Intelligent sensing system for power operation and maintenance

Abstract

The embodiment of the invention discloses an intelligent sensing system for power operation and maintenance. According to one specific embodiment of the intelligent sensing system for electric power operation and maintenance, an electric power data sensing terminal is in communication connection with an electric power data storage terminal and an electric power calculation terminal, and the electric power data sensing terminal comprises direct reading instrument equipment, electric power inspection unmanned aerial vehicle equipment, electric power inspection robot equipment and image acquisition equipment; the power data storage terminal is in communication connection with the power calculation terminal, and the power calculation terminal comprises an algorithm server and an edge calculation terminal; the algorithm server is in communication connection with the edge server, the algorithm server is used for identifying abnormal conditions in the electric power operation and maintenance image set, and the edge server is used for executing electric power operation and maintenance image preprocessing and electric power operation and maintenance image anomaly analysis of the electric power operation and maintenance image set. According to the embodiment, the automation level and accuracy of power equipment inspection can be improved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of electric power intelligent operation and maintenance, and specifically to an electric power operation and maintenance intelligent sensing system. Background Art

[0002] As the power grid area continues to expand, the requirements for operational safety, efficiency and reliability are constantly increasing. At present, when inspecting the power grid area, the usual method is to ensure the safe and stable operation of power equipment by combining traditional manual inspection with automated equipment monitoring.

[0003] However, in practice, it is found that when the above method is used to inspect the power grid area, the following technical problems often occur: Manual inspection requires a lot of manpower input, and the professional ability, experience and working status of the inspectors will directly affect the quality of the inspection, which may lead to missed faults or misjudgment.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0005] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0006] Some embodiments of the present disclosure propose an intelligent sensing system for power operation and maintenance to solve one or more of the technical problems mentioned in the above background technology section.

[0007] In the first aspect, some embodiments of the present disclosure provide an intelligent perception system for power operation and maintenance, which includes: the above-mentioned intelligent perception system for power operation and maintenance includes: a power data perception terminal, a power data storage terminal and a power computing terminal, characterized in that the power data perception terminal is communicatively connected with the power data storage terminal and the power computing terminal, wherein the power data perception terminal includes: a direct reading instrument device, a power inspection drone device, a power inspection robot device, and an image acquisition device; the power data storage terminal is communicatively connected with the power computing terminal, wherein the power computing terminal includes: an algorithm server and an edge computing terminal; the above-mentioned direct reading instrument device is used to collect a meter reading image set corresponding to each power equipment meter, wherein the above-mentioned direct reading instrument device includes: a direct reading instrument and a direct reading instrument gateway; the above-mentioned power inspection drone device is used to conduct aerial inspections of the target to be detected and collect a detection target image set of each target to be detected, wherein In the embodiment, the electric power inspection UAV equipment includes: UAVs and UAV parking stations; the electric power inspection robot equipment is used to inspect the inspection targets and collect the inspection target image sets of each ground inspection target, wherein the electric power inspection robot equipment includes: robots, base stations and robot workstations; the image acquisition equipment is used to collect the regional image sets of each area to be inspected, wherein the image acquisition equipment includes: fixed visible light gun camera, pan-tilt omnidirectional visible light ball camera, fixed infrared gun camera and pan-tilt omnidirectional infrared ball camera; the algorithm server is communicatively connected with the edge server, the algorithm server is used to identify abnormal conditions in the electric power operation and maintenance image set, and the edge server is used to perform electric power operation and maintenance image preprocessing and electric power operation and maintenance image abnormality analysis of the electric power operation and maintenance image set, wherein the electric power operation and maintenance image set includes: meter reading image set, detection target image set, inspection target image set and regional image set.

[0008] Optionally, the direct reading instrument is configured to: collect meter images corresponding to each power equipment meter to obtain a meter image set; perform grayscale processing on each meter image in the meter image set to generate a grayscale meter image to obtain a grayscale meter image set; perform image noise reduction processing on each grayscale meter image set in the grayscale meter image set to generate a noise-reduced meter image to obtain a noise-reduced meter image set; perform contrast enhancement processing on each noise-reduced meter image in the noise-reduced meter image set to generate an enhanced meter image. image, and obtain an enhanced meter image set; perform image binarization processing on each enhanced meter image in the enhanced meter image set to generate a binarized meter image, and obtain a binarized meter image set; perform edge detection processing on each binarized meter image in the binarized meter image set to generate an edge detection binarized meter image, and obtain an edge detection binarized meter image set; perform contour extraction processing on each edge detection binarized meter image in the edge detection binarized meter image set to generate a contour meter image, and obtain A contour meter image set, wherein an image boundary box exists in the contour meter image; the largest contour meter image in the contour meter image set is determined as a regional meter image; contour extraction processing is performed on the regional meter image to generate a contour regional meter image; in response to determining that the contour region area of ​​the contour regional meter image is less than a contour region area threshold of a preset contour regional meter image, a contour regional meter boundary box group is generated; based on the contour regional meter boundary box group, character sorting processing is performed on the regional meter image to generate a character sequence; each character in the character sequence is traversed to generate a traversed character set; character recognition processing is performed on each traversed character in the traversed character set to generate a recognized character and obtain a recognized character set; based on the recognized character set, a meter reading image set is output, and the meter reading image set is sent to a direct reading instrument gateway; and the direct reading instrument gateway is configured to: in response to receiving the meter reading image set, send the meter reading image set to a power data storage terminal and a power calculation terminal.

[0009] Optionally, the above-mentioned power inspection UAV equipment is configured to: in response to receiving an aerial patrol instruction, determine the current position of the patrol UAV and the position of each target to be detected, and plan a patrol path from the current position to the position of each target to be detected; control the UAV to move along the patrol path, and control the UAV to perform image acquisition and processing on each target to be detected to generate a detection target image, obtain a detection target image set, and send the acquired detection target image set to the power data storage terminal and the power computing terminal.

[0010] Optionally, the above-mentioned power inspection robot equipment is configured to: in response to receiving a ground inspection instruction, determine the current position of the inspection robot and the position of each target to be inspected, and plan an inspection path from the current position to the position of each target to be inspected; control the robot to move along the inspection path, and control the robot to perform image acquisition and processing on each target to be inspected to generate an inspection target image, obtain an inspection target image set, and upload the above-mentioned inspection target image set to the above-mentioned power data storage terminal and the power computing terminal through the above-mentioned base station.

