A power distribution network construction worker operation behavior monitoring method and system

Abstract

The application discloses a kind of power distribution network construction personnel job behavior monitoring method and system, comprising: first data set and second data set are constructed, first data set is personnel data set, and second data set is behavior data set;According to first data set and second data set, first identification model and second identification model are trained;The output of the first identification model after training is as the input of the second identification model, and a new third identification model is established;According to third identification model, power distribution network construction personnel job behavior monitoring is carried out.Compared with prior art, the present application extracts multi-level feature map information, so that the model can more accurately extract the characteristics of power distribution network construction site, thereby effectively improving the accuracy of construction personnel identification and the accuracy of construction personnel behavior detection in complex power distribution network construction environment, and through continuous learning and training, the accuracy and efficiency of power distribution network construction personnel identification and behavior detection can be further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution network construction personnel operation behavior monitoring, and particularly relates to a power distribution network construction personnel operation behavior monitoring method and system. BACKGROUND

[0002] In the power system, the safe operation of the power distribution network construction personnel is an important link to ensure the stable operation of the power grid. However, the traditional manual monitoring method often has problems such as low efficiency and difficulty in ensuring accuracy, especially in complex and changeable working environments, it is difficult to fully guarantee the safety of the construction personnel. The traditional manual inspection method requires the inspector to personally inspect the site, which not only consumes time and effort, but also is inefficient. In the existing solution, image acquisition is first performed by using an image acquisition device, and subsequent image analysis and recognition still need human participation, resulting in low inspection efficiency.

[0003] The existing image processing technology does not have ideal recognition and behavior detection effects on the power distribution network construction personnel in complex environments, because the number of power distribution network construction personnel is large, the structure is complex, and they may be affected by various factors, which limits the ability to accurately identify, classify and locate the power distribution network construction personnel, thereby failing to accurately monitor and inspect the power distribution network construction personnel. SUMMARY

[0004] This section is intended to summarize some aspects of the embodiments of the present application and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of the specification to avoid obscuring the purpose of this section, the abstract and the title, and such simplifications or omissions cannot be used to limit the scope of the present application.

[0005] In view of the above existing problems, the present application is proposed.

[0006] Therefore, the present application provides a power distribution network construction personnel operation behavior monitoring method and system, which can solve the problems mentioned in the background art.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] In a first aspect, the present application provides a power distribution network construction personnel operation behavior monitoring method, comprising:

[0009] constructing a first data set and a second data set, the first data set being a personnel data set, and the second data set being a behavior data set;

[0010] training a first recognition model and a second recognition model according to the first data set and the second data set;

[0011] The output of the first recognition model after training is taken as the input of the second recognition model, and a new third recognition model is established;

[0012] The power distribution network construction personnel operation behavior monitoring is performed according to the third recognition model.

[0013] As a preferred scheme of the power distribution network construction personnel operation behavior monitoring method, the training of the first recognition model and the second recognition model according to the first data set and the second data set comprises:

[0014] The first recognition model is any model with the first data set as the input and the personnel recognition result or the personnel recognition result related parameter as the output;

[0015] The second recognition model is any model with the second data set as the input and the personnel behavior result or the personnel behavior result related parameter as the output.

[0016] As a preferred scheme of the power distribution network construction personnel operation behavior monitoring method, the third recognition model comprises:

[0017] The third recognition model is any model with the output of the first recognition model as the input and the personnel behavior result or the personnel behavior result related parameter as the output.

[0018] As a preferred scheme of the power distribution network construction personnel operation behavior monitoring method, the first recognition model comprises a feature extraction module arranged in the first recognition model.

[0019] As a preferred scheme of the power distribution network construction personnel operation behavior monitoring method, the first recognition model further comprises a feature receptive field enhancement module arranged in the first recognition model.

[0020] As a preferred scheme of the power distribution network construction personnel operation behavior monitoring method, the first recognition model further comprises a feature fusion module arranged in the first recognition model.

[0021] As a preferred scheme of the power distribution network construction personnel operation behavior monitoring method, the power distribution network construction personnel operation behavior monitoring according to the third recognition model comprises:

[0022] A confidence threshold is preset, and the confidence threshold is compared with the output of the third recognition model;

[0023] The power distribution network construction personnel operation behavior monitoring is performed according to the comparison result.

[0024] In a second aspect, the present application provides a power distribution network construction worker behavior monitoring system, comprising:

[0025] a data set construction module for constructing a first data set and a second data set, the first data set being a personnel data set and the second data set being a behavior data set;

[0026] a model training module for training a first recognition model and a second recognition model according to the first data set and the second data set;

[0027] a model construction module for taking the output of the trained first recognition model as the input of the second recognition model to establish a new third recognition model;

[0028] a monitoring module for performing power distribution network construction worker behavior monitoring according to the third recognition model.

[0029] In a third aspect, the present application provides a computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the method as described above when executing the computer program.

[0030] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, the computer program being executed by a processor to implement the steps of the method as described above.

[0031] Compared with the prior art, the present application has the following beneficial effects: the present application proposes a power distribution network construction worker behavior monitoring method and system, constructs a first data set and a second data set, the first data set being a personnel data set and the second data set being a behavior data set, trains a first recognition model and a second recognition model according to the first data set and the second data set, takes the output of the trained first recognition model as the input of the second recognition model to establish a new third recognition model, and performs power distribution network construction worker behavior monitoring according to the third recognition model. Compared with the prior art, the present application extracts multi-level feature map information, so that the model can more accurately extract power distribution network construction site features, thereby effectively improving the construction worker recognition accuracy and the behavior detection accuracy of the construction workers in a complex power distribution network construction site environment. Moreover, through continuous learning and training, the accuracy and efficiency of the power distribution network construction worker recognition and behavior detection can be further improved. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments description will be briefly introduced as follows. Obviously, the drawings in the following description only represent some of the embodiments of the present application, and all other drawings obtained by those of ordinary skill in the art without creative effort based on these drawings should also fall within the protection scope of the present application.

