Power distribution room video monitoring method and system based on deep learning

Abstract

The invention provides a power distribution room video monitoring method and system based on deep learning, and the method comprises the following steps: carrying out the training based on visual information, and obtaining a deep learning model; and detecting, identifying and tracking a monitoring target by using the deep learning model to obtain a final video monitoring result. According to the technical scheme, unmanned, intelligent, standardized and processized analysis processing of the distribution room inspection image data is realized, the objectivity of data analysis is improved, and the efficiency of inspection business is improved.

Description

Background Art

[0002] With the rapid development of the modern national economy, most power companies are constantly accelerating the construction of smart grids, with the goal of building a core smart distribution network in the city and being able to serve economic and social development more comprehensively. The operation and inspection of the 10kV distribution room of the 66kV substation is the main link at the end of the power system and is an important basis and data source for distribution network management. The maintenance and testing of equipment depend on the daily inspection results of the equipment status, and the planning of the distribution room is closely linked to the analysis of the equipment status. There are a large number of equipment, a complex environment, and many inspection items. At present, the 10kV distribution room of the 66kV substation is still in a highly dependent stage, with a low level of intelligent operation and inspection. There is a lack of effective monitoring of hidden dangers and faults that affect the safe operation of the distribution room, and manual inspections require a lot of manpower, material and financial resources. The main problems are as follows:

[0003] (1) The information acquisition method is traditional and the data source is single. Currently, only staff members arrive at the site to inspect, measure temperature, and measure humidity. There are thermometers and hygrometers and exhaust fans in the switch stations and substations, and water pump facilities in some locations, all of which are manually monitored and started.

[0004] (2) Equipment status perception is still mainly based on power outage maintenance and offline testing. At present, the operation and maintenance of 10kV distribution rooms is heavy, inefficient, prone to errors, and has high operation and maintenance costs. The maintenance and testing of distribution rooms cannot be performed while conducting on-site operations.

[0005] (3) Data is recorded manually, which is not conducive to archiving. Manually recorded data cannot be uploaded in real time and is limited to the on-site environment. When problems arise, it is impossible to view the internal structure of the equipment and the equipment ledger in detail. This makes it difficult to perceive the equipment status in real time and conduct online centralized monitoring. When encountering emergencies, the inspection personnel are slow to deploy, which can easily cause major accidents.

[0006] (4) Lack of advanced means such as online monitoring. It mainly relies on manual inspections, which can only rely on experience and superficial inspections, resulting in many blind spots, accumulation of hidden dangers, lack of real-time alarms, and low alarm efficiency.

[0007] At present, the research on power distribution room operation and maintenance at home and abroad mainly makes full use of multi-source data such as equipment status, power grid operation and environmental meteorology for deep fusion analysis and machine learning to improve the real-time and accuracy of equipment status evaluation, fault diagnosis and prediction. Through the integration of artificial intelligence technology and equipment status diagnosis, the application of intelligent equipment evaluation and diagnosis has been significantly accelerated. Many research teams at home and abroad take practical problems as the starting point and use some advanced deep learning technology, video recognition technology, big data technology and other technical means, such as neural networks, environmental perception, wavelet analysis, expert systems, fuzzy diagnosis, pattern recognition, etc., to apply to status evaluation, prediction and defect diagnosis, so that the status detection and evaluation technology of power equipment in the power distribution room has developed rapidly.

[0008] Deep learning is to learn the internal laws and representation levels of sample data. The information obtained during these learning processes is very helpful for the interpretation of data such as text, images, and sounds. Its ultimate goal is to enable machines to have the ability of analytical learning like humans, and be able to recognize data such as text, images, and sounds. In practice, deep learning algorithms are adopted to automatically obtain the real-time distribution network topology structure and combine with device historical information, such as device fault records, maintenance records, trip records, etc. to evaluate the power supply reliability of the distribution room. At the same time, considering the importance level of the power supply to the receiving users, it is decided whether to change the operation mode of the system, and the optimal strategy can also be given based on the above deep learning algorithm to perform fault isolation and self-healing of the distribution room. Summary of the Invention

[0009] This application provides a distribution room video monitoring method and system based on deep learning to achieve unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the distribution room, improve the objectivity of data analysis, and improve the efficiency of the inspection business.

