Distribution network unmanned aerial vehicle inspection video acquisition and identification system

Abstract

The invention relates to a distribution network unmanned aerial vehicle inspection video acquisition and identification system, which is characterized in that a control module and a video receiving module are matched, intelligent scheduling and monitoring, remote control and high-definition video real-time transmission in an unmanned aerial vehicle inspection process are completed through '5G + edge calculation ', and the time delay of a 1080p high-definition video returned by an unmanned aerial vehicle can be kept within 300 ms; the preprocessing module can be used for carrying out weighted averaging gray processing and filtering noise reduction processing on a video image file, and the screening module can be used for carrying out comparative analysis on a plurality of similar images of the same power supply equipment; an initial model can be constructed by utilizing the modeling component, and the initial model is trained, so that a hidden danger recognition model is generated, the hidden danger recognition model can be packaged and deployed into a recognition server according to an actual application scene by utilizing the deployment module, and storage backup of image data can be performed by utilizing the storage module; the method has the advantages of intelligent analysis and identification, accuracy and high efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of distribution network unmanned aerial vehicle inspection, and in particular relates to a distribution network unmanned aerial vehicle inspection video acquisition and recognition system. Background Art

[0002] Distribution network refers to the power grid that receives electric energy from the transmission network or regional power plants, distributes it locally or step by step according to the voltage to various users through distribution facilities. It is composed of overhead lines, cables, towers, distribution transformers, disconnectors, reactive power compensators and some ancillary facilities, and plays an important role in distributing electric energy in the power grid. Distribution network lines are generally exposed outdoors, and they will inevitably be affected by the natural and social environment during operation. In particular, extreme weather such as lightning and storms may destroy or damage distribution network equipment, or cause accidents such as electric shock. At the same time, problems such as theft, random wiring, and deliberate destruction may also affect the normal operation of distribution equipment and even lead to safety accidents. Therefore, it is necessary to regularize the distribution network lines. Time inspection, traditional power inspection methods usually rely on manual inspection and equipment detection. During the specific operation, the labor intensity of the staff is high, the inspection efficiency is low, and under the harsh external environment, the personal safety of the inspectors cannot be guaranteed, and there is a great risk. Therefore, the existing inspection work is mostly completed by autonomous inspection by drones. However, in the process of inspection by drones, it is impossible to realize real-time inspection video and picture collection, which leads to manual collection operation every time, which is labor-intensive and time-consuming, and the storage is chaotic, seriously affecting the inspection management effect; therefore, in order to solve the above problems, it is necessary to develop a distribution network drone inspection video collection and recognition system that can perform intelligent analysis and recognition and is accurate and efficient. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a distribution network drone inspection video acquisition and identification system that can perform intelligent analysis and identification and is accurate and efficient.

[0004] The object of the present invention is achieved as follows: a distribution network drone inspection video acquisition and identification system, including a control module, a video receiving module, a preprocessing module, a screening module, a modeling component, a deployment module and a storage module, wherein the control module is used to remotely control and adjust the flight direction, flight speed and flight angle of the drone, so that the drone can shoot videos and images according to the specified required trajectory, and the video receiving module is used to collect the video image files shot by the drone, and the control module and the video receiving module complete the intelligent scheduling and monitoring, remote control, and real-time transmission of high-definition video during the drone inspection process through "5G + edge computing", and realize that the delay of the 1080p high-definition video transmitted by the drone is kept within 300 ms, meeting the application technical requirements of real-time analysis and judgment of fault hazards; The pre-processing module can perform weighted average grayscale processing and filtering noise reduction processing on the video image file, thereby increasing the clarity of the image; The screening module can screen the pre-processed video and image files, so as to select representative images with sharp pixels and clear image environment from multiple similar images of the same power supply equipment: The modeling component can build an initial model based on EasyDL, then upload the original image data to the built initial model, train the initial model using different scenarios, and finally evaluate the trained initial model using the evaluation module. If the evaluation is qualified, a hidden danger identification model is generated based on this model. If the evaluation is unqualified, the model is deleted, and the initial model is rebuilt to perform data training again. The deployment module can package and deploy the hidden danger identification model built and trained by the modeling component to the identification server according to the actual application scenario, and perform equipment fault identification operations on the images screened by the screening module; The storage module can be used for unified storage backup of video image files and image file information with device failure.