[0033] Figure 1 The method flow chart of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure:

[0034] Figure 2 The power distribution network construction personnel identification model network diagram of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure:

[0035] Figure 3 The Gs_Block module structure diagram of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure:

[0036] Figure 4 The RFEM module structure diagram of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure:

[0037] Figure 5 The FFM module structure diagram of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure:

[0038] Figure 6 The power distribution network construction personnel behavior detection model network diagram of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure:

[0039] Figure 7 The internal structure diagram of the computer device of the power distribution network construction personnel operation behavior monitoring method and system provided by an embodiment of the present application is shown in the following figure. DETAILED DESCRIPTION

[0040] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the protection scope of the present application.

[0041] Embodiment 1

[0042] Reference Figures 1-7For the first embodiment of the present application, the embodiment provides a power distribution network construction worker operation behavior monitoring method and system, comprising:

[0043] In the prior related art, there are some deficiencies, for example, in the power distribution network construction site, due to the complex and changeable environment and the diverse behaviors of the construction workers, the traditional monitoring method often cannot accurately identify the behaviors of the construction workers, resulting in low monitoring efficiency, and even false positives or false negatives may occur. The present application provides a method that can effectively solve the above-mentioned problems, and next, how to realize the power distribution network construction worker operation behavior monitoring method will be described in detail in combination with multiple embodiments;

[0044] Figure 1 A method flowchart of a power distribution network construction worker operation behavior monitoring method and system is shown, comprising:

[0045] S101, a first data set and a second data set are constructed, the first data set is a personnel data set, and the second data set is a behavior data set;

[0046] In an optional embodiment, the personnel data set can be obtained by collecting personnel image information at the construction site and obtaining feature information of the personnel such as height, body shape, clothing, etc. through manual labeling; the behavior data set is obtained by capturing behavior images of the construction workers in different operation states through a video monitoring system and labeling, including but not limited to safety helmet wearing, tool use, operation posture, etc.

[0047] In an optional embodiment, a camera can be used to collect images of the power distribution network construction site to construct a power distribution network construction worker image data set, i.e. a personnel data set;

[0048] In an optional embodiment, specific information of the construction workers can also be directly called from the network system of the power distribution network construction site to obtain detailed data of the personnel such as name, work number, post, etc. to enrich the content of the personnel data set.

[0049] In the embodiment of the present application, a high-definition monitoring camera is used to collect images of the power distribution network construction site to construct a power distribution network construction worker image data set, i.e. a personnel data set. The constructed power distribution network construction worker data set is labeled by manually framing the power distribution network construction workers in the images;

[0050] In an optional embodiment, the power distribution network construction workers in the images can be labeled using a labeling tool based on deep learning, which can automatically identify the construction workers in the images and mark their key points such as head, hand, foot, etc. to assist manual labeling for more accurate labeling. In this way, the labeling efficiency and accuracy can be significantly improved to provide a higher quality data set for subsequent model training.

[0051] In an optional embodiment, annotating the power distribution network construction personnel in the image can also use a machine learning-based automatic annotation system. The system first analyzes the image through a pre-trained deep learning model, identifies the construction personnel in the image, and predicts the key point positions. Then, the system compares these prediction results with the manually annotated results, learns the accuracy of manual annotation, and continuously optimizes the model parameters to improve the accuracy of automatic annotation. Finally, the system can realize fast and accurate annotation of the construction personnel in the power distribution network construction site image, providing efficient data support for subsequent monitoring methods.

[0052] In an optional embodiment, annotating the power distribution network construction personnel in the image can also use image annotation tools such as LabelMe image annotation tool and LabelImg, etc. These tools allow users to draw bounding boxes on images and assign labels to each bounding box, providing accurate annotation data for model training.

[0053] In the embodiments of the present application, the LabelMe image annotation tool is used to frame and annotate the position of the power distribution network construction personnel in the power distribution network construction site image as "power distribution network construction personnel". LabelMe is a commonly used tool for image data annotation, which annotates the position of the power distribution network construction personnel in the power distribution network construction site image by framing and annotates its category as "power distribution network construction personnel".

[0054] In an optional embodiment, the behavior dataset can be captured by using a video monitoring system to capture the behavior images of the construction personnel in different working states, and annotated. These behavior images include but are not limited to safety helmet wearing, tool use, working posture, etc. In this way, a dataset containing various behavior characteristics can be constructed to provide rich training samples for the behavior recognition model.

[0055] In an optional embodiment, the behavior dataset can also be obtained by collecting existing power distribution network construction personnel behavior image datasets and combining manual annotation to obtain the behavior characteristics of the construction personnel in different working states. These datasets can include behavior images of construction personnel in various different environments and conditions, such as work in adverse weather conditions, work in narrow spaces, and work at night, etc. In this way, the diversity and comprehensiveness of the behavior dataset can be ensured, thereby improving the adaptability and accuracy of the monitoring system. In addition, different construction scenarios can be simulated, and actors can be used to act as construction personnel to simulate behaviors, further enriching the content of the behavior dataset. Finally, these behavior datasets will be used to train and verify the behavior recognition model to ensure that the model can accurately identify and classify various working behaviors of the construction personnel.

[0056] In the embodiments of the present application, the behavior data set can be obtained by collecting existing power distribution network construction personnel behavior image data sets, including but not limited to collecting the above data sets from various public sources, and then constructing the power distribution network construction personnel behavior image data set of the present application.