[0010] In the first aspect, a distribution room video monitoring method based on deep learning is provided, including the following steps:

[0011] Based on visual information, a deep learning model is trained.

[0012] The deep learning model is used to detect, identify, and track the monitoring target to obtain the final video monitoring result.

[0013] In the above technical solution, by training a deep learning model based on visual information, and using the deep learning model to detect, identify, and track the monitoring target to obtain the final video monitoring result, the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the distribution room are realized, the objectivity of data analysis is improved, and the efficiency of the inspection business is improved.

[0014] In a specific feasible implementation, it further includes:

[0015] A video sensing device and a computer are used to simulate the human visual system to collect and process the visual information.

[0016] In a specific feasible implementation, the deep learning model includes an AlexNet network model based on deep learning.

[0017] In a specific feasible implementation, the final video monitoring result includes the fault location and type identification of the distribution room equipment.

[0018] In the second aspect, a distribution room video monitoring system based on deep learning is provided, including:

[0019] A model construction module for training a deep learning model based on visual information;

[0020] A detection and recognition module for detecting, recognizing, and tracking a monitoring target using the deep learning model to obtain a final video monitoring result.

[0021] In the above technical solution, a deep learning model is trained based on visual information; the deep learning model is used to detect, recognize, and track a monitoring target to obtain a final video monitoring result; realizing unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room, improving the objectivity of data analysis and the efficiency of the inspection operation.

[0022] In a specific feasible implementation, it further includes:

[0023] A data acquisition module for collecting and processing the visual information by using a video sensing device and a computer to simulate the human visual system.

[0024] In a specific feasible implementation, the deep learning model includes an AlexNet network model based on deep learning.

[0025] In a specific feasible implementation, the final video monitoring result includes fault location and type identification of the power distribution room equipment.

[0026] In a third aspect, an electronic device is provided, the electronic device includes a processor, the processor is coupled with a memory, and at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor so that the electronic device implements any one of the above-mentioned power distribution room video monitoring methods based on deep learning.

[0027] In the above technical solution, a deep learning model is trained based on visual information; the deep learning model is used to detect, recognize, and track a monitoring target to obtain a final video monitoring result; realizing unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room, improving the objectivity of data analysis and the efficiency of the inspection operation.

[0028] In a fourth aspect, a computer-readable storage medium is provided, and at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by a processor so that the computer-readable storage medium implements any one of the above-mentioned power distribution room video monitoring methods based on deep learning.

[0029] In the above technical solution, a deep learning model is trained based on visual information; the deep learning model is used to detect, identify, and track monitoring targets to obtain the final video monitoring result; the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room are realized, the objectivity of data analysis is improved, and the efficiency of the inspection operation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a flowchart of the power distribution room video monitoring method based on deep learning provided by an embodiment of the present application;

[0031] Figure 2 It is a block diagram of the structure of the power distribution room video monitoring system based on deep learning provided by an embodiment of the present application;

[0032] Figure 3 It is a schematic diagram of the structure of the AlexNet network model based on deep learning provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The present application will be further described in detail below with reference to the drawings and embodiments. Through these descriptions, the features and advantages of the present application will become more clearly defined.

[0034] The special term "exemplary" herein means "serving as an example, embodiment, or illustrative". Any embodiment described herein as "exemplary" is not necessarily to be construed as superior or better than other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings do not have to be drawn to scale unless otherwise specified.

