A visual intelligent identification device for hidden dangers in power channels

Abstract

The present application discloses a visual power channel hidden danger intelligent identification device, including: a high-definition image acquisition module, used to acquire images of the power channel; an image intelligent identification module, used to store the images and perform hidden danger intelligent identification on the images to obtain hidden danger information; a data transmission module, used to transmit the images and the hidden danger information to a remote terminal server; a power supply module, used to supply power to the high-definition image acquisition module, the image intelligent identification module and the data transmission module; and a remote configuration and upgrade module, used to provide remote parameter configuration, algorithm update and software upgrade to the high-definition image acquisition module, the image intelligent identification module, the data transmission module and the power supply module. The present application can provide a basic guarantee for the safe, reliable and stable operation of power lines and promote the innovative development of online monitoring technology for power lines.

Description

Technical Field

[0001] The present application relates to the field of power line monitoring, and in particular to a visual intelligent identification device for hidden dangers in power channels. Background Art

[0002] With the continuous expansion of the scale of the power system and the increasing complexity of the grid structure, the safe operation of power lines has become a key link in ensuring the stability of power supply. Hidden dangers such as wildfires, mechanical damage, and foreign objects within the power channel may affect the safe operation of the power line. Therefore, the hidden danger inspection and monitoring of the power channel is an important link in ensuring the stable operation of the power system. Traditional power channel hidden danger identification mainly relies on manual inspection, with professionals patrolling each section of the power line one by one to observe and record the existing safety hazards. However, manual patrols are not only inefficient, but also have low accuracy and timeliness. It is impossible to patrol power lines in some remote areas and harsh environments.

[0003] With the development of technology, although some modern power channel monitoring methods and equipment have emerged, such as aerial photography and infrared observation equipment, these devices still require manual complex data analysis and judgment. In addition, these devices are complex to maintain and have high energy consumption, which also limits their widespread application.

[0004] Therefore, there is an urgent need for an intelligent identification device for hidden dangers in power channels that can integrate advanced visual perception technology, efficient energy management and remote intelligent management functions to achieve real-time and accurate hidden danger monitoring, improve the efficiency and effectiveness of power line maintenance, reduce manpower input, reduce operating costs, and promote the sustainable development of the power industry. Based on this background, the present invention proposes a visual intelligent identification device for hidden dangers in power channels that integrates high-definition image acquisition, intelligent identification, efficient data transmission, composite energy supply and remote management, aiming to provide strong technical support for the safe operation of power lines. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the present application provides a visual intelligent identification device for hidden dangers in power channels, which can combine a self-supervised pre-training model, an artificial intelligence recognition algorithm, and photovoltaic power generation to realize intelligent detection and identification of safety hazards existing in the power channel by a visual device, and provide real-time warnings.

[0006] In order to achieve the above objectives, this application adopts the following technical solutions:

[0007] 1A visual intelligent identification device for hidden dangers in power channels, comprising:

[0008] A high-definition image acquisition module, used for acquiring images of the power channel;

[0009] An image intelligent recognition module, coupled to the high-definition image acquisition module, for storing the image and performing intelligent hidden danger recognition on the image to obtain hidden danger information;

[0010] A data transmission module, coupled to the image intelligent recognition module, for transmitting the image and the hidden danger information to a remote terminal server;

[0011] A power supply module, coupled to the high-definition image acquisition module, the image intelligent recognition module and the data transmission module respectively, for supplying power to the high-definition image acquisition module, the image intelligent recognition module and the data transmission module;

[0012] And a remote configuration and upgrade module, which is respectively coupled to the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module, and is used to provide remote parameter configuration, algorithm update and software upgrade for the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module.

[0013] Furthermore, the image intelligent recognition module includes:

[0014] An image storage unit is used to store the image; the image storage unit is a multi-level cache structure, and through an asynchronous data storage mechanism, when the high-definition image acquisition module acquires an image, the image stream is synchronously transmitted to the image storage unit and stored in layers; the image storage unit also includes a compression and encryption mechanism, and the image is adaptively compressed and AES encrypted before storage;

[0015] and a hidden danger intelligent identification unit, which is used to detect the image according to a self-supervised pre-training model and an artificial intelligence recognition algorithm to obtain the hidden danger information, and display the hidden danger information on the image in the form of a marked box, wherein the self-supervised pre-training model includes contrast learning, image restoration and time series prediction tasks; and the artificial intelligence recognition algorithm includes a convolutional neural network;

[0016] In the process of hidden danger identification, the convolutional neural network locates hidden dangers through the multi-scale convolution kernel in the convolution kernel structure. When the characteristics of the hidden danger are detected, a labeling box of the hidden danger is automatically generated, and the hidden danger position is displayed in the image in a highlighted form; the size and position of the labeling box are dynamically adjusted according to the hidden danger area in the image, and the text information of the hidden danger is attached.