[0011] Optionally, the fixed infrared gun camera is configured to: in response to receiving a face detection instruction, collect a first detection area image corresponding to each first area to be detected to obtain a first detection area image set; perform image preprocessing on each first detection area image in the first detection area image set to generate a preprocessed detection area image to obtain a preprocessed detection area image set; perform face positioning processing on each preprocessed detection area image in the preprocessed detection area image set to generate a face positioning image to obtain a face positioning image set; perform multi-scale detection processing on each face positioning image in the face positioning image set to generate a multi-scale detection face image to obtain a multi-scale detection face image set, wherein a face in the multi-scale detection face image has a bounding box; and perform multi-scale detection processing on each face positioning image in the face positioning image set to generate a multi-scale detection face image to obtain a multi-scale detection face image set. Each multi-scale detected face image in the detected face image set is subjected to face region cropping processing to generate a region cropped face image, and a region cropped face image set is obtained; each region cropped face image in the above region cropped face image set is subjected to face frame drawing processing to generate a frame drawn face image, and a frame drawn face image set is obtained; each frame drawn face image in the above frame drawn face image set is subjected to face image alignment processing to generate an aligned face image, and an aligned face image set is obtained; each aligned face image in the above aligned face image set is subjected to image denoising processing to generate a denoised face image, and a denoised face image set is obtained; each denoised face image in the above denoised face image set is subjected to image sharpening processing to generate a sharpened face image, and a sharpened face image set is obtained.

[0012] Optionally, the pan-tilt omnidirectional infrared ball camera is configured to, in response to receiving a target trajectory prediction instruction, collect a second detection area image corresponding to each second area to be detected to obtain a second detection area image set; perform target detection and target frame marking processing on each second detection area image in the second detection area image set to generate a target mark detection area image to obtain a target mark detection area image set, wherein the target mark detection area image has a bounding box of the marked target; perform target tracking processing on each target mark detection area image in the target mark detection area image set to generate a mark detection area image target position to obtain a mark detection area image target position set; perform trajectory prediction processing on the mark detection area image target position set to generate a mark detection area image target prediction trajectory; Each second detection area image in the domain image set is subjected to target point feature extraction processing to generate feature extraction target points and obtain a feature extraction target point set; each feature extraction target point in the above feature extraction target point set is subjected to target point feature matching processing to generate feature matching target points and obtain a feature matching target point set; based on the above feature matching target point set, each second detection area image in the above second detection area image set is subjected to target image registration processing to generate a registered target detection area image and obtain a registered target detection area image set; the registered target detection area image set is subjected to target image fusion processing to generate a fused target detection area image; based on the target prediction trajectory of the above marked detection area image, the above fused target detection area image is subjected to target locking processing to generate the future position of the target in the marked detection area image.

[0013] Optionally, the electric power data storage terminal is configured to: compress the electric power operation and maintenance image set to generate a compressed electric power operation and maintenance image set, and store the compressed electric power operation and maintenance image set in a hard disk recorder.

[0014] Optionally, the algorithm server is configured to: perform image recognition processing on each power operation and maintenance image in the power operation and maintenance image set to generate an identified power operation and maintenance image and obtain an identified power operation and maintenance image set; perform anomaly detection processing on the identified power operation and maintenance image set to generate an abnormality detection power operation and maintenance image result.

[0015] Optionally, the above-mentioned edge computing terminal is configured to: perform power operation and maintenance image preprocessing on the above-mentioned power operation and maintenance image set to generate a preprocessed power operation and maintenance image set; based on the above-mentioned abnormal detection power operation and maintenance image results, perform power operation and maintenance image abnormality analysis processing on the above-mentioned preprocessed power operation and maintenance image set to generate an abnormal analysis power operation and maintenance image result; in response to determining that the abnormal power operation and maintenance image result meets the preset power operation and maintenance abnormal alarm condition, control the associated alarm device to issue an abnormal alarm.

[0016] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the intelligent perception system of power operation and maintenance of some embodiments of the present disclosure, the automation level and accuracy of power equipment inspection can be improved. Specifically, the reason for missing faults or misjudgment is that manual inspection requires a lot of manual input, and the professional ability, experience and working status of the inspectors will directly affect the inspection quality. Based on this, the intelligent perception system of power operation and maintenance of some embodiments of the present disclosure, first, the power data perception terminal is connected with the power data storage terminal and the power computing terminal, wherein the power data perception terminal includes: direct reading instrument equipment, power inspection drone equipment, power inspection robot equipment, image acquisition equipment. Thus, real-time monitoring of the state of power equipment and collection of power operation and maintenance image sets can be realized, providing a basis for subsequent analysis and decision-making of power operation and maintenance image sets. Secondly, the power data storage terminal is connected with the power computing terminal, wherein the power computing terminal includes: algorithm server, edge computing terminal. Thus, efficient processing and analysis of the collected power operation and maintenance image sets can be realized, and the processing speed of the power operation and maintenance image sets can be improved. Secondly, the direct reading device is used to collect the meter reading accuracy image set corresponding to each power equipment meter, wherein the direct reading device includes: a direct reading device, a direct reading device gateway. Thus, the automation and accuracy of meter readings can be improved, and the errors of manual meter reading can be reduced. Then, the power inspection drone equipment is used to conduct aerial inspections of the targets to be detected, and collect the detection target image sets of each target to be detected, wherein the power inspection drone equipment includes: drones, drone parking stations. Thus, the inspection efficiency and coverage of remote or difficult-to-reach areas can be improved. Then, the power inspection robot equipment is used to inspect the targets to be inspected, and collect the inspection target image sets of each target to be inspected on the ground, wherein the power inspection robot equipment includes: a robot, a base station, and a robot workstation. Thus, the inspection efficiency and accuracy of ground facilities can be improved, especially in complex terrain or harsh environments. Then, the image acquisition device is used to collect regional image sets of each area to be detected, wherein the image acquisition device includes: a fixed visible light gun camera, a pan-tilt omnidirectional visible light ball camera, a fixed infrared gun camera, and a pan-tilt omnidirectional infrared ball camera. Thus, a more comprehensive monitoring perspective and data can be provided, and the monitoring capability of the equipment status can be enhanced. Finally, the algorithm server is connected to the edge server in communication, and the algorithm server is used to identify abnormal conditions in the power operation and maintenance image set, and the edge server is used to perform power operation and maintenance image preprocessing and power operation and maintenance image abnormality analysis of the power operation and maintenance image set, wherein the power operation and maintenance image set includes: meter reading image set, detection target image set, inspection target image set and regional image set. Thus, the intelligent monitoring and fault prediction of power equipment can be realized by using advanced computer vision technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0018] Figure 1 is a schematic system architecture diagram of a power operation and maintenance intelligent perception system suitable for implementing some embodiments of the present disclosure; Figure 2 It is a schematic system architecture diagram of a power data sensing terminal suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0019] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0020] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0021] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0022] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0023] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0024] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0025] Figure 11 is a schematic system architecture diagram of the power operation and maintenance intelligent perception system according to the present disclosure. The system architecture of the power operation and maintenance intelligent perception system 100 according to the present disclosure is shown. The power operation and maintenance intelligent perception system 100 includes: a power data perception terminal 101, a power data storage terminal 102 and a power calculation terminal 103.

