger identification model based on the convolutional neural network according to the preset batch size and the preset learning rate decay strategy:

[0058]

[0059]

[0060]

[0061]

[0062]

[0063] m t =β1m t-1 +(1-β1)g t

[0064] Among them, θ t represents the parameters of the hidden danger identification model based on convolutional neural network at time step t, θ t-1 represents the parameters of the hidden danger identification model based on convolutional neural network at the previous time step t-1, α represents the learning rate, represents the second-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, ε represents a small constant for numerical stability, represents the first-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, v trepresents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β2 represents the second-order moment estimate attenuation coefficient, and v t-1 represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at the previous time step t-1, g t represents the gradient of the hidden danger identification model based on convolutional neural network at time step t, represents the gradient operation, f represents the loss function, represents the gradient operation of the parameters of the hidden danger identification model based on the convolutional neural network at time step t, m t represents the first-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β1 represents the first-order moment estimate attenuation coefficient, and m t-1 Represents the first-order moment estimate used for the hidden danger identification model based on convolutional neural network at the previous time step t-1.

[0065] Furthermore, the device also includes:

[0066] A first push module is used to push warning information to the current level warning person in charge when the hidden danger identification model based on the convolutional neural network identifies that the target size image has hidden dangers;

[0067] The second push module is used to push the warning information to the next level warning person in charge through the MQ delayed message mechanism when the hidden danger is not handled within the preset time.

[0068] A third aspect of the present invention provides a system for automatically identifying hidden dangers of optical cable external force construction sites, comprising: a memory and one or more processors.

[0069] One or more application programs are stored in the memory, and the one or more application programs are suitable for being executed by one or more processors to implement a method for automatically identifying hidden dangers of an optical cable external force construction site according to the first aspect.

[0070] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which can be loaded and executed by a processor to provide a method for automatically identifying hidden dangers of an optical cable external force construction site according to the first aspect.

[0071] The beneficial effects of the present invention are as follows:

[0072] The method and device of the present invention, by constructing a hidden danger identification model based on a convolutional neural network, and using the target size image obtained after data standardization processing as the input of the hidden danger identification model based on the convolutional neural network, automatically identifies hidden dangers, no longer relies on manual inspections and the traditional three-watch management model, saves a lot of time and human resources, can realize real-time monitoring, avoids the phenomenon that hidden dangers are often not discovered in time due to the lack of connection in working hours, and reduces safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are only used to illustrate specific embodiments and are not considered to limit the present invention. In the entire drawings, the same reference symbols represent the same components. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0074] Figure 1 A schematic flow chart of a method for automatically identifying hidden dangers of external force on an optical cable construction site provided by an embodiment of the present invention;

[0075] Figure 2 A schematic diagram of the structure of a device for automatically identifying hidden dangers of an optical cable external force construction site provided by an embodiment of the present invention;

[0076] Figure 3 A structural schematic diagram of a system for automatically identifying hidden dangers of optical cable external force construction sites provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0077] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work should fall within the scope of protection of the present invention.

[0078] Furthermore, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily obscuring the concepts disclosed in the present invention.

[0079] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the orientation or position relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. The terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0080] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of methods and devices consistent with some aspects of the present invention as detailed in the appended claims.

[0081] The present invention proposes a method and device for automatically identifying hidden dangers at an optical cable external force construction site, in order to solve the technical problem that the existing technology mainly relies on manual inspections and the traditional three-monitor management mode, which not only requires a lot of time and human resources, but also cannot achieve real-time monitoring. In addition, due to the lack of connection in working hours, hidden dangers are often not discovered in time, thereby increasing safety risks.

[0082] Method Embodiment

[0083] Reference Manual Attached Figure 1 , showing a flow chart of a method for automatically identifying hidden dangers of optical cable external force construction sites provided by an embodiment of the present invention.

[0084] An embodiment of the present invention provides a method for automatically identifying hidden dangers of an optical cable external force construction site, comprising:

[0085] Specifically, the method includes steps S1 to S6.

[0086] S1, obtaining image data of the optical cable external force construction site.