[0017] Furthermore, the data transmission module includes:

[0018] An image compression and encryption unit, used to compress the image marked with hidden danger information and encrypt the compressed image data;

[0019] Multi-path communication interface, used to integrate multiple communication methods and select the best path for data transmission;

[0020] And a real-time status feedback unit is used to monitor the working status and network connection quality information of the data transmission module in real time, and transmit the working status and network connection quality information to the remote terminal server.

[0021] Furthermore, the power supply module includes:

[0022] A composite energy management unit, including a solar photovoltaic panel and a lithium-ion battery pack;

[0023] And a power monitoring unit, which is used to dynamically detect the energy consumption and remaining power of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module; the power monitoring unit monitors the power consumption of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module in real time through an integrated sensor, and the sensor dynamically captures the instantaneous power of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module, and determines the current remaining power by calculating the cumulative power consumption, and predicts the remaining power in combination with the discharge curve of the lithium-ion battery; the power monitoring unit also includes an abnormal power consumption alarm module, which generates an alarm and uploads it to the remote terminal server when abnormal power consumption of one or more of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module is detected.

[0024] Furthermore, the remote configuration and upgrade module includes:

[0025] A wireless upgrade unit, used to receive push software and algorithm models, and update and upgrade the high-definition image acquisition module, the image intelligent recognition module, the data transmission module, and the power supply module;

[0026] A configuration management unit, used to remotely set the working parameters of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module, view device status reports and logs, and roll back to a previous software version;

[0027] and a security authentication and data integrity check unit for identity authentication and encrypted communication protocols.

[0028] Furthermore, the data transmission module is connected to the remote terminal server via an operator's wireless network to perform remote data transmission.

[0029] Furthermore, the hidden danger intelligent identification unit has a built-in deep learning algorithm and a graphics processor, and uses the power channel environment image for learning and training; the learning and training include:

[0030] Self-supervised learning is performed through unlabeled data sets. Image restoration, contrastive learning, and time-series consistency prediction tasks are used to automatically learn and extract hidden danger features in the power channel environment, and pre-train image data collected at different time periods and under different lighting conditions.

[0031] Combine the information of different bands in the image and adaptively adjust the weight of feature extraction in night or daytime environments; and optimize the deep learning model of the convolution kernel structure.

[0032] Furthermore, the hidden danger intelligent identification unit analyzes the image through a self-supervised pre-training model and a deep learning algorithm, identifies the flame characteristics and the smoke diffusion pattern, converts the identification result into early warning information, and generates an image mark in the form of a hidden danger annotation box, wherein the identified flame characteristics and the smoke diffusion pattern include:

[0033] Collect multiple frames of images at the same point and at different times, perform feature embedding mapping on the collected multiple frames of images, and extract stable and distinguishing features by maximizing the similarity of images of the same type and minimizing the differences of images of different types, so as to obtain the difference features existing in the image, including the dynamic jump of flames or the diffusion path of smoke;

[0034] In the image restoration task, some areas are randomly occluded on the unlabeled image, and the occluded area is inferred through a multi-scale convolution kernel structure to generate a predicted feature map of the missing area, which is then aligned with the unoccluded feature map, and the occluded area is restored through iterative learning;

[0035] Dynamic features such as flame edges and smoke diffusion are extracted from continuous multi-frame images, changes in the image sequence are captured through convolution operations, and changes in each frame are analyzed through a temporal convolutional network to capture long-term hidden danger feature trends, including the continued expansion of flames or the spread speed of smoke. The information of past frames is processed using a long short-term memory network to predict the flame spread direction or smoke spread range in the next few frames.

[0036] Furthermore, the high-definition image acquisition module includes a visible light camera and an infrared camera, and the visible light camera and the infrared camera can switch automatically.

[0037] Furthermore, the lithium-ion battery pack is used to power the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module; the solar photovoltaic panel is used to charge the lithium-ion battery pack.