[0026] In some embodiments, the power data sensing terminal is in communication with the power data storage terminal and the power computing terminal. The power data sensing terminal includes: a direct reading device, a power inspection drone device, a power inspection robot device, and an image acquisition device.

[0027] Here, the power data sensing terminal can refer to Figure 2 is a schematic diagram of a power data sensing terminal structure suitable for implementing some embodiments of the present application, such as Figure 2 As shown, the above-mentioned power data sensing terminal includes: a direct reading device 1, a power inspection drone device 2, a power inspection robot device 3, and an image acquisition device 4. The above-mentioned direct reading device 1 includes: a direct reading device 4 and a direct reading device gateway 5. The above-mentioned power inspection drone device 2 includes: a drone 6 and a drone parking station 7. The above-mentioned power inspection robot device 3 includes: a robot 8, a base station 9, and a robot workstation 10. The above-mentioned image acquisition device 4 includes: a fixed visible light gun camera 11, a pan-tilt omnidirectional visible light ball camera 12, a fixed infrared gun camera 13, and a pan-tilt omnidirectional infrared ball camera 14. The direct reader 4 is connected to the direct reader gateway 5; the drone 6 is connected to the drone parking station 7; the robot 8 is connected to the base station 9 and the robot workstation 10 respectively; the base station 9 is connected to the robot workstation 10; the fixed visible light gun camera 11 is connected to the pan-tilt omnidirectional visible light ball camera 12, the fixed infrared gun camera 13, and the pan-tilt omnidirectional infrared ball camera 14 respectively; the pan-tilt omnidirectional visible light ball camera 12 is connected to the fixed infrared gun camera 13 and the pan-tilt omnidirectional infrared ball camera 14 respectively; the fixed infrared gun camera 13 is connected to the pan-tilt omnidirectional infrared ball camera 14.

[0028] The direct reading instrument 4 adopts a pixel image sensor, has an automatic / manual shutter function, and supports automatic fill light; the direct reading instrument 4 supports a shooting distance within a certain range, has a specific frame rate and field of view; the direct reading instrument 4 uses a wireless transmission protocol, supports real-time wake-up and timed image transmission, is powered by a lithium battery, has low power consumption and a high number of image acquisitions. The direct reading instrument gateway 5 is used to wake up the direct reading instrument, set the image acquisition frequency of the direct reading instrument 4, receive the image data collected by the direct reading instrument 4, and perform image reading processing.

[0029] The drone 6 has a long flight time and a suitable working environment temperature; the drone 6 has a specific working frequency and transmission power; the drone 6 supports fast charging and has a large charging time and charging power; the gimbal camera carried by the drone 6 adopts a high-pixel image sensor, has a variety of shooting parameters, and can capture high-definition images. The drone parking station 7 realizes automatic power on and off, automatic inspection control, image reception and storage, and precise landing; the drone parking station 7 has multiple protection measures to ensure the safety and stable charging of the drone 6; the drone parking station 7 realizes the automatic power on and off and automatic charging functions of the drone 6 through the automatic centering mechanism.

[0030] The robot 8 realizes autonomous navigation through a simultaneous positioning and mapping navigation system, which performs positioning and mapping by using a lidar and an inertial measurement unit; the robot 8 collects images and videos, and transmits them to the robot workstation 10 in real time through the base station 9; the robot 8 generates alarm information when an abnormality of the power equipment is detected, and records and stores the inspection information.

[0031] The above-mentioned fixed visible light gun machine 11 adopts a high-performance image sensor with high pixel resolution, can work in low illumination environment, and output high-definition images; the above-mentioned fixed visible light gun machine 11 supports high frame rate video output to ensure smooth picture, supports encoding technology, and realizes high compression ratio and low code stream transmission; the above-mentioned fixed visible light gun machine supports multiple monitoring modes, and the above-mentioned fixed visible light gun machine 11 supports encoding technology to adapt to different bandwidths and storage environments; the above-mentioned fixed visible light gun machine 11 supports operation without a secure digital card and has alarm input and output functions; the above-mentioned fixed visible light gun machine 11 supports large-capacity storage without a secure digital card; the above-mentioned fixed visible light gun machine 11 supports multiple power supply methods; the above-mentioned fixed visible light gun machine 11 supports twelve volt power return and has maximum current requirements; the above-mentioned fixed visible light gun machine 11 supports voltage abnormality alarm function to facilitate engineering installation.

[0032] The above-mentioned pan-tilt omnidirectional visible light ball camera 12 supports multiple intelligent monitoring functions; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 adopts a high-performance image sensor with high pixel resolution, can work in a low-light environment, and output high-definition images; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 supports high frame rate video output to ensure smooth pictures; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 supports encoding technology to achieve high compression ratio and low bit rate transmission; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 has a built-in infrared fill light and can monitor in an environment without visible light; the above-mentioned pan-tilt omnidirectional visible light ball camera supports multiple monitoring modes to adapt to different monitoring environments; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 supports encoding technology to adapt to different bandwidths and storage environments; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 supports large-capacity storage cards to meet long-term video storage needs; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 supports multiple power supply methods to adapt to different power supply environments; the above-mentioned pan-tilt omnidirectional visible light ball camera 12 has a wide voltage design and can work stably under different voltages.