[0087] Specifically, install a camera at the target location of the optical cable external force construction site, debug the screen and related functions after power on, and connect the camera to the smart optical network camera management system. The system monitors the online and offline status of the camera in real time through the heartbeat mechanism. At the same time, the user binds the external force construction hazards at the construction site and configures the hierarchical warning responsible persons: the first-level warning responsible person (hazard reporter and reviewer), the second-level warning responsible person (municipal leader), the third-level warning responsible person (provincial leader), and the fourth-level warning responsible person (group leader). The camera then automatically captures the on-site image data every 5 minutes and uploads it to the optical network platform, replacing manual sign-in.

[0088] Optionally, the camera supports solar energy storage power supply and 4G / 5G network connection, has AI engineering machinery recognition and warning, live streaming, picture and video transmission functions, and can realize pan-tilt control and remote shouting operation.

[0089] In the present invention, the camera is used to automatically capture images at regular intervals and upload them to the optical network platform, realizing real-time and automated data collection, replacing the traditional manual sign-in method and greatly improving efficiency. By binding construction hazards and configuring hierarchical early warning responsible persons, a hierarchical management mechanism is formed to ensure timely reporting, processing and tracking of hazard information, thereby improving the standardization and traceability of management. The accumulation of image data can provide a rich data foundation for the optimization of machine learning models, further improving the accuracy of hazard identification and the intelligence level of the system.

[0090] S2, performing data standardization processing on the image data to obtain an image of a target size.

[0091] Among them, the target size image is an image of 224×224 pixels.

[0092] In the present invention, the image size is unified to 224×224 pixels to ensure that all images accepted by the network input layer have the same size, which helps the consistency and stability of the model. Standardizing the image size can reduce the consumption of computing resources, avoid performance degradation caused by processing images of different sizes, and improve the efficiency of the training and reasoning process.

[0093] S3, build a hidden danger identification model based on convolutional neural network.

[0094] Among them, the convolutional neural network includes: an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, a Dropout layer and an output layer.

[0095] In the present invention, convolutional neural networks are particularly good at extracting spatial features from images. Through multiple convolutional layers, CNN can automatically learn and extract image features from low to high levels, such as edges, textures, shapes, and patterns, and then identify potential construction hazards. This feature extraction process does not require manual feature design, allowing the model to flexibly respond to complex construction environments. The convolutional neural network structure can automatically classify according to the features in the image, making the hazard identification process more efficient and accurate, without the need for human intervention, and suitable for real-time monitoring and automated identification.

[0096] S4, uses the target size image as the input of the hidden danger identification model based on convolutional neural network to automatically identify hidden dangers.

[0097] In the present invention, by using the target size image as input, the model can focus more on identifying hidden dangers in the construction site, rather than noise or unnecessary background interference caused by different image sizes. In this way, the model can more accurately capture the subtle features of construction hidden dangers, thereby improving the accuracy of recognition.

[0098] In a possible implementation, the convolution layer includes a first convolution layer, a second convolution layer, and a third convolution layer, the pooling layer includes a first pooling layer and a second pooling layer, and S4 specifically includes sub-steps S401 to S409:

[0099] S401, in the input layer, receiving an image of a target size.

[0100] S402, in the first convolutional layer, extract the image features of the target size image, and generate a low-level feature map through the ReLU activation function.

[0101] It should be noted that the first convolutional layer includes 32 3x3 convolution kernels, a step size of 1 and uses the ReLU activation function.

[0102] S403: In the first pooling layer, the low-level feature map is subjected to dimensionality reduction processing by maximum pooling.

[0103] It should be noted that the first pooling layer uses 2x2 maximum pooling with a step size of 2.

[0104] S404, in the second convolutional layer, extract the image features of the low-level feature map after dimensionality reduction processing, and generate a mid-level feature map through a ReLU activation function.

[0105] It should be noted that the second convolutional layer includes 64 3x3 convolution kernels, a step size of 1 and uses a ReLU activation function.

[0106] S405, in the second pooling layer, the intermediate feature map is subjected to dimensionality reduction processing by maximum pooling.

[0107] It should be noted that the second pooling layer uses 2x2 maximum pooling with a step size of 2.

[0108] S406, in the third convolutional layer, extract the image features of the intermediate feature map after the dimensionality reduction processing, and generate a high-level feature map through the ReLU activation function.