[0057] In the embodiments of the present application, the data set contains power distribution network construction personnel behavior images, including the following three types of sub-data sets related to power distribution network construction personnel behavior: power distribution network construction personnel safety helmet image data set, power distribution network construction personnel work clothes image data set, and power distribution network construction personnel bad construction behavior image data set. The collected power distribution network construction personnel behavior data set (i.e., the behavior data set) is labeled according to the behavior of the construction personnel. The specific types of labeling include "power distribution network construction personnel not wearing safety helmet", "power distribution network construction personnel not wearing work clothes", and "power distribution network construction personnel having bad construction behavior in the construction site".

[0058] It should be noted that in actual construction, public data sets can be collected on roboflow or other open source data set platforms. Roboflow is a large computer vision platform that provides a large number of open source data sets.

[0059] It should also be noted that the construction of the first data set and the second data set can provide more comprehensive and accurate personnel behavior information, so that the monitoring system can more effectively identify and analyze the behavior patterns of the construction personnel. By constructing a detailed data set containing personnel features and behavior features, the monitoring system can realize real-time monitoring and analysis of the construction personnel's work behavior, timely discover potential safety risks and violations, and thus improve the safety management level and work efficiency of the construction site. In addition, by continuously updating and expanding the data set, the monitoring system can adapt to different construction environments and conditions, ensuring the accuracy and effectiveness of the monitoring. Ultimately, the construction and application of these data sets not only help to improve the intelligent level of power distribution network construction personnel work behavior monitoring, but also provide strong technical support for construction safety management and quality control.

[0060] S102, training a first recognition model and a second recognition model according to the first data set and the second data set;

[0061] In the embodiments of the present application, training a first recognition model and a second recognition model according to the first data set and the second data set includes:

[0062] The first recognition model is any model whose input is the first data set and whose output is a personnel recognition result or a personnel recognition result related parameter;

[0063] The second recognition model is any model whose input is the second data set and whose output is a personnel behavior result or a personnel behavior result related parameter.

[0064] In an optional embodiment, the first and second recognition models can be constructed and trained using deep learning frameworks such as TensorFlow or PyTorch. These frameworks provide rich tools and libraries that enable researchers to quickly implement complex neural network structures and perform efficient model training and optimization. During the training process, various techniques can be employed to improve the generalization ability of the models, such as data augmentation, regularization, and transfer learning using pre-trained models. Data augmentation techniques transform training images through rotation, scaling, cropping, and other means to increase data diversity and reduce the risk of overfitting. Regularization techniques such as dropout or weight decay help prevent the model from overfitting to the training data. The use of pre-trained models can leverage the feature representations obtained from large-scale datasets, accelerating the convergence speed of the model on specific tasks and improving the final recognition performance. Through these methods, the first and second recognition models can more accurately identify the identities and behaviors of construction personnel, providing strong technical support for the power distribution network construction personnel operation behavior monitoring system.

[0065] In an optional embodiment, the first and second recognition models can also be further enhanced by integrating other advanced machine learning algorithms. For example, algorithms such as support vector machines (SVM), random forests, or gradient boosting decision trees can be used in combination to enhance the decision-making ability of the model when dealing with complex scenarios. These algorithms can handle nonlinear problems and perform well in feature selection and classification, helping to improve the accuracy and robustness of the monitoring system in actual applications.

[0066] In an optional embodiment, the first and second recognition models can also be designed in other ways, but the present application designs similar models, i.e., the first recognition model with input as the first data set and output as the personnel identification result or personnel identification result related parameters; the second recognition model with input as the second data set and output as the personnel behavior result or personnel behavior result related parameters, should be within the scope of protection of the present application.

[0067] In an optional embodiment, the personnel identification result related parameters can include personnel identification probabilities or other parameters that can directly or indirectly obtain personnel identification results, such as personnel identification labels, personnel identification confidence, etc. Similarly, the personnel behavior result related parameters can include behavior identification labels, behavior identification confidence, behavior identification timestamps, etc. These parameters are crucial to the monitoring system as they not only provide direct results of behavior identification but also provide additional information such as the accuracy of identification and the time of behavior occurrence, which are very useful for subsequent safety analysis and decision support.

[0068] In an optional embodiment, the personnel behavior result related parameters can include personnel behavior result probabilities or other parameters that can directly or indirectly obtain personnel behavior results, such as behavior recognition labels, behavior recognition confidence, behavior recognition timestamps, etc. These parameters are crucial for the monitoring system, as they not only provide direct results of behavior recognition, but also provide additional information such as the accuracy of recognition and the time of behavior occurrence, which are very useful for subsequent safety analysis and decision support. Through these parameters, the monitoring system can record and analyze the behavior patterns of construction personnel, timely discover and warn potential safety risks, and thus achieve more effective risk management and intervention during the construction process. In addition, these parameters can also be used to generate detailed construction personnel behavior reports, providing data support for construction management, helping managers better understand the actual situation of the construction site, optimizing the work process, and improving the overall construction efficiency and safety level.

[0069] In the embodiments of the present application, the first recognition model and the second recognition model can be designed by using a deep learning framework.

[0070] In the embodiments of the present application, the first recognition model includes a feature extraction module configured in the first recognition model.

[0071] For example, a feature extraction module Gs_Block is designed. Gs_Block uses a lightweight convolution module Gs-Conv and combines linear variation to quickly generate mirror feature images, as shown in Figure 3 which reduces the parameter amount of the model, thereby improving the overall operation speed of the model, and Gs_Block expands the detail feature receptive field of the power distribution network construction site image, thereby better capturing key information in the image and strengthening the model's recognition ability of construction personnel.