[0033] The above-mentioned fixed infrared gun camera 13 has a built-in graphics processor chip, supports deep learning algorithms, and improves detection accuracy; the above-mentioned fixed infrared gun camera 13 supports intelligent resource switching, and is a general behavior analysis, face detection, and crowd counting; the above-mentioned fixed infrared gun camera 13 supports face detection function; the above-mentioned fixed infrared gun camera 13 supports multiple face attributes; the above-mentioned fixed infrared gun camera 13 supports crowd counting; the above-mentioned fixed infrared gun camera 13 supports multiple behavior detection; the above-mentioned fixed infrared gun camera 13 supports multi-stream function to meet multi-channel high-definition video display; the above-mentioned fixed infrared gun camera 13 adopts high-performance image sensors to output high-definition images; the above-mentioned fixed infrared gun camera 13 supports encoding technology to achieve high compression ratio and low stream transmission; the above-mentioned fixed infrared gun camera 13 supports multiple monitoring modes; the above-mentioned fixed infrared gun camera 13 supports encoding to adapt to different bandwidths and storage environments; the above-mentioned fixed infrared gun camera 13 supports multiple alarm input and output and audio input and output; the above-mentioned fixed infrared gun camera supports 13 large-capacity non-secure digital card storage; the above-mentioned fixed infrared gun camera 13 supports multiple power supply methods.

[0034] The above-mentioned pan-tilt all-round infrared ball camera 14 supports vehicle feature recognition and can upload information to the power data storage terminal and the power calculation terminal; the above-mentioned pan-tilt all-round infrared ball camera 14 can cover multiple lanes and can perform video analysis on the near-end lanes; the above-mentioned pan-tilt all-round infrared ball camera 14 supports face detection and crowd counting; the above-mentioned pan-tilt all-round infrared ball camera 14 supports heat maps and has high-power optical zoom and digital zoom capabilities; the above-mentioned pan-tilt all-round infrared ball camera 14 supports super-starlight-level ultra-low illumination and can work in extremely dark environments; the above-mentioned pan-tilt all-round infrared ball camera 14 has a high signal-to-noise ratio and supports privacy shielding; the above-mentioned pan-tilt all-round infrared ball camera 14 can process multiple areas at the same time and can display multiple areas on the same screen; the above-mentioned pan-tilt all-round infrared ball camera 14 supports wide dynamic effects and image noise reduction functions, and can display images during the day and at night; the above-mentioned pan-tilt all-round infrared ball camera 14 supports wiper function and built-in infrared light fill light; the above-mentioned pan-tilt all-round infrared ball camera 14 supports wide voltage input, reaches a high protection level, and has lightning protection and surge protection and anti-surge protection; the above-mentioned pan-tilt all-round infrared ball camera 14 supports open application programming interface for application integration; the above-mentioned pan-tilt all-round infrared ball camera 14 supports three-stream technology, and can perform single-scene tracking, multi-scene tracking, and panoramic tracking; the above-mentioned pan-tilt all-round infrared ball camera 14 supports two tracking methods: manual tracking and alarm tracking; the above-mentioned pan-tilt all-round infrared ball camera 14 supports multiple behavior detections; the above-mentioned pan-tilt all-round infrared ball camera 14 supports target filtering, tracking algorithm, automatic target locking, and automatic adjustment of the panoramic pan-tilt focal length to obtain the best monitoring image; the above-mentioned pan-tilt all-round infrared ball camera 14 has a panoramic pan-tilt and panoramic algorithm, and can seamlessly stitch panoramic images; the above-mentioned pan-tilt all-round infrared ball camera 14 can rotate continuously in the horizontal direction, and continuously monitor after automatically flipping in the vertical direction, without blind spots; the above-mentioned pan-tilt all-round infrared ball camera 14 can adjust the pan-tilt positioning accurately with horizontal keying speed and vertical keying speed; the above-mentioned pan-tilt all-round infrared ball camera 014 supports multiple preset positions, can complete multiple cruise paths, and can set multiple patrol paths.

[0035] For example, the pixel image sensor may be a complementary metal oxide semiconductor (CMOS) sensor. The automatic / manual shutter function may be an exposure control function. The shooting distance may be 500 meters. The specific frame rate and field of view may be 30 frames per second (fps) and 60 degrees. The wireless transmission protocol may be Wi-Fi, Bluetooth. Real-time wake-up and scheduled image transmission may be timer trigger and active wake-up mode. The lithium battery may be a 3.6V lithium battery. Low power consumption and high image acquisition times may be 6000 times. The flight time may be 21 minutes. The working environment temperature may be -10°C~40°C. The specific working frequency may be 2.400~2,4835GHz, 725-5.850GHz. The transmission power may be 2.400-2.4835GHz, FCC≤26dBm, CE≤20dBm, SRRC≤20dBm, MIC≤20dBm / 5.725-5.850GHz. The charging time may be: 95 minutes. The charging power may be: 80W. The pixel image sensor may be: 1 / 2.3 inch CMOS. The various shooting parameters may be: light time, white balance. The multiple protection measures may be: overvoltage protection, overtemperature protection. The automatic centering mechanism may be: a motor drive mechanical structure. The laser radar may be: a Hersel radar. The high-performance image sensor may be: a complementary metal oxide semiconductor (CMOS) sensor. The encoding technology may be: H.264, H.265. The non-secure digital card may be: an SD card. The infrared fill light may be: a Vstarcam infrared fill light. The large-capacity storage card may be: a 128G Micro SD card. The various power supply modes may be: a DC12V mode, a POE mode. The graphics processor chip may be: a GPU chip. The deep learning algorithm may be: a convolutional neural network (CNN) algorithm. The intelligent resources may be: general behavior analysis, face detection, and people counting. The general behavior analysis may be: health behavior analysis. The face detection function may be: tracking, optimization, capture, face enhancement, face exposure, and face attribute extraction. Multiple behavior detections can be: tripwire intrusion, area intrusion, fence crossing, wandering detection, item left behind, item removal, and fast movement. Multi-stream functions can be: main stream, sub-stream. Multiple monitoring modes can be: corridor mode, wide dynamic, strong light suppression, backlight compensation, and digital watermark. Multiple lanes can be: 2-3 lanes. Multiple behavior detections can be: tripwire intrusion, area intrusion, fence crossing, wandering detection, item left behind, item removal, and fast movement. Face detection functions can be: tracking, optimization, snapshot, face enhancement, face exposure, and face attribute extraction. Panoramic PTZ can be: omnidirectional rotating PTZ. Panoramic algorithm can be: panoramic image stitching algorithm. Multiple preset positions can be: 300. Multiple patrol paths can be: 5. Multiple intelligent monitoring functions can be: tripwire intrusion, area intrusion, fast movement, item left behind, item removal, and face detection.Crowd counting can be: queue management and area crowd counting. Vehicle feature recognition can be: license plate recognition, body color recognition, vehicle model recognition, and vehicle logo recognition.