[0005] Furthermore, a comparison and analysis unit is provided inside the screening module, and the comparison and analysis unit can compare and analyze multiple similar image files of the same power supply device.

[0006] Furthermore, a marking unit is provided inside the deployment module, and the marking unit can be used to mark the fault point during the process of identifying device faults in the image.

[0007] Furthermore, a statistical unit is provided inside the deployment module, and the statistical unit can be used to organize and count image files with equipment failures, and then perform real-time display operations through its external display.

[0008] Beneficial effects of the invention: The invention sets a control module and a video receiving module, utilizes the cooperation of the control module and the video receiving module, and uses "5G + edge computing" to complete intelligent scheduling and monitoring, remote control, and real-time transmission of high-definition video during the drone inspection process, so that the delay of the 1080p high-definition video transmitted by the drone is kept at 300 ms. Within, it meets the application technical requirements of real-time analysis of fault hazards; by setting a preprocessing module and a screening module, the preprocessing module can be used to perform weighted average grayscale processing and filtering noise reduction processing on the video image file, so as to increase the clarity of the image, and the screening module can be used in conjunction with the internal comparison and analysis unit to compare and analyze multiple similar images of the same power supply equipment, so as to screen out representative images; by setting a modeling component, it can be used to build an initial model, and the initial model can be trained to finally generate a hidden danger identification model, by setting a deployment module and a storage module, the deployment module can be used to package the hidden danger identification model according to the actual application scenario and deploy it to the identification server, and the storage module can be used to uniformly store and back up video image files and image file information with equipment failures, so as to facilitate subsequent review of image data; in general, the present invention has the advantages of intelligent analysis and identification and accurate and efficient. DETAILED DESCRIPTION

[0009] The present invention is further described below.

[0010] Embodiment: A distribution network drone inspection video acquisition and recognition system includes a control module, a video receiving module, a preprocessing module, a screening module, a modeling component, a deployment module and a storage module. The control module is used to remotely control and adjust the flight direction, flight speed and flight angle of the drone, so that the drone can shoot videos and images according to the specified required trajectory. The video receiving module is used to collect the video image files shot by the drone, and the control module and the video receiving module complete the intelligent scheduling and monitoring, remote control, and real-time transmission of high-definition video during the drone inspection process through "5G + edge computing", and realize that the delay of 1080p high-definition video transmitted by the drone is kept at 300 ms. Within, it meets the application technical requirements of real-time fault hidden danger analysis; the preprocessing module can perform weighted average grayscale processing and filtering noise reduction processing on the video image file, so as to increase the clarity of the image; the screening module can screen the video and image files after the preprocessing, so as to select representative images with clear and sharp pixels and clear image environment from multiple similar images of the same power supply equipment: the modeling component can build an initial model based on EasyDL, and then upload the original image data to the constructed initial model, train the initial model with different scenarios, and finally evaluate the trained initial model with the evaluation module. If the evaluation is qualified, a hidden danger identification model is generated based on this model. If the evaluation is unqualified, the model is deleted, and the initial model is rebuilt, and data training is performed again; the deployment module can package and deploy the hidden danger identification model built and trained by the modeling component to the recognition server according to the actual application scenario, and perform equipment fault identification operations on the images screened by the screening module; the storage module can be used for unified storage and backup of video image files and image file information with equipment faults.

[0011] The screening module is internally provided with a comparison and analysis unit, which can compare and analyze multiple similar image files of the same power supply equipment; the deployment module is internally provided with a marking unit, which can be used to mark the fault point during the process of identifying equipment faults in the image; the deployment module is internally provided with a statistical unit, which can be used to organize and count image files with equipment faults, and then perform real-time display operations through its external display.