[0038] Compared with the prior art, this application has the following advantages:

[0039] The visualized power channel hidden danger intelligent identification device of the present application combines a self-supervised pre-training model, an artificial intelligence recognition algorithm, and photovoltaic power generation to realize the device's intelligent detection and identification of safety hazards within the power channel, and real-time warning, making hidden danger identification more accurate and efficient. The accuracy and speed of power channel hidden danger identification can be enhanced through self-supervised pre-training and artificial intelligence algorithms. The integrated solar photovoltaic panels achieve energy self-sufficiency and reduce dependence on external power supply. The remote management function improves the maintainability and upgrade convenience of the system, ensures the safety and reliability of power lines, and promotes the sustainable development of the power industry. Overall, it achieves the purpose of promoting the intelligent process of power line monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application but do not constitute an improper limitation on the present application. In the drawings:

[0041] Figure 1 The module schematic diagram of the device for visualizing hidden dangers in power channels for this application;

[0042] Figure 2 Visualize the identification result diagram of the power channel hidden danger identification device for this application; DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0044] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms and the like is based on the orientation or positional relationship shown in the accompanying drawings, which is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device, element, module, system, platform or device referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. The following description of the present application is only understood as a description of individual embodiments of the technical solution of the present application. Other embodiments are not reflected in the following description, but it does not mean that the present application excludes these other embodiments, and the technical solution of the present application is not limited to the specific implementation methods described below, and the protection scope of the present application is not limited to only the specific implementation methods described below. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present application.

[0045] It should be noted that if the terms "first", "second", etc. appear in the specification and claims of the present application and the above-mentioned drawings, the description is only used to distinguish similar objects, and is not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0046] In some embodiments, Figure 1 As shown, a visual power channel hidden danger intelligent identification device includes:

[0047] A high-definition image acquisition module, used for acquiring images of the power channel;

[0048] An image intelligent recognition module, coupled to the high-definition image acquisition module, for storing the image and performing intelligent hidden danger recognition on the image to obtain hidden danger information;

[0049] A data transmission module, coupled to the image intelligent recognition module, for transmitting the image and the hidden danger information to a remote terminal server;

[0050] A power supply module, coupled to the high-definition image acquisition module, the image intelligent recognition module and the data transmission module respectively, for supplying power to the high-definition image acquisition module, the image intelligent recognition module and the data transmission module;

[0051] And a remote configuration and upgrade module, which is respectively coupled to the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module, and is used to provide remote parameter configuration, algorithm update and software upgrade for the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module.

[0052] Specifically, the storage of the image may be temporary storage; the high-definition image acquisition module, the image intelligent recognition module, the data transmission module, and the remote configuration and upgrade module are connected via a data bus to perform data transmission within the device.

[0053] The present application can be applied to the identification of wildfires or smoke in power channels. The device is installed on a power pole tower and includes a high-definition image acquisition module, an image intelligent recognition module, a data transmission module, a power supply module and a remote configuration and upgrade module; the modules are connected through a data bus; the data transmission module automatically selects a communication network to connect to a remote terminal server through a multi-path communication interface; the real-time status feedback unit monitors the working status of the device and the quality of the network connection in real time to ensure the stability of data transmission.

[0054] The operation and maintenance personnel view the recognition results, images, and collection time information through the monitoring interface on the remote terminal server. After confirming the hidden danger warning information, they notify the relevant maintenance personnel to handle it and record the time and place of the hidden danger. The operation and maintenance personnel remotely configure the parameters of the device through the remote configuration and upgrade module, adjust the shooting interval, hidden danger confidence and other parameters according to different seasons and climates, and remotely update the intelligent recognition algorithm and upgrade the software firmware.

[0055] In some embodiments, the image intelligent recognition module includes:

[0056] An image storage unit, used for storing the image;

[0057] Specifically, the image storage unit is a multi-level cache structure that supports image storage in multiple formats, including high-resolution visible light images and infrared thermal imaging images. Through an asynchronous data storage mechanism, the image stream is synchronously transmitted to the storage unit when the image is acquired, and the acquired visible light images and infrared images are stored in layers. When storing multimodal data, the system binds timestamps and location information to each frame of the image for subsequent image retrieval and multi-frame data processing. The storage unit has a built-in intelligent compression and encryption mechanism. The image will be adaptively compressed before storage to reduce storage space occupancy; at the same time, in order to prevent data tampering and loss, the storage unit performs AES encryption on the image data.

[0058] The hidden danger intelligent identification unit is used to detect the image according to the self-supervised pre-training model and the artificial intelligence recognition algorithm to obtain the hidden danger information, and display the hidden danger information on the image in the form of a marked box, such as Figure 2 shown.