[0035] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0036] To facilitate the understanding of the deep learning-based power distribution room video monitoring method and system provided by the embodiments of the present application, the application scenario will be described first. The deep learning-based power distribution room video monitoring method and system provided by the embodiments of the present application are used to realize the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room, improve the objectivity of data analysis, and improve the efficiency of the inspection business. At present, the 10kV power distribution room of the 66kV substation is still highly dependent on manual work, with a low degree of intelligent operation and maintenance, and there is a lack of effective monitoring of potential faults that affect the safe operation of the power distribution room. Manual inspections require a large amount of manpower, material resources, and financial resources. The main problems are as follows: (1) The information acquisition method is traditional and the data source is single. Currently, only staff members arrive at the scene for inspections, temperature measurements, humidity measurements, etc. There are temperature and humidity meters and exhaust fans in the switchgear station and substation, and there are water pump facilities at individual locations, all of which rely on manual monitoring and startup. (2) The equipment status perception still mainly relies on power-off maintenance and off-line tests. Currently, the operation and maintenance workload of the 10kV power distribution room is large, the efficiency is low, it is easy to make mistakes, and the operation and maintenance cost is high. Information access cannot be carried out while the power distribution room is being repaired and tested on-site. (3) Manual data recording is not conducive to archiving. The manually recorded data cannot be uploaded in real time, is limited by the on-site environment, and the internal structure and equipment ledger of the equipment cannot be viewed in detail when problems occur. This makes it difficult to perceive the equipment status in real time and conduct online centralized monitoring. When encountering emergencies, the deployment speed of inspection personnel is slow, which is likely to cause major accidents. (4) There is a lack of advanced means such as online monitoring. Mainly relying on manual inspections, only experience can be relied on during the inspection process, and only the surface can be seen, resulting in many blind spots and large accumulations of potential hazards, lack of real-time alarms, and low alarm efficiency. Therefore, the embodiments of the present application provide a deep learning-based power distribution room video monitoring method and system to realize the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room, improve the objectivity of data analysis, and improve the efficiency of the inspection business. The following will be described in detail with specific drawings by way of examples.

[0037] Reference Figures 1 to 3 , Figure 1 is the flow block diagram of the deep learning-based power distribution room video monitoring method provided by the embodiments of the present application; Figure 2 is the structural block diagram of the deep learning-based power distribution room video monitoring system provided by the embodiments of the present application; Figure 3 is the schematic diagram of the structure of the AlexNet network model provided by the embodiments of the present application.

[0038] In Figure 1 the embodiments of the present application provide a deep learning-based power distribution room video monitoring method, including the following steps:

[0039] Based on visual information, a deep learning model is trained;

[0040] Use the deep learning model to detect, identify, and track the monitoring target to obtain the final video monitoring result.

[0041] In the above technical solution, a deep learning model is trained based on visual information; the deep learning model is used to detect, identify, and track the monitoring target to obtain the final video monitoring result; realizing unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room, improving the objectivity of data analysis, and improving the efficiency of the inspection business.

[0042] Specifically, during the operation and inspection process of the current 10kV power distribution room, the monitoring and video systems generally rely on manual analysis and lack intelligence. For a small number of local points with video intelligent analysis, the analysis algorithms are not flexible, one function for one set of hardware, and the video global networking ability is insufficient. To solve the above problems, an intelligent video monitoring method based on deep learning technology is proposed, which specifically includes: detecting, analyzing, filtering interference, etc. of the operation and inspection video images based on deep learning technology, researching intelligent video monitoring methods, and performing intelligent analysis and accurate identification on the facilities and equipment of the power distribution room video monitoring, realizing functions such as motion detection, automatic upload of video images, remote monitoring and analysis of the power distribution room video system.

[0043] In the above technical solution, a deep learning model is trained based on visual information and applied to the detection, identification, and tracking of monitoring targets to realize unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room. The beneficial effects include:

[0044] Improve the objectivity of data analysis: The deep learning model can learn and extract features based on a large amount of training data, avoiding the subjectivity and errors that may be brought by manual analysis.

[0045] Through an automated analysis process, it can ensure that the data analysis results of each inspection are obtained based on the same standards and algorithms, thereby improving the objectivity and consistency of the data.

[0046] Improve the efficiency of the inspection business: Automated inspections can significantly reduce the time and labor costs required for manual inspections.

[0047] The deep learning model can quickly process a large amount of image data and provide inspection results in real time or near real time, enabling a shorter inspection cycle and a faster response speed.

[0048] Realize unmanned inspections: The introduction of the deep learning model enables inspections to be carried out without personnel on duty, reducing the need for and dependence on personnel. This is particularly important for inspections of power distribution rooms in remote or dangerous environments and can ensure the safety of personnel.

[0049] Intelligent analysis and processing: The deep learning model has powerful learning and reasoning capabilities, and can accurately identify various equipment and abnormal situations in the power distribution room.

[0050] Through continuous learning and optimization, the recognition accuracy and generalization ability of the model will be continuously improved, so as to better adapt to various complex inspection scenarios.

[0051] Standardized and process-based processing: The automated inspection process can ensure that each inspection follows the same steps and standards, thus improving the standardization and consistency of inspections.