[0072] Specifically, input the power distribution network construction site feature map T1 with size HxWxC into the Gs-Conv layer with convolution kernel number C, size 3x3, padding value 2, and step 1, then input the result into the BN batch normalization layer and the Swish activation function layer to obtain the power distribution network construction site feature map T2 with size HxWxC. Perform linear variation operation on T2 to generate mirror feature image T3 with size HxWxC; then input T3 into the Gs-Conv layer with convolution kernel number C, size 3x3, padding value 2, and step 1, and then input into the BN batch normalization layer and the Swish activation function layer to obtain the power distribution network construction site feature map T4 with size HxWxC.

[0073] T2 = Swish(BN(GC 3×3 (T1)))

[0074] T3 = LNV(T2)

[0075] T4 = Swish(BN(GC 3×3 (T3)))

[0076] wherein GC 3×3 represents a Gs-Conv layer with a size of 3x3, BN represents batch normalization, and Swish represents an activation function.

[0077] Further, T3 and T4 are subjected to a channel concatenation operation to obtain a power grid construction site feature map T5 with a size of HxWx2C. T5 is input to a Gs-Conv layer with a number of convolution kernels of 2C, a size of 3x3, a padding value of 2, and a step of 2, and then input to a BN batch normalization layer and a Swish activation function layer to obtain and output a power grid construction site feature map T6 with a size of H / 2xW / 2x2C.

[0078] T5 = Concat(T2, T4)

[0079] T6 = Swish(BN(GC 3×3 (T4)))

[0080] wherein GC 3×3 represents a Gs-Conv layer with a size of 3x3, Concat represents a channel concatenation operation, BN represents a batch normalization layer, and Swish represents an activation function layer.

[0081] It should be noted that an image with a size of HxWxC refers to an image with a length of H pixels, a width of W pixels, and a number of channels of C. The number of channels of a feature map obtained after an image is calculated by a convolution layer is the same as the number of convolution kernels in the convolution layer. GS-Conv is an existing lightweight deep learning convolution module, which is mainly used to reduce the computational complexity and parameter amount of a model while maintaining or improving the performance of the model. The concatenation (Concat) operation is used to connect two or more feature matrices together in a certain dimension to generate a larger feature matrix. For example, a feature matrix with a size of HxW1 and a feature matrix with a size of HxW2 are subjected to a Concat operation to obtain a feature matrix with a size of Hx(W1+W2).

[0082] In the embodiments of the present application, the first recognition model further includes a feature receptive field enhancement module configured in the first recognition model.

[0083] Exemplarily, a receptive field enhancement module (RFEM) is designed to associate each pixel point in the power distribution network construction site image and improve the global receptive field of each pixel point in the feature image, thereby enhancing the model's recognition ability of construction personnel. The specific structure of the RFEM module is as shown in Figure 4

[0084] wherein the input size of the power distribution network construction site feature map L1 of HxWxC is input to the Conv layer with C convolution kernel numbers, 3x3 size, 2 padding value and 1 step in the second branch to obtain the power distribution network construction site feature map L2 with HxWxC size.

[0085] L2=C 3×3 (L1)

[0086] L3=C 5×5 (L1)

[0087] wherein C 3×3 represents the Conv layer with 3x3 size, and C 5×5 represents the Conv layer with 5x5 size.

[0088] Further, the channel concatenation operation is performed on L2 and L3 to obtain the power distribution network construction site feature map L4 with HxWx2C size. Then, L4 is input to the Conv layer with C convolution kernel numbers, 3x3 size, 2 padding value and 1 step to obtain the power distribution network construction site feature map L5 with HxWxC size.

[0089] L4=Concat(L2,L3)

[0090] L5=C 3×3 (L4)

[0091] wherein Concat represents the channel concatenation operation.

[0092] Further, the channel concatenation operation is performed on L1 and L5, and the result is input to the CBS module with C convolution kernel numbers, 3x3 size, 2 padding value and 1 step to obtain and output the power distribution network construction site feature map L6 with HxWxC size.

[0093] L6=CBS 3×3 (Concat(L1,L5)) ​

[0094] wherein: CBS 3×3 represents sequentially passing through a 3x3 Conv layer (convolution layer), a BN layer (normalization layer), and a Swish (activation function layer) to calculate.

[0095] It should be noted that the CBS module is a commonly used deep learning module composed of a Conv layer (convolution layer), a BN layer (batch normalization layer), and a Swish (activation function layer), and is mainly used for feature extraction and conversion. It extracts the features of the input data through convolution operation, normalizes it through the BN layer, and provides nonlinear characteristics through the Swish activation function.

[0096] In the embodiments of the present application, the first recognition model further includes a feature fusion module configured in the first recognition model.

[0097] For example, a feature fusion module FFM (Feature Fusion module) is designed. FFM extracts important features in the image through double-branch power distribution network construction site image input, such as Figure 5 as shown, and fuses the double-branch power distribution network construction site features through Concat operation to improve the model's recognition ability of power distribution network construction personnel. The double-branch input of FFM enhances the redundancy of model recognition. When one end of the input power distribution network construction site feature map is disturbed by noise, the other end of the input feature map can still provide effective feature information support. This redundancy makes the model have stronger robustness and fault tolerance when facing complex and variable inputs, and can more stably and reliably output the construction personnel recognition result. In addition, FFM also improves the randomness of image features through shuffle operation, improves the generalization performance of the network, avoids the problem of gradient explosion when the model training weight is updated due to regular data, thereby preventing the model from overfitting or underfitting.

[0098] wherein, the power distribution network construction site feature maps A1 and A2 are input from the fourth branch and the fifth branch respectively, and the sizes of the feature maps A1 and A2 are both HxWxC. A1 is input into a CBS module with a convolution kernel number of C, a size of 3x3, a padding value of 2, and a step of 1, to obtain a power distribution network construction site feature map A3 with a size of HxWxC. A2 is input into a CBS module with a convolution kernel number of C, a size of 3x3, a padding value of 2, and a step of 1, to obtain a power distribution network construction site feature map A4 with a size of HxWxC.