[0036] In some embodiments, the power data storage terminal is connected to the power computing terminal in communication, and the power computing terminal includes: an algorithm server and an edge computing terminal. For example, the algorithm server may be: a graphics processing server. The edge computing terminal may be: an Internet of Things edge computing terminal.

[0037] In some embodiments, the direct reading device is used to collect a set of meter reading images corresponding to each power equipment meter. The direct reading device includes: a direct reading instrument and a direct reading instrument gateway. The direct reading instrument is used to directly read the meter readings of the power equipment meter and convert the meter readings into digital signal entities. The direct reading instrument gateway is used to receive the digital signal transmitted by the direct reading instrument and forward the signal to the power data storage terminal and the power calculation terminal. For example, the direct reading instrument can be: a direct reading spectrometer. The direct reading instrument gateway can be: a spectrum data gateway. The power equipment meter can be: a meter.

[0038] In some embodiments, the above-mentioned power inspection drone equipment is used to conduct aerial inspections of the inspection targets and collect detection target image sets of each target to be inspected. Among them, the above-mentioned power inspection drone equipment includes: drones and drone parking stations. The drone is used to conduct aerial inspections of the inspection targets and the detection target sets of each target to be inspected. The drone parking station is used to store and charge the drone to ensure that the drone can automatically return after performing a mission and prepare for the next mission. For example, the drone can be: a multi-rotor drone. The drone parking station can be: an intelligent charging parking station.

[0039] In some embodiments, the above-mentioned power inspection robot device is used to inspect the inspection targets and collect inspection target image sets of each ground target to be inspected. Among them, the above-mentioned power inspection robot device includes: a robot, a base station, and a robot workstation. The robot is used to inspect the inspection targets and collect inspection target image sets of each ground target to be inspected. The base station is used to communicate with the robot, send control instructions and receive the inspection target image sets collected by the robot. The robot workstation is used to monitor the operating status of the robot and analyze the inspection target image set. For example, the robot can be: an industrial collaborative robot. The base station can be: a wireless communication base station. The robot workstation can be: an automatic maintenance workstation.

[0040] In some embodiments, the image acquisition device is used to acquire a set of regional images of each area to be detected. Among them, the image acquisition device includes: a fixed visible light gun camera, a pan-tilt omnidirectional visible light ball camera, a fixed infrared gun camera, and a pan-tilt omnidirectional infrared ball camera. The fixed visible light gun camera is used to monitor the power equipment area and provide high-definition visible light images. The pan-tilt omnidirectional visible light ball camera is used to monitor the power equipment area in all directions, and can automatically track moving targets and provide continuous image data. The fixed infrared gun camera is used to monitor the power equipment area in bad weather or night conditions, and provide infrared images to detect thermal anomalies. The fixed visible light gun camera is used to monitor the power equipment area in all directions, provide infrared images to detect thermal anomalies, and can automatically track heat sources. For example, the fixed visible light gun camera can be: a high-definition surveillance camera. The pan-tilt omnidirectional visible light ball camera can be: an intelligent tracking ball camera. The fixed infrared gun camera can be: a thermal imaging surveillance camera. The pan-tilt omnidirectional infrared ball camera can be: a panoramic thermal imaging surveillance ball camera. The power equipment can include: a transformer, a circuit breaker, and a meter.

[0041] In some embodiments, the algorithm server is in communication with the edge server, the algorithm server is used to identify abnormal conditions in the power operation and maintenance image set, and the edge server is used to perform power operation and maintenance image preprocessing and power operation and maintenance image abnormality analysis of the power operation and maintenance image set. The power operation and maintenance image set includes: meter reading image set, detection target image set, inspection target image set and area image set.

[0042] Optionally, the direct reader is configured as: The first step is to collect meter images corresponding to each power equipment meter to obtain a meter image set. For example, the direct reading instrument can use its built-in imaging device to collect meter images corresponding to each power equipment meter to obtain a meter image set. The built-in imaging device can be: a camera.

[0043] In the second step, each meter image in the meter image set is grayed to generate a grayed meter image and obtain a grayed meter image set. For example, the direct reading instrument can convert each meter image in the meter image set into a grayed meter image by a preset graying algorithm to obtain a grayed meter image set. The graying algorithm can be: an average value algorithm (taking the average value of the three RGB components as the gray value).

[0044] The third step is to perform image denoising on each grayscale meter image set in the grayscale meter image set to generate a denoised meter image and obtain a denoised meter image set. For example, the direct reading instrument can use a filter to perform image denoising on each grayscale meter image set in the grayscale meter image set to generate a denoised meter image and obtain a denoised meter image set. The filter can be a Gaussian filter.

[0045] In the fourth step, each denoised meter image in the denoised meter image set is subjected to contrast enhancement processing to generate an enhanced meter image, thereby obtaining an enhanced meter image set. For example, the direct reading instrument may perform contrast enhancement processing on each denoised meter image in the denoised meter image set by using a preset histogram equalization algorithm to generate an enhanced meter image, thereby obtaining an enhanced meter image set. The preset histogram equalization algorithm may be a contrast-limited adaptive histogram equalization algorithm.

[0046] The fifth step is to perform image binarization processing on each enhanced meter image in the enhanced meter image set to generate a binary meter image and obtain a binary meter image set. For example, first, the direct reading instrument can set a preset pixel threshold, and in response to determining that the pixel value in the enhanced meter image is greater than or equal to the preset pixel threshold, the enhanced meter image is determined as a white area meter image. Secondly, in response to determining that the pixel value in the enhanced meter image is less than the preset pixel threshold, the enhanced meter image is determined as a black area meter image. Finally, using a preset binarization algorithm, the white area meter image and the black area meter image are subjected to pixel value classification processing to generate a binary meter image set. The preset binarization algorithm can be: Otsu algorithm. The preset pixel threshold can be: 0-255.