[0109] It should be noted that the third convolutional layer includes 128 3x3 convolution kernels, a step size of 1 and uses a ReLU activation function.

[0110] S407, in the fully connected layer, according to the high-level feature graph, the hidden dangers are automatically identified through the ReLU activation function.

[0111] It should be noted that the fully connected layer contains 256 neurons and uses the ReLU activation function.

[0112] S408, in the Dropout layer, a dropout rate of 0.5 is used to reduce overfitting.

[0113] S409: In the output layer, the probability of the hidden danger category is output through the Softmax activation function.

[0114] In the present invention, by using multiple convolutional layers (first, second and third convolutional layers) to gradually extract different levels of features of the image, the model can be helped to recognize image information from simple to complex. Through the pooling layer (first pooling layer and second pooling layer), the feature map is reduced in dimension, reducing the computational burden and memory usage of the model. The use of the ReLU activation function can ensure the nonlinear features in each layer of the network, thereby enhancing the model's ability to process complex images. The fully connected layer can accurately classify hidden dangers by combining the high-level feature map information extracted by the convolutional layer.

[0115] In a possible implementation, the training method of the hidden danger identification model based on the convolutional neural network specifically includes:

[0116] Obtain image data samples of the optical cable external force construction site.

[0117] Furthermore, the labelimg tool is used to accurately label the image data samples, including labeling various types of construction machinery and potential hidden dangers.

[0118] Normalize the image data samples.

[0119] Specifically, the image data samples are normalized to standardize the pixel values ​​to between 0 and 1.

[0120] Perform data enhancement processing on the normalized image data samples.

[0121] Optionally, the data augmentation processing includes: random rotation, scaling, cropping and flipping.

[0122] With the goal of minimizing the cross entropy loss function, the Adam optimizer is used to perform batch training on the hidden danger identification model based on the convolutional neural network according to the preset batch size and the preset learning rate decay strategy:

[0123]

[0124]

[0125]

[0126]

[0127]

[0128] m t =β1m t-1 +(1-β1)g t

[0129] Among them, θ t represents the parameters of the hidden danger identification model based on convolutional neural network at time step t, θ t-1 represents the parameters of the hidden danger identification model based on convolutional neural network at the previous time step t-1, α represents the learning rate, represents the second-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, ε represents a small constant for numerical stability, represents the first-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, v t represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β2 represents the second-order moment estimate attenuation coefficient, and v t-1 represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at the previous time step t-1, g t represents the gradient of the hidden danger identification model based on convolutional neural network at time step t, represents the gradient operation, f represents the loss function, represents the gradient operation of the parameters of the hidden danger identification model based on the convolutional neural network at time step t, m t represents the first-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β1 represents the first-order moment estimate attenuation coefficient, and m t-1 Represents the first-order moment estimate used for the hidden danger identification model based on convolutional neural network at the previous time step t-1.

[0130] Optionally, the preset batch size is 32 or 64.

[0131] Optionally, the preset learning rate decay strategy is to decay the learning rate by 10% every 10 epochs.

[0132] Furthermore, the model performance is monitored on the validation set through the Early Stopping method. If the performance does not improve for multiple consecutive epochs, the training is stopped early. Dropout and L2 regularization are combined to control the model complexity to reduce overfitting. Grid search or random search is used to tune hyperparameters such as learning rate, batch size, and number of network layers to find the best parameter combination, thereby further improving the accuracy and generalization ability of the model.

[0133] In the present invention, the labelimg tool is used to annotate the image data of the optical cable external force construction site to ensure high-quality labels and improve the efficiency and accuracy of the model. Data enhancement processing generates more samples to enhance the robustness and generalization ability of the model under different construction sites and hidden dangers. The Adam optimizer is used in combination with first-order and second-order moment estimation to dynamically adjust the gradient, accelerate convergence, and avoid learning rate problems. Batch training makes efficient use of computing resources, reduces memory consumption, and accelerates training. Hyperparameters are tuned through grid or random search to find the optimal learning rate, batch size, and number of network layers to improve test performance. Combine Dropout and L2 regularization to prevent overfitting. Early stopping can terminate training early to avoid computational waste.