[0099] A3 = CBS 3×3 (A1)

[0100] A4 = CBS 3×3 (A2)

[0101] The A3 and A4 are first subjected to a shuffle operation and then subjected to a channel superposition operation to obtain a power distribution network construction site feature map A5 with a size of HxWx2C. Finally, A5 is input into a CBS module with a convolution kernel number of 2C, a size of 3x3, a padding value of 2, and a step of 1 to obtain and output a power distribution network construction site feature map A6 with a size of HxWx2C.

[0102] A5 = Concat(shuffle(A3), shuffle(A4))

[0103] A6 = CBS 3×3 (A5)

[0104] Wherein: Concat represents a channel superposition operation, and shuffle represents a shuffle operation.

[0105] It should be noted that the shuffle operation is a deep learning operation, and the shuffle operation randomly shuffles the order of samples in the data set. The purpose of this is to ensure that the model sees random samples at each iteration during training, thereby avoiding the model from learning the order characteristics of the data and improving the generalization ability of the model.

[0106] In an optional embodiment, based on the Gs_Block module, the RFEM module and the FFM module, a power distribution network construction personnel identification model is proposed. The power distribution network construction site image is input into the model, and the model automatically identifies and locates the construction personnel position in the construction site image. The model network diagram is as shown in Figure 2 The specific implementation process can be as follows:

[0107] A power distribution network construction site image F1 is input into a CBS module to obtain a power distribution network construction site feature map F2. F2 is input into a Gs_Block module for feature extraction to obtain a power distribution network construction site feature map F3. F3 is input into the Gs_Block module and the RFEM module to obtain a power distribution network construction site feature map F4. F4 is input into the Gs_Block module and the RFEM module to obtain a power distribution network construction site feature map F5. F5 is input into the Gs_Block module and the RFEM module to obtain a power distribution network construction site feature map F6. F6 is input into a SPPF to obtain a power distribution network construction site feature map F7.

[0108] Further, F7 is subjected to a linear interpolation operation to obtain a power distribution network construction site feature map F8. F5 and F8 are input into an FFM for calculation, and then the calculation result is input into an RFEM module to obtain a power distribution network construction site feature map F9. F9 is subjected to a linear interpolation operation to obtain a power distribution network construction site feature map F 10 . F4 and F 10The calculation result is input into the RFEM module to obtain the power distribution network construction site feature map F 11 The calculation result is input into the RFEM module to obtain the power distribution network construction site feature map F 11 The calculation result is input into the RFEM module to obtain the power distribution network construction site feature map F 12 The calculation result is input into the RFEM module to obtain the power distribution network construction site feature map F 12 The calculation result is input into the RFEM module to obtain the power distribution network construction site feature map F 13 .

[0109] Further, the power distribution network construction site feature map F 11 , the feature map F 12 , and the feature map F 13 are input into the target detection head head to obtain a power distribution network construction personnel identification result, which includes the category, the regression box, and the confidence of the construction personnel in the image.

[0110] It should be noted that training the first identification model and the second identification model according to the first data set and the second data set can enable the model to be trained on different data sets, thereby improving the accuracy and adaptability of the model in identifying power distribution network construction personnel in different scenarios. The first data set may contain more types of construction environments and personnel postures, while the second data set may focus more on detailed features in specific environments. By combining the two data sets to train the model, the advantages of both can be integrated, so that the model can maintain a high recognition rate when facing various complex scenarios.

[0111] S103, inputting the output of the trained first identification model into the second identification model to establish a new third identification model;

[0112] In the embodiments of the present application, the third identification model includes:

[0113] The third identification model is any model whose input is the output of the first identification model and whose output is the personnel behavior result or the personnel behavior result related parameters.

[0114] In an optional embodiment, the third identification model can be implemented through deep learning, machine learning, or neural network technology. For example, a convolutional neural network (CNN) can be used as the basic architecture of the third identification model to learn the behavior patterns of power distribution network construction personnel through training. In the training process, the backpropagation algorithm and gradient descent optimization method can be used to adjust the network weights to minimize the difference between the predicted results and the actual behavior results.

[0115] In the embodiments of the present application, a power distribution network construction personnel behavior detection model (i.e., the trained second identification model) is constructed, as shown in Figure 6As shown, the construction personnel identification result output by the power distribution network construction personnel identification model is input into the model to detect whether the power distribution network construction personnel has behavior. The model uses the AD-Conv cavity convolution to extract image features, increases the receptive field of each convolution kernel while maintaining the size of the feature map, helps to extract more context information from the input feature map, and thus improves the detection accuracy and efficiency.

[0116] Specifically, the power distribution network construction personnel feature map K1 with an input size of HxWxC is input into the upper and lower two different branches. In the upper branch, K1 is input into the AD-Conv cavity convolution layer with C number of convolution kernels, a size of 3x3, a padding value of 2, and a step of 1, and then the result is subjected to Avg-Pooling average pooling operation to obtain a power distribution network construction personnel feature map K2 with a size of HxWxC. In the lower branch, K1 is input into the AD-Conv cavity convolution layer with C number of convolution kernels, a size of 3x3, a padding value of 2, and a step of 1, and then the result is subjected to Max-Pooling maximum pooling operation to obtain a power distribution network construction personnel feature map K3 with a size of HxWxC.

[0117] K2 = ADConv 3×3 (AvgPooling(K1))

[0118] K3 = ADConv 3×3 (MaxPooling(K1))

[0119] ADConv 3×3 ADConv represents an AD-Conv cavity convolution layer with a size of 3x3, AvgPooling represents average pooling, and MaxPooling represents maximum pooling.