[0047] In the sixth step, edge detection processing is performed on each binary meter image in the above-mentioned binary meter image set to generate an edge-detected binary meter image and obtain an edge-detected binary meter image set. For example, the above-mentioned direct reading instrument can perform meter image boundary extraction processing on each binary meter image in the above-mentioned binary meter image set by using a preset edge detection algorithm to generate an edge-detected binary meter image and obtain an edge-detected binary meter image set. The preset edge detection algorithm can be: Canny edge detection algorithm.

[0048] In the seventh step, contour extraction processing is performed on each edge detection binary meter image in the edge detection binary meter image set to generate a contour meter image and obtain a contour meter image set. There is an image boundary box in the contour meter image. For example, the direct reading instrument can perform contour extraction processing on each edge detection binary meter image in the edge detection binary meter image set through a preset function to generate a contour meter image and obtain a contour meter image set. The preset function can be: find contour (cv2.findContours function).

[0049] In the eighth step, the largest contour meter image in the above contour meter image set is determined as the regional meter image.

[0050] In the ninth step, the contour extraction process is performed on the regional meter image to generate a contour regional meter image. For example, the direct reading instrument can perform contour extraction process on the regional meter image through a preset function to generate a contour regional meter image.

[0051] In the tenth step, in response to determining that the contour area of ​​the contour area meter image is less than the contour area area threshold of the preset contour area meter image, a contour area meter boundary box group is generated. For example, the direct reading instrument can perform contour filtering processing on the contour area meter image to generate a filtered contour area. Secondly, the filtered contour area is screened using the preset function to generate a screened filtered contour area. Finally, a contour area meter boundary box group is extracted from the screened filtered contour area. The contour area area threshold of the contour area meter image can be: 100.

[0052] In the eleventh step, based on the above-mentioned contour area meter boundary box group, the above-mentioned area meter image is subjected to character sorting processing to generate a character sequence. For example, first, the above-mentioned direct reading instrument can perform horizontal sorting processing on the above-mentioned contour area meter boundary box group to generate a horizontal sorting meter boundary box group. According to the above-mentioned horizontal sorting meter boundary box group, the above-mentioned area meter image is cropped to generate a character sequence.

[0053] In the twelfth step, each character in the character sequence is traversed to generate a traversed character set. For example, the direct reader can perform feature extraction on each character in the character sequence to generate feature-extracted characters and obtain a feature-extracted character sequence as the traversed character set.

[0054] In step 13, character recognition processing is performed on each traversal character in the traversal character set to generate a recognition character and obtain a recognition character set. For example, the direct reader can perform character recognition processing on each traversal character in the traversal character set using a preset character recognition algorithm to generate a recognition character and obtain a recognition character set. The preset character recognition algorithm can be: an optical character recognition algorithm (OCR).

[0055] Step 14: Based on the recognition character set, output a meter reading image set, and send the meter reading image set to the direct reading instrument gateway. For example, the direct reading instrument can perform image format conversion processing on each recognition character in the recognition character set to generate a meter reading image set, and send the meter reading image set to the direct reading instrument gateway.

[0056] Optionally, the direct reader gateway is configured as follows: In some embodiments, in response to receiving the meter reading image set, the meter reading image set is sent to a power data storage terminal and a power calculation terminal.

[0057] Optionally, the above-mentioned power inspection drone equipment is configured as: The first step is to determine the current position of the patrol drone and the position of each target to be detected in response to receiving an aerial patrol instruction, and plan a patrol path from the current position to the position of each target to be detected. For example, first, the above-mentioned power inspection drone device can respond to determining that the battery power and signal are normal, unlock the motor and control the above-mentioned drone to start taking off to a preset height. Then, in response to an error encountered during takeoff, the above-mentioned drone is controlled to perform an emergency landing procedure. Then, the current position of the drone is obtained, and a patrol path to the next inspection point is planned. Then, the drone is controlled to move along the planned patrol path to monitor obstacles in real time. Finally, in response to detecting an obstacle, the patrol path is replanned and the flight route of the above-mentioned drone is adjusted. The aerial patrol instruction can be: check the operating status of the targets to be detected along the patrol path. The targets to be detected can be: transmission lines, wind turbines, and solar panels.

[0058] The second step is to control the drone to move along the patrol path, and control the drone to perform image acquisition and processing on each target to be detected to generate a detection target image, obtain a detection target image set, and send the collected detection target image set to the power data storage terminal and the power calculation terminal. For example, first, the above-mentioned power inspection drone device can control the above-mentioned drone to move along the patrol path until it reaches the position of the target to be detected. Secondly, according to the position of the above-mentioned target to be detected, the drone is controlled to fly to each target position to be detected in turn. Thirdly, at each target position to be detected, the onboard camera is used to perform image acquisition to obtain a detection target image set. Then, the above-mentioned detection target image set is saved and sent to the power data storage terminal and the power calculation terminal. Then, the above-mentioned detection target image set is subjected to abnormal detection processing to generate an abnormal detection detection target recognition result. Finally, in response to receiving the detection target image set of the target to be detected, the above-mentioned drone is controlled to return to the above-mentioned drone parking station. The camera can be: an infrared camera. The abnormal detection detection target recognition result can be: the running state of the target to be detected is abnormal, and the running state of the target to be detected is normal.

[0059] Optionally, the above-mentioned power inspection robot device is configured as: The first step is to determine the current position of the inspection robot and the positions of each target to be inspected in response to receiving a ground inspection instruction, and plan an inspection path from the current position to the positions of each target to be inspected. For example, the ground inspection instruction may include: setting the inspection path, target equipment, and detection parameters. For example, the inspection path from the current position to the positions of each target to be inspected may be planned by a preset path planning algorithm. The preset path planning algorithm may be: Dijkstra algorithm, RRT (Rapidly-exploring Random Trees) algorithm, MILP (Mixed Integer Linear Programming) algorithm.