[0134] In a possible implementation manner, after S4, the method further includes:

[0135] S5: When the hidden danger identification model based on the convolutional neural network identifies that there are hidden dangers in the target size image, the warning information is pushed to the person in charge of the current level of warning.

[0136] S6: When the hidden danger is not handled within the preset time, the warning information is pushed to the next level warning person in charge through the MQ delayed message mechanism.

[0137] Furthermore, the person in charge of early warning can jump to the smart cloud protection module of the APP by clicking on the SMS link, view the early warning pictures and videos, and take corresponding actions according to the on-site situation. The person in charge of early warning can also turn on the live broadcast function of the equipment, remotely control the camera to adjust the angle and observe the construction site. When the situation is serious, the device can first send a pre-made early warning sound, make a remote call, or trigger a light flash warning to remind the on-site construction personnel, and then go to the site for disposal. According to the specific hidden dangers, take emergency measures to ensure the safety of the optical cable, such as suspending construction and setting warning signs.

[0138] Furthermore, the system will record the hidden dangers that have been dealt with in detail and feed back the processing results to the platform. The system will optimize and improve the machine learning model based on the feedback results. For example, an electronic fence is added to the location of the optical cable in the monitoring area, and an early warning is triggered when the relevant machinery enters the fenced area, thereby improving the recognition accuracy and stability of the model.

[0139] In the present invention, when the convolutional neural network model identifies a hidden danger, the warning information is pushed to the warning person in charge at the current level in a timely manner to ensure rapid response and handling of potential problems. Through the MQ delayed message mechanism, when the hidden danger is not handled in time, it is pushed to the person in charge at the next level, further speeding up the processing speed and preventing the hidden danger from being ignored. The person in charge of the warning can view the warning pictures and videos through the APP, quickly understand the on-site situation, and make more accurate decisions.

[0140] The beneficial effects of the present invention are as follows:

[0141] By constructing a hidden danger identification model based on convolutional neural networks, and using the target size image obtained after data standardization processing as the input of the hidden danger identification model based on convolutional neural networks, hidden dangers can be automatically identified, no longer relying on manual inspections and the traditional three-watch management model, saving a lot of time and human resources, and realizing real-time monitoring, avoiding the phenomenon that hidden dangers are often not discovered in time due to the lack of connection in working hours, thereby reducing safety risks.

[0142] Device Embodiment

[0143] Reference Manual Attached Figure 2 , showing a schematic structural diagram of a device for automatically identifying hidden dangers of optical cable external force construction sites provided by an embodiment of the present invention.

[0144] An automatic identification device 20 for hidden dangers of external force on an optical cable construction site provided by an embodiment of the present invention comprises:

[0145] The acquisition module 201 is used to acquire image data of the optical cable external force construction site.

[0146] The processing module 202 is used to perform data standardization processing on the image data to obtain an image of a target size.

[0147] The construction module 203 is used to construct a hidden danger identification model based on a convolutional neural network.

[0148] The identification module 204 is used to use the target size image as the input of the hidden danger identification model based on the convolutional neural network to automatically identify hidden dangers.

[0149] Furthermore, the convolutional neural network includes: an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, a Dropout layer and an output layer.

[0150] Furthermore, the convolution layer includes a first convolution layer, a second convolution layer and a third convolution layer, the pooling layer includes a first pooling layer and a second pooling layer, and the recognition module is specifically used for:

[0151] In the input layer, the target size image is received.

[0152] In the first convolutional layer, the image features of the target size image are extracted and the low-level feature map is generated through the ReLU activation function.

[0153] In the first pooling layer, the low-level feature maps are reduced in dimensionality through maximum pooling.

[0154] In the second convolutional layer, the image features of the low-level feature map after dimensionality reduction are extracted, and the mid-level feature map is generated through the ReLU activation function.

[0155] In the second pooling layer, the mid-level feature maps are reduced in dimensionality through maximum pooling.

[0156] In the third convolutional layer, the image features of the intermediate feature map after dimensionality reduction are extracted, and the high-level feature map is generated through the ReLU activation function.

[0157] In the fully connected layer, hidden dangers are automatically identified based on the high-level feature map through the ReLU activation function.

[0158] In the Dropout layer, a dropout rate of 0.5 is used to reduce overfitting.