[0120] The power distribution network construction personnel feature maps K2 and K3 are subjected to channel superposition operation, and the result is input into the AD-Conv cavity convolution layer with C number of convolution kernels, a size of 3x3, a padding value of 2, and a step of 1 to obtain a power distribution network construction personnel feature map K4 with a size of HxWxC. K4 is input into the Full-connection full connection layer for power distribution network construction personnel behavior result calculation and classification. Finally, the power distribution network construction personnel behavior detection result is output.

[0121] It should be noted that the second recognition model and the third recognition model can be the same model or use the same model, because the inputs and outputs of the second recognition model and the third recognition model are the same, only the input of the third recognition model is provided by the output of the first recognition model, and the second recognition model and the third recognition model are used in the present application to distinguish the two models. The essence of the third recognition model is to input the output of the first recognition model into the second recognition model, which can also be considered as a combination of the first recognition model and the second recognition model.

[0122] In S104, the power distribution network construction personnel operation behavior monitoring is performed according to the third recognition model.

[0123] In the embodiment of the present application, the power distribution network construction personnel operation behavior monitoring according to the third recognition model includes:

[0124] The confidence threshold is pre-set, and the confidence threshold is compared with the output of the third recognition model;

[0125] The power distribution network construction personnel operation behavior monitoring is performed according to the comparison result.

[0126] In an optional embodiment, after the power distribution network construction personnel recognition model and the power distribution network construction personnel behavior detection model are trained, the two models are applied to the monitoring and early warning of the power distribution network construction personnel behavior. First, it is ensured that the model can receive the construction personnel behavior video frame data from the power distribution station in real time, and then a reasonable recognition and detection threshold is set in the system to ensure that the confidence of the construction personnel identity recognition and behavior detection meets the requirements. Once the model recognizes the power distribution network construction personnel or detects the behavior, and the confidence exceeds the preset alarm threshold, the system will automatically trigger the alarm mechanism. The mechanism will record the specific time of the behavior, the identity and position of the construction personnel involved in the power distribution station and other key details, and then immediately inform the on-duty personnel, relevant team managers and safety personnel through multiple channels such as email, SMS or system notification, to ensure timely identification and response to potential safety risks. This can effectively improve the supervision of the power distribution network construction personnel behavior, avoid safety problems caused by behavior, and further ensure the stable operation and safety of the power distribution network.

[0127] In an optional embodiment, the power distribution network construction personnel dataset is divided, and the power distribution network construction personnel identification model is trained and verified. The constructed power distribution network construction personnel dataset is divided into a training set and a verification set according to a ratio of 4:1, and the power distribution network construction personnel identification model is trained. First, all parameters of the network are initialized, and hyperparameters related to training are configured. The training set and the verification set will be divided into several small batches, and each batch of training data will be input into the model for learning, and the training loss value of the batch is calculated and recorded. Then, the batch data of the verification set is input into the model, and the corresponding verification loss value is calculated. The algorithm adjusts and optimizes itself according to the loss of each time, and as the training proceeds, the loss value gradually converges, and finally the training of the power distribution network construction personnel identification model is completed.

[0128] In an optional embodiment, the power distribution network construction personnel behavior dataset is divided, and the power distribution network construction personnel behavior detection model is trained and verified. The constructed power distribution network construction personnel behavior dataset is divided into a training set and a verification set according to a ratio of 4:1, and the power distribution network construction personnel behavior detection model is trained. First, all parameters of the network are initialized, and hyperparameters related to training are configured. The training set and the verification set will be divided into several small batches, and each batch of training data will be input into the model for learning, and the training loss value of the batch is calculated and recorded. Then, the batch data of the verification set is input into the model, and the corresponding verification loss value is calculated. The algorithm adjusts and optimizes itself according to the loss of each time, and as the training proceeds, the loss value gradually converges, and finally the training of the power distribution network construction personnel behavior detection model is completed.

[0129] In summary, the application provides a power distribution network construction personnel operation behavior monitoring method, a first dataset and a second dataset are constructed, the first dataset is a personnel dataset, and the second dataset is a behavior dataset; a first identification model and a second identification model are trained according to the first dataset and the second dataset; the output of the trained first identification model is used as the input of the second identification model, and a new third identification model is established; and the third identification model is used for power distribution network construction personnel operation behavior monitoring. Compared with the prior art, the application extracts multi-level feature map information, so that the model can more accurately extract power distribution network construction site features, thereby effectively improving the construction personnel identification accuracy and the behavior detection accuracy of the construction personnel in a complex power distribution network construction environment. Through continuous learning and training, the accuracy and efficiency of power distribution network construction personnel identification and behavior detection can be further improved.

[0130] Embodiment 2

[0131] The embodiment also provides a power distribution network construction personnel operation behavior monitoring system, which comprises:

[0132] a data set construction module, configured to construct a first data set and a second data set, the first data set being a personnel data set, and the second data set being a behavior data set;

[0133] a model training module, configured to train a first recognition model and a second recognition model according to the first data set and the second data set;

[0134] a model construction module, configured to take the output of the trained first recognition model as the input of the second recognition model, and establish a new third recognition model;

[0135] a monitoring module, configured to perform power distribution network construction personnel operation behavior monitoring according to the third recognition model.

[0136] The above modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above modules.

[0137] The embodiment also provides a computer device which can be a terminal, and an internal structure diagram of the computer device can be as shown in Figure 7 The computer device includes a processor, a memory, a communication interface, a display screen and an input device which are connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a power distribution network construction personnel operation behavior monitoring method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0138] The embodiment also provides a computer readable storage medium having a computer program stored thereon. The computer program is executed by the processor to implement the following steps:

[0139] constructing a first data set and a second data set, the first data set being a personnel data set, and the second data set being a behavior data set;

[0140] training a first recognition model and a second recognition model according to the first data set and the second data set;

[0141] The output of the first recognition model after training is taken as the input of the second recognition model, and a new third recognition model is established;

[0142] According to the third recognition model, the operation behavior of the power distribution network construction personnel is monitored.