[0060] The second step is to control the robot to move along the inspection path, and control the robot to perform image acquisition and processing on each target to be inspected to generate an inspection target image, obtain an inspection target image set, and upload the inspection target image set to the power data storage terminal and the power calculation terminal through the base station. For example, first, the power inspection robot device can control the robot to move autonomously along the inspection path according to the planned inspection path using the front wheel drive and differential control system, while relying on the laser radar to avoid obstacles. Secondly, control the robot to perform image acquisition and processing on each target to be inspected to generate an inspection target image and obtain an inspection target image set. Thirdly, the collected inspection target image set is transmitted to the robot workstation in real time through the base station. Then, in response to determining that the robot workstation receives the inspection target image set, the inspection target image set is subjected to abnormal detection processing to generate an abnormal detection inspection target recognition result. Then, in response to determining that the abnormal detection inspection target recognition result meets the preset abnormal alarm condition, the associated alarm device is controlled to perform an abnormal alarm. Finally, in response to determining that the battery power is low after the inspection command is completed, the automatic docking system is used to control the robot to autonomously navigate back to the charging pile for charging. The inspection targets to be inspected may be: power distribution equipment, substation equipment. The inspection target recognition result of abnormal detection may be: the operating status of the inspection target to be inspected is abnormal, and the operating status of the inspection target to be inspected is normal.

[0061] Optionally, the above-mentioned fixed infrared gun camera is configured as: In the first step, in response to receiving a face detection instruction, a first detection area image corresponding to each first to-be-detected area is collected to obtain a first detection area image set. For example, the face detection instruction may be a voice instruction or a programming instruction.

[0062] In the second step, each first detection area image in the above-mentioned first detection area image set is subjected to image preprocessing to generate a preprocessed detection area image and obtain a preprocessed detection area image set. For example, first, the above-mentioned fixed infrared gun machine can perform image cleaning processing on each first detection area image in the above-mentioned first detection area image set to generate a cleaned detection area image and obtain a cleaned detection area image set. Secondly, each cleaned detection area image in the above-mentioned cleaned detection area image set is subjected to image feature selection processing to generate a feature-selected detection area image and obtain a feature-selected detection area image set. Then, each feature-selected detection area image in the above-mentioned feature-selected detection area image set is subjected to image standardization processing to generate a preprocessed detection area image and obtain a preprocessed detection area image set.

[0063] The third step is to perform face positioning processing on each preprocessed detection area image in the preprocessed detection area image set to generate a face positioning image and obtain a face positioning image set. For example, the fixed infrared gun can perform face positioning processing on each preprocessed detection area image in the preprocessed detection area image set using a deep learning model and a classifier to generate a face positioning image and obtain a face positioning image set. The deep learning model can be: a residual network model (ResNet), a multi-task cascade convolutional neural network (MTCNN). The classifier can be: a feature cascade classifier (Haar).

[0064] The fourth step is to perform multi-scale detection processing on each face positioning image in the above face positioning image set to generate a multi-scale detection face image and obtain a multi-scale detection face image set. Among them, the face in the multi-scale detection face image has a bounding box. For example, the above-mentioned fixed infrared gun camera can use multiple detection scales to perform multi-scale detection processing on each face positioning image in the above face positioning image set to generate a multi-scale detection face image and obtain a multi-scale detection face image set. The multiple detection scales can be: 1:1, 1:1.25, 1:1.5, 1:1.75, 1:2.

[0065] The fifth step is to perform face region cropping processing on each multi-scale detected face image in the above multi-scale detected face image set to generate a regional cropped face image and obtain a regional cropped face image set. For example, first, the above fixed infrared gun can obtain the bounding box coordinates of each face from the above multi-scale detected face image set. Secondly, based on the bounding box coordinates of each face, the face region corresponding to the multi-scale detected face image is cropped from the above multi-scale detected face image set to obtain a regional cropped face image set. The bounding box coordinates can be: x, y, z.

[0066] In the sixth step, a face frame drawing process is performed on each of the regional cropped face images in the regional cropped face image set to generate a frame-drawn face image and obtain a frame-drawn face image set. For example, the fixed infrared gun camera can perform face frame drawing process on each of the regional cropped face images in the regional cropped face image set by using a preset face frame drawing function to generate a frame-drawn face image and obtain a frame-drawn face image set. The preset face frame drawing function can be: detect faces (detect_faces).

[0067] In the seventh step, each frame-drawn face image in the frame-drawn face image set is subjected to face image alignment processing to generate an aligned face image and obtain an aligned face image set. For example, the fixed infrared gun camera can perform face image alignment processing on each frame-drawn face image in the frame-drawn face image set by using a preset alignment algorithm to generate an aligned face image and obtain an aligned face image set. The preset alignment algorithm can be: a key point detection algorithm, a geometric transformation algorithm.

[0068] In the eighth step, each aligned face image in the aligned face image set is subjected to image denoising to generate a denoised face image and obtain a denoised face image set. For example, the fixed infrared gun camera can perform image denoising on each aligned face image in the aligned face image set by using a preset image denoising algorithm to generate a denoised face image and obtain a denoised face image set. The preset image denoising algorithm can be: a fast mean denoising algorithm.

[0069] In the ninth step, each denoised face image in the denoised face image set is subjected to image sharpening processing to generate a sharpened face image, thereby obtaining a sharpened face image set. For each denoised face image in the denoised face image set, a filter is used to perform image sharpening processing on the denoised face image to generate a sharpened face image, thereby obtaining a sharpened face image set. The filter may be a convolution filter.

[0070] Optionally, the above pan / tilt omnidirectional infrared ball camera is configured as: The first step is to acquire a second detection area image corresponding to each second area to be detected in response to receiving a target trajectory prediction instruction, to obtain a second detection area image set.

[0071] In the second step, target detection and target frame marking are performed on each second detection area image of the second detection area image set to generate a target mark detection area image and obtain a target mark detection area image set. The target mark detection area image has a bounding box of the marked target. For example, the pan-tilt omnidirectional infrared ball camera can use an initialization tracker to perform target detection and target frame marking on each second detection area image of the second detection area image set to generate a target mark detection area image and obtain a target mark detection area image set. The target mark detection area image has a bounding box of the marked target. The initialization tracker can be: a multiple instance learning (MIL) tracker.