[0159] In the output layer, the probability of the hidden danger category is output through the Softmax activation function.

[0160] Furthermore, the training method of the hidden danger identification model based on the convolutional neural network specifically includes:

[0161] Obtain image data samples of the optical cable external force construction site.

[0162] Normalize the image data samples.

[0163] Perform data enhancement processing on the normalized image data samples.

[0164] With the goal of minimizing the cross entropy loss function, the Adam optimizer is used to perform batch training on the hidden danger identification model based on the convolutional neural network according to the preset batch size and the preset learning rate decay strategy:

[0165]

[0166]

[0167]

[0168]

[0169]

[0170] m t =β1m t-1 +(1-β1)g t

[0171] Among them, θ t represents the parameters of the hidden danger identification model based on convolutional neural network at time step t, θ t-1 represents the parameters of the hidden danger identification model based on convolutional neural network at the previous time step t-1, α represents the learning rate, represents the second-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, ε represents a small constant for numerical stability, represents the first-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, v t represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β2 represents the second-order moment estimate attenuation coefficient, and v t-1 represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at the previous time step t-1, g t represents the gradient of the hidden danger identification model based on convolutional neural network at time step t, represents the gradient operation, f represents the loss function, represents the gradient operation of the parameters of the hidden danger identification model based on the convolutional neural network at time step t, m t represents the first-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β1 represents the first-order moment estimate attenuation coefficient, and m t-1 Represents the first-order moment estimate used for the hidden danger identification model based on convolutional neural network at the previous time step t-1.

[0172] Furthermore, the device also includes:

[0173] The first push module 205 is used to push warning information to the current level warning person in charge when the hidden danger identification model based on the convolutional neural network identifies that there is a hidden danger in the target size image.

[0174] The second push module 206 is used to push warning information to the next level warning person in charge through the MQ delayed message mechanism when the hidden danger is not handled within a preset time.

[0175] The automatic identification device 20 for hidden dangers at an optical cable external force construction site provided by the present invention can implement each of the methods for automatic identification of hidden dangers at an optical cable external force construction site in the above-mentioned method embodiment and can obtain the same technical effect. To avoid repetition, the present invention will not go into details.

[0176] The beneficial effects of the present invention are as follows:

[0177] By constructing a hidden danger identification model based on convolutional neural networks, and using the target size image obtained after data standardization processing as the input of the hidden danger identification model based on convolutional neural networks, hidden dangers can be automatically identified, no longer relying on manual inspections and the traditional three-watch management model, saving a lot of time and human resources, and realizing real-time monitoring, avoiding the phenomenon that hidden dangers are often not discovered in time due to the lack of connection in working hours, thereby reducing safety risks.

[0178] System Example

[0179] Reference Manual Attached Figure 3 , showing a structural schematic diagram of a system for automatically identifying hidden dangers of optical cable external force construction sites provided by an embodiment of the present invention.

[0180] The present invention provides an automatic identification system 30 for hidden dangers of optical cable external force construction sites, comprising: a memory 303 and one or more processors 301.

[0181] The memory 303 stores one or more application programs, and the one or more application programs are suitable for being executed by the one or more processors 301 to implement a method for automatically identifying hidden dangers at an optical cable external force construction site as described in a method embodiment.

[0182] An automatic identification system 30 for hidden dangers of external force on optical cable construction sites includes: a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, through a bus 302.

[0183] The structure of the automatic identification system 30 for hidden dangers of optical cable external force construction sites does not constitute a limitation on the embodiments of the present invention.

[0184] The beneficial effects of the present invention are as follows:

[0185] By constructing a hidden danger identification model based on convolutional neural networks, and using the target size image obtained after data standardization processing as the input of the hidden danger identification model based on convolutional neural networks, hidden dangers can be automatically identified, no longer relying on manual inspections and the traditional three-watch management model, saving a lot of time and human resources, and realizing real-time monitoring, avoiding the phenomenon that hidden dangers are often not discovered in time due to the lack of connection in working hours, thereby reducing safety risks.