[0059] The self-supervised pre-training model and artificial intelligence recognition algorithm specifically include a built-in self-supervised pre-training model, which includes contrastive learning, image restoration, and time series prediction tasks; a built-in customized convolutional neural network structure;

[0060] Built-in multi-modal data fusion strategy, integrating image data from visible light and infrared cameras. The fusion strategy uses a weighted fusion algorithm to dynamically adjust the weight of each modality data. In night conditions, the system will give priority to infrared images.

[0061] Built-in annotation box generation and display strategy. During the hidden danger identification process, the hidden danger is located through the multi-scale convolution kernel in the convolution kernel structure. When the flame or smoke feature is detected, the hidden danger annotation box is automatically generated and the hidden danger position is displayed in the image in a highlighted form. The size and position of the annotation box are dynamically adjusted according to the hidden danger area in the image, and the text information of the hidden danger is attached, such as flame intensity, smoke diffusion speed, etc.

[0062] In some embodiments, the data transmission module includes:

[0063] An image compression and encryption unit, used to compress the image marked with hidden danger information and encrypt the compressed image data;

[0064] Multi-path communication interface, used to integrate multiple communication methods and select the best path for data transmission;

[0065] And a real-time status feedback unit is used to monitor the working status and network connection quality information of the data transmission module in real time, and transmit the working status and network connection quality information to the remote terminal server.

[0066] In some embodiments, the power supply module includes:

[0067] A composite energy management unit, including a solar photovoltaic panel and a lithium-ion battery pack;

[0068] And a power monitoring unit is used to dynamically detect the energy consumption and remaining power of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module.

[0069] Specifically, the power supply module has a built-in solar photovoltaic panel to convert light energy into electrical energy to provide the device with the power required for normal operation; the excess electrical energy converted from solar energy is stored in a lithium-ion battery pack so that the lithium-ion battery can continue to supply power when the solar energy supply is insufficient at night or on cloudy days, ensuring a continuous energy source for the device; the power monitoring and scheduling unit detects the energy consumption and remaining power of each component of the device in real time, and sends an alarm message to the remote terminal server when it finds that the power is unhealthy.

[0070] Specifically, the power monitoring unit uses integrated current and voltage sensors to monitor the power consumption of the high-definition image acquisition module, image intelligent recognition module, data transmission module, and remote configuration and upgrade module in real time. The sensor dynamically captures the instantaneous power of each module, and determines the current remaining power by calculating the cumulative power consumption. It collects sensor data regularly and predicts the remaining power in combination with the discharge curve of the lithium-ion battery. The power monitoring unit also has an abnormal power consumption alarm function. When abnormal power consumption of a module is detected, an alarm is generated and uploaded to the remote server.

[0071] The high-definition image acquisition module automatically switches to use the visible light camera during the day and the infrared camera at night based on the image grayscale value. It automatically acquires power channel images every 5 minutes, and the acquisition interval can be adjusted through the remote configuration and upgrade module.

[0072] In some embodiments, the remote configuration and upgrade module includes:

[0073] A wireless upgrade unit, used to receive push software and algorithm models, and update and upgrade the high-definition image acquisition module, the image intelligent recognition module, the data transmission module, and the power supply module;

[0074] A configuration management unit, used to remotely set the working parameters of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module, view device status reports and logs, and roll back to a previous software version;

[0075] and a security authentication and data integrity check unit for identity authentication and encrypted communication protocols.

[0076] In some embodiments, the data transmission module is connected to the remote terminal server via an operator wireless network to perform remote data transmission.

[0077] In some embodiments, the hidden danger intelligent identification unit has a built-in deep learning algorithm and a graphics processor, and uses the power channel environment image for learning and training.

[0078] Specifically, self-supervised learning is performed through unlabeled data sets, using image restoration, contrastive learning, and time-series consistency prediction tasks to achieve automatic learning and extraction of hidden danger features (such as flames and smoke) in the power channel environment, and pre-training of image data collected at different time periods and under different lighting conditions to enhance recognition capabilities in complex environments;

[0079] The hidden danger intelligent identification unit receives data input from visible light cameras and infrared cameras, uses a multimodal fusion algorithm to combine information from different bands in the image, and adaptively adjusts the weight of feature extraction in different environments such as night or day. Infrared images are used to detect temperature changes and smoke diffusion patterns, and visible light images are used to identify the color and shape characteristics of flames;

[0080] Through the integration of graphics processing units (GPUs) for parallel computing, the deep learning model with optimized convolution kernel structure can efficiently process large amounts of image data and improve the real-time performance of hidden danger detection. Multi-scale convolution kernels are used to identify hidden dangers of different sizes and shapes (such as small-scale flames, widely diffused smoke, etc.). Convolution extracts the edge features of the hidden danger area to ensure recognition accuracy.