[0012] When the present invention is used, first, through "5G + edge computing", intelligent scheduling and monitoring, remote control, and real-time transmission of high-definition video during the drone inspection process are completed, so that the delay of the 1080p high-definition video transmitted by the drone is kept within 300 ms, meeting the application technical requirements of real-time analysis of fault hazards. In this process, the control module can be used to remotely control and adjust the flight direction, flight speed, and flight angle of the drone, so that the drone can shoot videos and images according to the specified required trajectory. The video receiving module is used to collect the video image files shot by the drone, and then the pre-processing module is used to perform weighted average grayscale processing and filtering noise reduction processing on the video image files, so as to increase the clarity of the image. Then, the screening module is used in conjunction with the internal comparison and analysis unit to compare and analyze multiple similar images of the same power supply equipment, so as to screen out representative images with clear and sharp pixels and clear image environment; finally, the modeling component is used to construct the initial model, and the initial model is trained. After the evaluation is qualified, , a hidden danger identification model is generated based on the initial model. After completing the above operations, the deployment module is used to package the hidden danger identification model according to the actual application scenario and deploy it to the identification server, and the equipment fault identification operation is performed on the image screened by the screening module. In this process, the marking unit inside the deployment module can be used to complete the marking of the fault point in the process of equipment fault identification of the image; the statistical unit inside the deployment module can be used to organize and count the image files with equipment faults, and then perform real-time display operations through its external display; at the same time, the video image files and the image file information with equipment faults can also be uniformly stored and backed up through the storage module to facilitate the subsequent review of the image data; in general, the present invention has the advantages of intelligent analysis and identification, accuracy and efficiency.

[0013] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A distribution network drone inspection video acquisition and recognition system, including a control module, a video receiving module, a preprocessing module, a screening module, a modeling component, a deployment module and a storage module, characterized in that: The control module is used to remotely control and adjust the flight direction, flight speed and flight angle of the drone, so that the drone can shoot videos and images according to the specified required trajectory. The video receiving module is used to collect the video image files shot by the drone. The control module and the video receiving module complete the intelligent scheduling and monitoring, remote control, and real-time transmission of high-definition video during the drone inspection process through "5G + edge computing", and realize that the delay of the 1080p high-definition video sent back by the drone is kept within 300 ms, meeting the application technical requirements of real-time analysis of fault hazards; The pre-processing module can perform weighted average grayscale processing and filtering noise reduction processing on the video image file, thereby increasing the clarity of the image; The screening module can screen the pre-processed video and image files, so as to select representative images with sharp pixels and clear image environment from multiple similar images of the same power supply equipment: The modeling component can build an initial model based on EasyDL, then upload the original image data to the built initial model, train the initial model using different scenarios, and finally evaluate the trained initial model using the evaluation module. If the evaluation is qualified, a hidden danger identification model is generated based on this model. If the evaluation is unqualified, the model is deleted, and the initial model is rebuilt to perform data training again. The deployment module can package and deploy the hidden danger identification model built and trained by the modeling component to the identification server according to the actual application scenario, and perform equipment fault identification operations on the images screened by the screening module; The storage module can be used for unified storage backup of video image files and image file information with device failure.

2. A distribution network drone inspection video acquisition and recognition system as claimed in claim 1, characterized in that: A comparison and analysis unit is provided inside the screening module, and the comparison and analysis unit can compare and analyze a plurality of similar image files of the same power supply device.

3. A distribution network drone inspection video acquisition and recognition system as claimed in claim 1, characterized in that: The deployment module is internally provided with a marking unit, and the marking unit can be used to complete the marking of the fault point during the process of identifying the device fault in the image.

4. A distribution network drone inspection video acquisition and recognition system as claimed in claim 1, characterized in that: The deployment module is internally provided with a statistical unit, which can be used to sort and count the image files with equipment failures, and then perform real-time display operations through its external display.