[0186] Processor 301 may be a CPU, a general purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of the present invention. Processor 301 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0187] The bus 302 may include a path to transmit information between the above components. The bus 302 may be a PCI bus or an EISA bus, etc. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0188] The memory 303 can be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an EEPROM, a CD-ROM or other optical disk storage, an optical disk storage (including a compressed optical disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0189] Computer Readable Storage Medium Embodiments

[0190] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program can be loaded and executed by a processor to implement the method for automatically identifying hidden dangers of an optical cable external force construction site as described in the first aspect.

[0191] The applicant of the present invention has made a detailed explanation and description of the implementation examples of the present invention in conjunction with the drawings in the specification. However, those skilled in the art should understand that the above implementation examples are only preferred implementation schemes of the present invention, and the detailed description is only to help readers better understand the spirit of the present invention, but not to limit the scope of protection of the present invention. On the contrary, any improvements or modifications based on the inventive spirit of the present invention should fall within the scope of protection of the present invention.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the above embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or replacements that can be easily thought of by those skilled in the art within the technical scope disclosed in the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for automatically identifying hidden dangers of external force on optical cable construction sites, characterized in that: include: S1, obtaining image data of the optical cable external force construction site; S2, performing data standardization processing on the image data to obtain an image of a target size; S3, building a hidden danger identification model based on convolutional neural network; S4, using the target size image as input to the hidden danger identification model based on convolutional neural network to automatically identify hidden dangers.

2. The method for automatically identifying hidden dangers of optical cable external force construction site according to claim 1 is characterized in that: The convolutional neural network includes: an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, a Dropout layer and an output layer.

3. The method for automatically identifying hidden dangers of optical cable external force construction site according to claim 2 is characterized in that: The convolution layer includes a first convolution layer, a second convolution layer and a third convolution layer, the pooling layer includes a first pooling layer and a second pooling layer, and S4 specifically includes: S401, receiving the target size image in the input layer; S402, in the first convolutional layer, extracting image features of the target size image, and generating a low-level feature map through a ReLU activation function; S403, in the first pooling layer, performing dimensionality reduction processing on the low-level feature map by maximum pooling; S404, in the second convolutional layer, extracting image features of the low-level feature map after dimensionality reduction processing, and generating a mid-level feature map through a ReLU activation function; S405, in the second pooling layer, performing dimensionality reduction processing on the intermediate feature map by maximum pooling; S406, in the third convolutional layer, extracting image features of the intermediate feature map after dimensionality reduction processing, and generating a high-level feature map through a ReLU activation function; S407, in the fully connected layer, according to the high-level feature graph, using a ReLU activation function, automatically identifying hidden dangers; S408, in the Dropout layer, a dropout rate of 0.5 is used to reduce overfitting; S409: In the output layer, the probability of the hidden danger category is output through the Softmax activation function.

4. The method for automatically identifying hidden dangers of optical cable external force construction site according to claim 1 is characterized in that: The training method of the hidden danger identification model based on the convolutional neural network specifically includes: Obtain image data samples of the optical cable external force construction site; Performing normalization processing on the image data samples; Perform data enhancement processing on the normalized image data samples; With the goal of minimizing the cross entropy loss function, the Adam optimizer is used to perform batch training on the hidden danger identification model based on the convolutional neural network according to the preset batch size and the preset learning rate decay strategy: m t =β1m t-1 +(1-β1)g t Among them, θ t represents the parameters of the hidden danger identification model based on convolutional neural network at time step t, θ t-1 represents the parameters of the hidden danger identification model based on convolutional neural network at the previous time step t-1, α represents the learning rate, represents the second-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, ε represents a small constant for numerical stability, represents the first-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t. Value, v t represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β2 represents the second-order moment estimate attenuation coefficient, and v t-1 represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at the previous time step t-1, g t represents the gradient of the hidden danger identification model based on convolutional neural network at time step t, represents the gradient operation, f represents the loss function, represents the gradient operation of the parameters of the hidden danger identification model based on the convolutional neural network at time step t, m t represents the first-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β1 represents the first-order moment estimate attenuation coefficient, and m t-1 Represents the first-order moment estimate used for the hidden danger identification model based on convolutional neural network at the previous time step t-1.