[0143] Embodiment 3

[0144] Reference Figures 1-6 For an embodiment of the present application, a power distribution network construction personnel operation behavior monitoring method and system are provided. In order to verify the beneficial effects of the present application, scientific demonstration is carried out through experiments.

[0145] In this embodiment, 550 power distribution network site construction images are collected by using a high-definition monitoring camera to construct a power distribution network construction personnel dataset, and the positions of the power distribution network construction personnel in the power distribution network site construction images are labeled by using a LabelMe image labeling tool. Each image contains one or more power distribution network construction personnel position targets, and multiple anchor boxes need to be selected for the power distribution network construction personnel targets. After all the power distribution network construction personnel target labeling is completed, 550 xml format labeling files can be obtained. The name of each labeling file is consistent with the name of the corresponding power distribution network site construction image file. Each labeling file can contain multiple target labeling anchor box information and category information.

[0146] The public dataset is collected by roboflow, and then the power distribution network construction personnel behavior dataset is constructed. The dataset content includes power distribution network construction personnel behavior images, including the following sub-datasets: 376 pictures of power distribution network construction personnel safety helmet dataset, 234 pictures of power distribution network construction personnel work clothes dataset, and 1700 pictures of power distribution network construction personnel site bad construction behavior dataset, a total of 2310 pictures. “Power distribution network construction personnel without safety helmet”, “power distribution network construction personnel without work clothes”, “power distribution network construction personnel with bad construction behavior in construction site”.

[0147] The constructed power distribution network construction personnel dataset is divided into training set and validation set according to the ratio of 4:1, that is, 440 power distribution network construction personnel data training sets and 110 test sets. According to the power distribution network construction personnel recognition model designed in the present scheme, the power distribution network site construction image is trained, and the construction of the power distribution network construction personnel model is completed. First, the parameters and hyperparameters of algorithm training are initialized. In this embodiment, several important parameters in algorithm training are described: the initialization optimizer is AdaGrad optimizer, the initialization training epoch is 500, the batch_size is 32, and the initial learning rate is 0.0015. These parameters need to be adaptively adjusted according to the training effect of the algorithm multiple times until the algorithm modeling reaches the optimal effect.

[0148] The constructed power distribution network construction personnel behavior dataset is divided into a training set and a validation set according to a ratio of 4:1, that is, 1848 power distribution network construction personnel behavior training sets and 462 test sets. The power distribution network construction personnel model designed according to the scheme is trained on the power distribution network construction site image, and the construction of the power distribution network construction personnel model is completed. First, the parameters and hyperparameters of the algorithm training are initialized, and several important parameters in the algorithm training of the embodiment are described: the initialized optimizer is the Adam optimizer, the initialized training epoch is 500, the batch_size is 32, the initial learning rate is 0.001, and the weight initialization method is He initialization. These parameters need to be adaptively adjusted according to the effect of multiple algorithm training until the algorithm modeling reaches the optimal effect.

[0149] After the basic parameter initialization of the algorithm training is completed, the power distribution network construction personnel identification model is trained. The power distribution network construction site feature image is input into the power distribution network construction personnel identification model of the application, and it is assumed that the input is a power distribution network construction site feature image F1. The power distribution network construction site feature image F1 is input into the CBS module to obtain the power distribution network construction site feature image F2. F2 is input into the Gs_Block module for feature extraction to obtain the power distribution network construction site feature image F3. F3 is input into the Gs_Block module and the RFEM module to obtain the power distribution network construction site feature image F4. F4 is input into the Gs_Block module and the RFEM module to obtain the power distribution network construction site feature image F5. F5 is input into the Gs_Block module and the RFEM module to obtain the power distribution network construction site feature image F6. F6 is input into the SPPF to obtain the power distribution network construction site feature image F7. F7 is obtained through a linear interpolation operation to obtain the power distribution network construction site feature image F8. F5 and F8 are input into the FFM, and the result is input into the RFEM module to obtain the power distribution network construction site feature image F9. F9 is obtained through a linear interpolation operation to obtain the power distribution network construction site feature image F 10 . F4 and F 10 are input into the FFM, and the result is input into the RFEM module to obtain the power distribution network construction site feature image F 11 . F9 and F 11 are input into the FFM, and the result is input into the RFEM module to obtain the power distribution network construction site feature image F 12 . F8 and F 12 are input into the FFM to obtain the power distribution network construction site feature image F 13 . The power distribution network construction site feature image F 11 , the feature image F 12 , and the feature image F 13And input into the target detection head, the training of the detection algorithm of the application updates the parameters inside the algorithm through the back propagation of the loss function, the confidence loss of the bounding box uses binary cross-entropy loss, the above loss function is updated round by round in the gradient back propagation process of algorithm training, the algorithm gradually converges in the training round by round, and the detection accuracy is continuously improved. After all rounds of training are completed, the algorithm will perform double verification of accuracy and recall rate on the divided verification data, and use mAP50 as the evaluation standard for testing the effectiveness of algorithm training. Finally, the distribution network construction personnel recognition result is obtained, which includes the category of construction personnel in the image, the regression box and the confidence.