[0052] This helps to establish a standardized inspection report and data analysis system, facilitating subsequent maintenance and management.

[0053] Enhanced fault warning and emergency response capabilities: By real-time monitoring and analyzing the image data of the power distribution room, the deep learning model can timely detect potential faults or abnormal situations and issue warning signals. This helps the operation and maintenance personnel to take measures in advance to avoid the occurrence or expansion of faults, thus ensuring the safe operation of the power distribution room.

[0054] Improved data traceability and analyzability: The automated inspection system can record the image data and analysis results of each inspection, facilitating subsequent data traceability and analysis. This helps the operation and maintenance personnel to understand the historical operating status of the power distribution room and provide data support for future maintenance and optimization.

[0055] In summary, through the application of the deep learning model based on visual information in the inspection of the power distribution room, the objectivity of data analysis is significantly improved, the inspection business efficiency is enhanced, unmanned inspection, intelligent analysis and processing, standardized and process-based processing are realized, and the fault warning and emergency response capabilities are enhanced; jointly promoting the modernization and intelligentization process of the inspection work in the power distribution room.

[0056] In a specific feasible implementation, it further includes:

[0057] Using video sensing devices and computers to simulate the human visual system to collect and process the visual information.

[0058] In a specific feasible implementation, the deep learning model includes the AlexNet network model based on deep learning.

[0059] Specifically, the AlexNet network model is a convolutional neural network of milestone significance in the field of deep learning, including:

[0060] I. Network hierarchy

[0061] AlexNet consists of 8 weight layers (including 5 convolutional layers and 3 pooling layers) and 3 fully connected layers. These layers, through specific connection methods and parameter configurations, jointly achieve efficient processing of images.

[0062] Input layer: Usually receives images of a fixed size as input, such as images of 224x224x3 (width / height / number of channels) or 32x32x3. These images need to be preprocessed, such as normalized, before being input into the network.

[0063] Convolutional layers: AlexNet contains multiple convolutional layers. Each convolutional layer uses multiple convolutional kernels to perform convolution operations on the input image to extract features in the image. The size, stride, and number of output channels of these convolutional kernels are designed according to specific tasks. For example, the first convolutional layer uses a convolutional kernel of size 11x11, a stride of 4, and the number of output channels is 96. Subsequent convolutional layers use smaller convolutional kernels (such as 3x3) and more output channels.

[0064] Pooling layers: After the convolutional layers, pooling layers are usually connected. The main role of the pooling layer is to perform downsampling on the output of the convolutional layer to reduce the dimension of the data and extract the main features. The main pooling operation used in AlexNet is the max pooling operation, and the pooling window size and stride are usually set to 2x2.

[0065] Fully connected layers: After the convolutional and pooling layers, AlexNet contains three fully connected layers. These fully connected layers map the features extracted by the convolutional layers to the final output classes. Each fully connected layer contains a large number of neurons. These neurons are connected to the output of the previous layer through weights and use activation functions for non-linear transformation.

[0066] II. Network Features

[0067] Multi-GPU parallel computing: A significant feature of AlexNet is its use of two GPUs for parallel computing. This design not only improves the computing speed of the network but also enables the network to process larger-scale data and more complex models.

[0068] Data augmentation: To improve the generalization ability of the model, AlexNet uses data augmentation techniques during training. This includes operations such as random cropping, mirror flipping, and color perturbation to increase the diversity of training data and reduce the risk of overfitting.

[0069] ReLU activation function: AlexNet widely uses the ReLU (Rectified Linear Unit) activation function in convolutional and fully connected layers. The ReLU function has a simple form and efficient computational performance, and can effectively alleviate the problem of gradient vanishing.

[0070] Dropout Technique: After the fully connected layer, AlexNet also uses the Dropout technique to prevent overfitting. Dropout reduces the model's dependence on the training data by randomly discarding the outputs of some neurons during training, thereby improving the model's generalization ability.

[0071] In summary, through its unique hierarchical structure, multi-GPU parallel computing, data augmentation, ReLU activation function, and Dropout technique, the AlexNet network model achieves efficient processing and accurate classification of image data.

[0072] In a specific feasible implementation, the final video surveillance result includes the fault location and type identification of the power distribution room equipment.