5. The method for automatically identifying hidden dangers of optical cable external force construction site according to claim 1 is characterized in that: After S4, the method further includes: S5: When the hidden danger identification model based on the convolutional neural network identifies that the target size image has hidden dangers, push warning information to the person in charge of the current level of warning; S6: When the hidden danger is not handled within the preset time, the warning information is pushed to the next level warning person in charge through the MQ delayed message mechanism.

6. An automatic identification device for hidden dangers of external force on optical cable construction sites, characterized in that: The device for automatically identifying hidden dangers at an optical cable external force construction site is used to implement the method for automatically identifying hidden dangers at an optical cable external force construction site according to any one of claims 1 to 5, and the device comprises: An acquisition module, used to acquire image data of the optical cable external force construction site; A processing module, used for performing data standardization processing on the image data to obtain an image of a target size; A building module for building a hidden danger identification model based on convolutional neural network; The recognition module is used to use the target size image as the input of the hidden danger recognition model based on the convolutional neural network to automatically identify hidden dangers.

7. The automatic identification device for hidden dangers of external force on optical cable construction sites according to claim 6 is characterized in that: The convolutional neural network includes: an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, a Dropout layer and an output layer.

8. The automatic identification device for hidden dangers of external force on optical cable construction sites according to claim 7 is characterized in that: The convolution layer includes a first convolution layer, a second convolution layer and a third convolution layer, the pooling layer includes a first pooling layer and a second pooling layer, and the recognition module is specifically used for: In the input layer, receiving the target size image; In the first convolutional layer, image features of the target size image are extracted, and a low-level feature map is generated through a ReLU activation function; In the first pooling layer, the low-level feature map is subjected to dimensionality reduction processing by maximum pooling; In the second convolutional layer, the image features of the low-level feature map after dimensionality reduction are extracted, and the mid-level feature map is generated through the ReLU activation function; In the second pooling layer, the intermediate feature map is subjected to dimensionality reduction processing by maximum pooling; In the third convolutional layer, image features of the intermediate feature map after dimensionality reduction are extracted, and a high-level feature map is generated through a ReLU activation function; In the fully connected layer, hidden dangers are automatically identified according to the high-level feature map through the ReLU activation function; In the Dropout layer, a dropout rate of 0.5 is used to reduce overfitting; In the output layer, the probability of the hidden danger category is output through the Softmax activation function.

9. The automatic identification device for hidden dangers of external force on optical cable construction sites according to claim 6 is characterized in that: The training method of the hidden danger identification model based on the convolutional neural network specifically includes: Obtain image data samples of the optical cable external force construction site; Performing normalization processing on the image data samples; Perform data enhancement processing on the normalized image data samples; With the goal of minimizing the cross entropy loss function, the Adam optimizer is used to perform batch training on the hidden danger identification model based on the convolutional neural network according to the preset batch size and the preset learning rate decay strategy: m t =β1m t-1 +(1-β1)g t Among them, θ t represents the parameters of the hidden danger identification model based on convolutional neural network at time step t, θ t-1 represents the parameters of the hidden danger identification model based on convolutional neural network at the previous time step t-1, α represents the learning rate, represents the second-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, ε represents a small constant for numerical stability, represents the first-order moment estimate after correction of the deviation of the hidden danger identification model based on the convolutional neural network at time step t, v t represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β2 represents the second-order moment estimate attenuation coefficient, and v t-1 represents the second-order moment estimate of the hidden danger identification model based on convolutional neural network at the previous time step t-1, g t represents the gradient of the hidden danger identification model based on convolutional neural network at time step t, represents the gradient operation, f represents the loss function, represents the gradient operation of the parameters of the hidden danger identification model based on the convolutional neural network at time step t, m t represents the first-order moment estimate of the hidden danger identification model based on convolutional neural network at time step t, β1 represents the first-order moment estimate attenuation coefficient, and m t-1 Represents the first-order moment estimate used for the hidden danger identification model based on convolutional neural network at the previous time step t-1.

10. The automatic identification device for hidden dangers of optical cable external force construction site according to claim 6, characterized in that: The device also includes: A first push module is used to push warning information to the current level warning person in charge when the hidden danger identification model based on the convolutional neural network identifies that the target size image has hidden dangers; The second push module is used to push the warning information to the next level warning person in charge through the MQ delayed message mechanism when the hidden danger is not handled within the preset time.

Images