[0150] After the basic parameter initialization of the algorithm training is completed, the distribution network construction personnel behavior detection model is trained. Assuming that the input is a distribution network construction personnel feature map F1, the input size of the distribution network construction site feature map K1 of HxWxC is input into the upper and lower two different branches. In the upper branch, K1 is input into the AD-Conv hollow convolution layer with the number of convolution kernels C, the size of 3x3, the padding value of 2, and the step of 1, then the result is subjected to Avg-Pooling average pooling operation to obtain the distribution network construction personnel feature map K2 with the size of HxWxC. In the lower branch, K1 is input into the AD-Conv hollow convolution layer with the number of convolution kernels C, the size of 3x3, the padding value of 2, and the step of 1, then the result is subjected to Max-Pooling maximum pooling operation to obtain the distribution network construction personnel feature map K3 with the size of HxWxC. The distribution network construction personnel feature maps K2 and K3 are subjected to channel superposition operation, and the result is input into the AD-Conv hollow convolution layer with the number of convolution kernels C, the size of 3x3, the padding value of 2, and the step of 1 to obtain the distribution network construction personnel feature map K4 with the size of HxWxC. K4 is input into the Full-connection full connection layer for result calculation and classification. Finally, the distribution network construction personnel behavior detection result is output.

[0151] The training of the distribution network construction personnel behavior detection model of the application updates the internal parameters through the back propagation mechanism of the loss function, wherein the classification loss uses the cross-entropy loss function. In the algorithm training process, the model weight is updated step by step through gradient back propagation in each iteration, and as the number of rounds increases, the model gradually converges and the classification accuracy continuously improves. After all training rounds are completed, the algorithm will use the divided verification set data to perform double verification of accuracy and F1 score, and use the overall classification accuracy and average F1 score as the evaluation standard for testing the training effect of the algorithm. Finally, the model can output the behavior classification result, including the category label of the behavior in the image and the corresponding confidence.

[0152] After the power distribution network construction personnel identification model and the power distribution network construction personnel behavior detection model are trained, the models are connected, and the process is as follows: inputting the power distribution network site construction image into the power distribution network construction personnel identification model to identify the power distribution network construction personnel, and then inputting the power distribution network construction personnel identification result into the power distribution network construction personnel behavior detection model to detect the behavior of the power distribution network construction personnel.

[0153] The model is integrated into the existing power distribution network site monitoring system to ensure that it can receive image or video data from the power distribution network site area in real time. In the system, an alarm threshold of 0.6 is set in the system to capture the confidence of the behavior of the power distribution network construction personnel. Once the model detects that the power distribution network construction personnel has bad behavior and the confidence exceeds the preset alarm threshold, the system will automatically trigger the alarm mechanism. The mechanism will record the specific time of the behavior, the identity and position of the construction personnel involved in the power distribution station, and other key details, and then immediately inform the on-duty personnel, relevant team managers and safety personnel through multiple channels such as email, SMS or system notification, to ensure timely identification and response to potential safety risks. This can effectively improve the supervision of the behavior of the power distribution network construction personnel, avoid safety problems caused by behavior, and further ensure the stable operation and safety of the power distribution network.

[0154] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

[0155] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages, such as object-oriented programming languages Java and interpreted scripting language JavaScript.

[0156] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0157] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0158] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0159] While the preferred embodiments of the application have been described, additional variations and modifications can be employed by those skilled in the art. Therefore, the appended claims intend to cover all such modifications and variations as fall within the true spirit and scope of the application.

[0160] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A method of monitoring the behavior of a power distribution network construction worker, characterized by, The method comprises the following steps: constructing a first data set and a second data set, the first data set being a personnel data set, and the second data set being a behavior data set; training a first recognition model and a second recognition model according to the first data set and the second data set; establishing a new third recognition model by taking the output of the trained first recognition model as the input of the second recognition model; conducting power distribution network construction personnel operation behavior monitoring according to the third recognition model; the first recognition model comprises a feature extraction module arranged in the first recognition model; the feature extraction module is used to generate a mirror feature image and expand the detail feature receptive field of the power distribution network construction site image to capture the key information in the image; the first recognition model further comprises a feature receptive field enhancement module arranged in the first recognition model; the feature receptive field enhancement module is used to associate each pixel point in the power distribution network construction site image to improve the global receptive field of each pixel point in the feature image; the first recognition model further comprises a feature fusion module arranged in the first recognition model; the feature fusion module is used to extract features in the image through double-branch power distribution network construction site image input, fuse the double-branch power distribution network construction site features through superposition operation, and increase the randomness of image features through shuffling operation.

2. The power distribution network worker behavior monitoring method of claim 1, wherein, The training of the first recognition model and the second recognition model according to the first data set and the second data set comprises: the first recognition model is any model with the first data set as input and the personnel recognition result or the personnel recognition result related parameters as output; the second recognition model is any model with the second data set as input and the personnel behavior result or the personnel behavior result related parameters as output.

3. The power distribution network worker behavior monitoring method of claim 2, wherein, The third recognition model comprises: the third recognition model is any model with the output of the first recognition model as input and the personnel behavior result or the personnel behavior result related parameters as output.

4. The power distribution network worker behavior monitoring method of claim 3, wherein, The power distribution network construction personnel operation behavior monitoring according to the third recognition model comprises: pre-setting a confidence threshold and comparing the confidence threshold with the output of the third recognition model; conducting power distribution network construction personnel operation behavior monitoring according to the comparison result.

5. A power distribution network construction worker behavior monitoring system for use in the method of claim 1, wherein The method comprises the following steps: a data set construction module is used to construct a first data set and a second data set, the first data set being a personnel data set, and the second data set being a behavior data set; a model training module is used to train a first recognition model and a second recognition model according to the first data set and the second data set; a model construction module is used to establish a new third recognition model by taking the output of the trained first recognition model as the input of the second recognition model; a monitoring module is used to conduct power distribution network construction personnel operation behavior monitoring according to the third recognition model. 6.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-5 when the computer program is executed by the processor. The processor executes the computer program to realize the steps of the method in any one of claims 1 to 4.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method in any one of claims 1 to 4.

Images