[0081] In some embodiments, the hidden danger intelligent identification unit analyzes the image through a self-supervised pre-training model and a deep learning algorithm, identifies the flame characteristics and smoke diffusion pattern, converts the identification result into warning information, and generates an image mark in the form of a hidden danger annotation box, such as Figure 2 shown.

[0082] Specifically, the self-supervised pre-training model includes contrastive learning, image restoration, and time series prediction tasks.

[0083] In the contrastive learning task, multiple frames of images are collected at the same point and at different times, and these images are embedded and mapped with features. By maximizing the similarity of images of the same type and minimizing the difference between images of different types, stable and distinguishing features are extracted, thereby learning the difference features in the image, including the dynamic jump of flames or the diffusion path of smoke. This task improves the model's recognition stability of hidden danger features under different lighting conditions;

[0084] In the image restoration task, we randomly block some areas on the unlabeled image, and use the multi-scale convolution kernel structure to infer the blocked area, generate the predicted feature map of the missing area, and align it with the unblocked feature map, and restore the blocked area through iterative learning. This enables the model to learn to understand the relationship between the local and global aspects of the image, such as the flames or smoke that may be hidden under the branches covering the wires, thereby improving its ability to identify hidden dangers in complex scenes;

[0085] In the time series prediction task, the model extracts dynamic features such as flame edges and smoke diffusion from multiple consecutive frames of images, captures changes in the image sequence through convolution operations, and then analyzes the changes in each frame through a time series convolutional network to capture long-term hidden danger feature trends, such as the continued expansion of flames or the spread speed of smoke. Finally, the long short-term memory network is used to process the information of past frames to predict the direction of flame spread or the range of smoke spread in the next few frames. The model will adjust its focus on short-term or long-term frame changes according to dynamic changes in the environment to ensure the accuracy of hidden danger identification in different environments.

[0086] The customized convolutional neural network structure specifically includes customization for hidden dangers such as flames and smoke. The convolution kernel is customized into a dual-path structure. One path is used to process visible light images to identify the color, shape, and edge features of flames; the other path processes infrared images to detect temperature anomalies and smoke diffusion patterns. This dual-path structure ensures the comprehensive extraction of hidden danger features.

[0087] In some embodiments, the high-definition image acquisition module includes a visible light camera and an infrared camera, and the visible light camera and the infrared camera can switch automatically.

[0088] In some embodiments, the lithium-ion battery pack is used to power the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module; the solar photovoltaic panel is used to charge the lithium-ion battery pack.