[0073] In Figure 2 and Figure 3 this application embodiment provides a power distribution room video surveillance system based on deep learning, including:

[0074] A model construction module, used to train a deep learning model based on visual information;

[0075] A detection and recognition module, used to detect, recognize, and track the monitoring target using the deep learning model to obtain the final video surveillance result.

[0076] In the above technical solution, by training a deep learning model based on visual information, and using the deep learning model to detect, recognize, and track the monitoring target to obtain the final video surveillance result, it realizes the unmanned, intelligent, standardized, and process-based analysis and processing of the power distribution room inspection image data, improves the objectivity of data analysis, and improves the efficiency of the inspection operation.

[0077] In a specific feasible implementation, it further includes:

[0078] A data acquisition module, used to collect and process the visual information by using video sensing devices and computers to simulate the human visual system.

[0079] In a specific feasible implementation, the deep learning model includes an AlexNet network model based on deep learning.

[0080] Specifically, the construction steps of the AlexNet network model based on deep learning include:

[0081] I. Model Architecture Design

[0082] The AlexNet model consists of five convolutional layers and three fully connected layers. When building the model, first define the initialization function (init) and the forward propagation function (forward).

[0083] Initialization function (init):

[0084] In the initialization function, each sub-module required in forward propagation is defined, including convolutional layers, pooling layers, activation layers, etc.

[0085] Convolutional layers are used to extract image features, pooling layers are used to compress the computational amount, and activation layers are used to increase the non-linearity of the model.

[0086] Forward propagation function (forward):

[0087] Call the first defined module (convolutional layer + pooling layer + activation layer) to obtain features.

[0088] Unroll the previous layer of the fully connected layer into a one-dimensional vector because the input of the fully connected layer must be one-dimensional.

[0089] Call the second defined module (fully connected layer + activation layer + Dropout layer) for classification.

[0090] II. Weight Initialization

[0091] Weight initialization in deep learning has an important impact on the model convergence speed and model quality. In AlexNet, corresponding initial values need to be set for each parameter. Common initialization methods include random initialization, zero initialization, etc., and the most suitable initialization method is selected according to the specific problem and model structure.

[0092] III. Data Preparation and Preprocessing

[0093] Data download: Download image data from a public dataset (such as ImageNet).

[0094] Data preprocessing: Preprocess the image data, including resizing the image, normalizing, etc.

[0095] Data augmentation: Use data augmentation methods (such as flipping, cropping, color transformation, etc.) to expand the dataset and alleviate the overfitting phenomenon.

[0096] IV. Model Training

[0097] Instantiate the network model: Create an instance of the AlexNet model.

[0098] Define the loss function and optimizer: Select a suitable loss function (such as cross-entropy loss) and optimizer (such as SGD) to train the model.

[0099] Training process:

[0100] Enable the training mode (model.train()), and enable Batch Normalization and Dropout.

[0101] Iterate through the training set, perform forward propagation and backward propagation, and update the model parameters.

[0102] At the end of each epoch, use the validation set to calculate the accuracy of the model.

[0103] V. Model evaluation and testing

[0104] Instantiate the model: Create an instance of the AlexNet model for testing.

[0105] Load the weight file: Load the weight file saved during the training process.

[0106] Model evaluation: Use the test set to evaluate the model and calculate metrics such as accuracy.

[0107] Turn off Dropout: During testing, it is necessary to turn off Dropout (model.eval()) to ensure that all network connections are utilized.

[0108] VI. Supplementary knowledge points

[0109] Batch Normalization: During the training process, Batch Normalization can utilize the mean and variance of each batch of data, which helps the model converge.

[0110] Dropout: Dropout is a regularization technique that prevents overfitting by randomly discarding a portion of the neural network units during the training process.

[0111] Activation function: The ReLU activation function is used in AlexNet. Compared with the sigmoid function, ReLU has advantages such as less computational complexity, less gradient vanishing problem, and sparsity.

[0112] In a specific implementable embodiment, the final video surveillance result includes fault location and type identification of the power distribution room equipment.

[0113] Specifically, the AlexNet network model based on deep learning is adopted because the power operation and inspection field mainly deals with image and video data. Deep learning technology uses video sensing devices and computers to simulate the human visual system, collect and process external visual information. Digital signals such as images, videos, or multi-dimensional data are collected through imaging devices such as cameras as information input, and then the computer processes this information to achieve functions such as target detection, recognition, and tracking, and finally obtains judgments and explanations that meet the requirements. The ImageNet dataset contains a vast number of high-resolution pictures of various types. Combining with the effective overfitting technology of AlexNet, the convolutional neural network model is fully trained, greatly improving the image recognition accuracy and generalization performance of the convolutional neural network. In this embodiment, the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data are realized, improving the objectivity of data analysis, enhancing the efficiency of the inspection operation, quickly realizing the fault location and type recognition of the power distribution room equipment, ensuring the safe and stable operation of the power system, reducing the operation and maintenance costs, and improving the operation and inspection efficiency.

[0114] The embodiment of the present application also provides an electronic device, which includes a processor. The processor is coupled with a memory, and at least one computer program is stored in the memory. The at least one computer program is loaded and executed by the processor to enable the electronic device to implement any one of the above-mentioned deep learning-based power distribution room video monitoring methods.

[0115] In the above technical solution, a deep learning model is trained based on visual information; the deep learning model is used to detect, recognize, and track the monitoring target to obtain the final video monitoring result; the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room are realized, improving the objectivity of data analysis and enhancing the efficiency of the inspection operation.

[0116] The embodiment of the present application also provides a computer-readable storage medium, in which at least one computer program is stored. The at least one computer program is loaded and executed by a processor to enable the computer-readable storage medium to implement any one of the above-mentioned deep learning-based power distribution room video monitoring methods.

[0117] In the above technical solution, a deep learning model is trained based on visual information; the deep learning model is used to detect, recognize, and track the monitoring target to obtain the final video monitoring result; the unmanned, intelligent, standardized, and process-based analysis and processing of the inspection image data of the power distribution room are realized, improving the objectivity of data analysis and enhancing the efficiency of the inspection operation.

[0118] Those skilled in the art of the technical field know that the present application can be implemented as a system, a method, or a computer program product.

[0119] Accordingly, the present disclosure may be embodied in the following forms, namely: it may be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit", "module", or "system". Additionally, in some embodiments, the present application may also be implemented in the form of a computer program product in one or more computer-readable media, which contain computer-readable program code.

[0120] Any combination of one or more computer-readable media may be employed. The computer-readable media may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this document, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0121] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application. On this basis, various substitutions and improvements can be made to the present application, and all of these fall within the protection scope of the present application.

Claims

1. A distribution room video monitoring method based on deep learning, characterized in that: The following steps are involved: Based on visual information, a deep learning model is trained; The deep learning model is used to detect, identify and track the monitoring target to obtain the final video monitoring result.

2. The method for video monitoring of a power distribution room based on deep learning according to claim 1 is characterized in that: Also includes: The visual information is collected and processed by using video sensing equipment and computers to simulate the human visual system.

3. The method for video monitoring of a power distribution room based on deep learning according to claim 2 is characterized in that: The deep learning model includes an AlexNet network model based on deep learning.

4. The method for video monitoring of a power distribution room based on deep learning according to claim 3 is characterized in that: The final video monitoring result includes fault location and type identification of the equipment in the power distribution room.

5. A distribution room video monitoring system based on deep learning, characterized in that: include: Model building module, used to train deep learning models based on visual information; The detection and recognition module is used to detect, identify and track the monitoring target using the deep learning model to obtain the final video monitoring result.

6. The power distribution room video monitoring system based on deep learning according to claim 5 is characterized in that: Also includes: The data acquisition module is used to utilize video sensing equipment and computers to simulate the human visual system and to collect and process the visual information.

7. The power distribution room video monitoring system based on deep learning according to claim 6 is characterized in that: The deep learning model includes an AlexNet network model based on deep learning.

8. The power distribution room video monitoring system based on deep learning according to claim 7 is characterized in that: The final video monitoring result includes fault location and type identification of the equipment in the power distribution room.

9. An electronic device, characterized in that: The electronic device includes a processor, which is coupled to a memory, in which at least one computer program is stored, and the at least one computer program is loaded and executed by the processor so that the electronic device implements the deep learning-based video monitoring method for a distribution room as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that: At least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by the processor so that the computer-readable storage medium implements the deep learning-based distribution room video monitoring method as described in any one of claims 1 to 4.

Images