[0089] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0090] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A visual intelligent identification device for hidden dangers in power channels, characterized in that: include: A high-definition image acquisition module, used for acquiring images of the power channel; An image intelligent recognition module, coupled to the high-definition image acquisition module, for storing the image and performing intelligent hidden danger recognition on the image to obtain hidden danger information; A data transmission module, coupled to the image intelligent recognition module, for transmitting the image and the hidden danger information to a remote terminal server; A power supply module, coupled to the high-definition image acquisition module, the image intelligent recognition module and the data transmission module respectively, for supplying power to the high-definition image acquisition module, the image intelligent recognition module and the data transmission module; and a remote configuration and upgrade module, respectively coupled to the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module, for providing remote parameter configuration, algorithm update and software upgrade to the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module; The image intelligent recognition module includes: an image storage unit for storing the image; the image storage unit is a multi-level cache structure, and through an asynchronous data storage mechanism, when the high-definition image acquisition module acquires the image, the image stream is synchronously transmitted to the image storage unit and stored in layers; the image storage unit also includes a compression and encryption mechanism, and the image is adaptively compressed and AES encrypted before storage; and a hidden danger intelligent recognition unit for detecting the image according to a self-supervised pre-training model and an artificial intelligence recognition algorithm to obtain the hidden danger information, and displaying the hidden danger information on the image in the form of a marked box, the self-supervised pre-training model includes contrast learning, image restoration and time series prediction tasks; the artificial intelligence recognition algorithm includes a convolutional neural network; In the process of hidden danger identification, the convolutional neural network locates hidden dangers through the multi-scale convolution kernel in the convolution kernel structure. When the characteristics of the hidden danger are detected, a labeling box of the hidden danger is automatically generated, and the hidden danger position is displayed in the image in a highlighted form; the size and position of the labeling box are dynamically adjusted according to the hidden danger area in the image, and the text information of the hidden danger is attached; The hidden danger intelligent identification unit has a built-in deep learning algorithm and a graphics processor, and uses the power channel environment image for learning and training; the learning and training include: Self-supervised learning is performed through unlabeled data sets. Image restoration, contrastive learning, and time-series consistency prediction tasks are used to automatically learn and extract hidden danger features in the power channel environment, and pre-train image data collected at different time periods and under different lighting conditions. Combine the information of different bands in the image and adaptively adjust the weight of feature extraction in night or day environment; and deep learning models that optimize convolution kernel structures; The hidden danger intelligent identification unit analyzes the image through a self-supervised pre-training model and a deep learning algorithm, identifies the flame characteristics and the smoke diffusion pattern, converts the identification result into early warning information, and generates an image mark in the form of a hidden danger annotation box. The identification of the flame characteristics and the smoke diffusion pattern includes: Collect multiple frames of images at the same point and at different times, perform feature embedding mapping on the collected multiple frames of images, and extract stable and distinguishing features by maximizing the similarity of images of the same type and minimizing the differences of images of different types, so as to obtain the difference features existing in the image, including the dynamic jump of flames or the diffusion path of smoke; In the image restoration task, some areas are randomly occluded on the unlabeled image, and the occluded area is inferred through a multi-scale convolution kernel structure to generate a predicted feature map of the missing area, which is then aligned with the unoccluded feature map, and the occluded area is restored through iterative learning; The dynamic features of flame edges and smoke diffusion are extracted from continuous multi-frame images, and the changes in the image sequence are captured through convolution operations. The changes in each frame are analyzed through a temporal convolutional network to capture long-term hidden danger feature trends, including the continued expansion of flames or the spread speed of smoke. The information of past frames is processed using a long short-term memory network to predict the flame spread direction or smoke spread range in the next few frames.

2. The visual power channel hidden danger intelligent identification device according to claim 1 is characterized in that: The data transmission module comprises: An image compression and encryption unit, used to compress the image marked with hidden danger information and encrypt the compressed image data; Multi-path communication interface, used to integrate multiple communication methods and select the best path for data transmission; And a real-time status feedback unit is used to monitor the working status and network connection quality information of the data transmission module in real time, and transmit the working status and network connection quality information to the remote terminal server.

3. The visual power channel hidden danger intelligent identification device according to claim 2 is characterized in that: The power supply module comprises: A composite energy management unit, including a solar photovoltaic panel and a lithium-ion battery pack; And a power monitoring unit, which is used to dynamically detect the energy consumption and remaining power of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module; the power monitoring unit monitors the power consumption of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module in real time through an integrated sensor, and the sensor dynamically captures the instantaneous power of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module, and determines the current remaining power by calculating the cumulative power consumption, and predicts the remaining power in combination with the discharge curve of the lithium-ion battery; the power monitoring unit also includes an abnormal power consumption alarm module, which generates an alarm and uploads it to the remote terminal server when abnormal power consumption of one or more of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module is detected.

4. The visual power channel hidden danger intelligent identification device according to claim 3 is characterized in that: The remote configuration and upgrade module includes: A wireless upgrade unit, used to receive push software and algorithm models, and update and upgrade the high-definition image acquisition module, the image intelligent recognition module, the data transmission module, and the power supply module; A configuration management unit, used to remotely set the working parameters of the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the power supply module, view device status reports and logs, and roll back to a previous software version; and a security authentication and data integrity check unit for identity authentication and encrypted communication protocols.

5. The visual intelligent identification device for hidden dangers in power channels according to claim 4 is characterized in that: The data transmission module is connected to the remote terminal server via the operator's wireless network to perform remote data transmission.

6. The visual intelligent identification device for hidden dangers in power channels according to claim 5 is characterized by: The high-definition image acquisition module includes a visible light camera and an infrared camera, and the visible light camera and the infrared camera can be automatically switched.

7. The visual intelligent identification device for hidden dangers in power channels according to claim 6 is characterized by: The lithium-ion battery pack is used to supply power to the high-definition image acquisition module, the image intelligent recognition module, the data transmission module and the remote configuration and upgrade module; the solar photovoltaic panel is used to charge the lithium-ion battery pack.

Images