[0072] In the third step, target tracking processing is performed on each target mark detection area image in the target mark detection area image set to generate a mark detection area image target position, and obtain a mark detection area image target position set. For example, the pan-tilt omnidirectional infrared ball camera can use a tracker to perform target tracking processing on each target mark detection area image in the target mark detection area image set to generate a mark detection area image target position, and obtain a mark detection area image target position set. The tracker can be: a kernel correlation filter (KCF) tracker.

[0073] The fourth step is to perform trajectory prediction processing on the above-mentioned target position set of the mark detection area image to generate a predicted trajectory of the target in the mark detection area image. For example, the above-mentioned pan-tilt omnidirectional infrared ball camera can perform trajectory prediction processing on the above-mentioned target position set of the mark detection area image through a filter to generate a predicted trajectory of the target in the mark detection area image. The filter can be: a Kalman filter.

[0074] In the fifth step, target point feature extraction processing is performed on each second detection area image in the second detection area image set to generate feature extraction target points and obtain a feature extraction target point set. For example, the pan-tilt omnidirectional infrared ball camera can perform target point feature extraction processing on each second detection area image in the second detection area image set by using a preset target point feature extraction algorithm to generate feature extraction target points and obtain a feature extraction target point set. The preset target point feature extraction algorithm can be: Oriented FAST and Rotated BRIEF (ORB) algorithm.

[0075] In the sixth step, target point feature matching processing is performed on each feature extraction target point in the feature extraction target point set to generate a feature matching target point and obtain a feature matching target point set. For example, the pan-tilt omnidirectional infrared ball camera can perform target point feature matching processing on each feature extraction target point in the feature extraction target point set through a preset matcher to generate a feature matching target point and obtain a feature matching target point set. The preset matcher can be: feature matcher (BFMatcher).

[0076] In the seventh step, based on the above-mentioned feature matching target point set, target image registration processing is performed on each second detection area image in the above-mentioned second detection area image set to generate a registered target detection area image and obtain a registered target detection area image set. For example, the above-mentioned pan-tilt omnidirectional infrared ball camera can perform target image registration processing on each second detection area image in the above-mentioned second detection area image set by using a preset target image registration matrix to generate a registered target detection area image and obtain a registered target detection area image set. The preset target image registration matrix can be: calculating a homography matrix (Homography).

[0077] In the eighth step, the above-mentioned registered target detection area image set is subjected to target image fusion processing to generate a fused target detection area image. For example, the above-mentioned pan-tilt omnidirectional infrared ball camera can use a preset target image fusion technology to perform target image fusion processing on the above-mentioned registered target detection area image set to generate a fused target detection area image. The preset target image fusion technology can be: perspective transformation (warpPerspective) technology.

[0078] In the ninth step, based on the target prediction trajectory of the above-mentioned mark detection area image, the above-mentioned fused target detection area image is subjected to target locking processing to generate the future position of the target in the mark detection area image. For example, the above-mentioned pan-tilt omnidirectional infrared ball camera can perform target locking processing on the above-mentioned fused target detection area image through a preset target locking algorithm to generate the future position of the target in the mark detection area image. The preset target locking algorithm can be: tracking algorithm (KCF).

[0079] Optionally, the above-mentioned power data storage terminal is configured as: In some embodiments, the above-mentioned power operation and maintenance image set is compressed to generate a compressed power operation and maintenance image set, and the compressed power operation and maintenance image set is stored in a hard disk recorder. For example, the above-mentioned power data storage terminal can compress the above-mentioned power operation and maintenance image set through a preset compression algorithm to generate a compressed power operation and maintenance image set, and store the compressed power operation and maintenance image set in a hard disk recorder. The hard disk recorder can be: a personal video recorder (PVR), a network video recorder (NVR). The local hard disk can be: a serial (SATA) hard disk. The preset compression algorithm can be: a Joint Photographic Experts Group (JPEG) algorithm.

[0080] Optionally, the algorithm server is configured as follows: In the first step, each power operation and maintenance image in the power operation and maintenance image set is subjected to image recognition processing to generate an identified power operation and maintenance image, thereby obtaining an identified power operation and maintenance image set. For example, the algorithm server may perform image recognition processing on each power operation and maintenance image in the power operation and maintenance image set by using a preset image recognition algorithm to generate an identified power operation and maintenance image, thereby obtaining an identified power operation and maintenance image set. The preset image recognition algorithm may be a support vector machine (SVM) algorithm.

[0081] In the second step, the above-mentioned identified power operation and maintenance image set is processed for anomaly detection to generate an abnormal detection power operation and maintenance image result. For example, the above-mentioned algorithm server can perform anomaly detection processing on the above-mentioned identified power operation and maintenance image set through a preset anomaly detection algorithm to generate an abnormal detection power operation and maintenance image result. The abnormal detection power operation and maintenance image result can be: the operating state of the power equipment is abnormal, and the operating state of the power equipment is normal. The preset anomaly detection algorithm can be: an outlier detection algorithm.

[0082] Optionally, the edge computing terminal is configured as follows: The first step is to perform power operation and maintenance image preprocessing on the power operation and maintenance image set to generate a preprocessed power operation and maintenance image set. For example, first, the edge computing terminal can perform power operation and maintenance image cleaning on the power operation and maintenance image set to generate a cleaned power operation and maintenance image set. Secondly, perform power operation and maintenance image feature selection on the cleaned power operation and maintenance image set to generate a feature-selected power operation and maintenance image set. Then, perform table standardization on the feature-selected power operation and maintenance image set to generate a preprocessed power operation and maintenance image set.

[0083] In the second step, based on the above abnormal detection power operation and maintenance image results, the above pre-processed power operation and maintenance image set is subjected to power operation and maintenance image abnormality analysis processing to generate abnormal analysis power operation and maintenance image results. For example, the above edge computing terminal can perform power operation and maintenance image abnormality analysis processing on the above pre-processed power operation and maintenance image set through a preset abnormal analysis algorithm to generate abnormal analysis power operation and maintenance image results. The preset abnormal analysis algorithm can be: isolation forest algorithm.

[0084] In the third step, in response to determining that the abnormal power operation and maintenance image result meets the preset power operation and maintenance abnormal alarm condition, the associated alarm device is controlled to issue an abnormal alarm. The preset power operation and maintenance abnormal alarm condition may be: the operating state of the power device is abnormal. For example, the alarm device may be a speaker that emits an alarm prompt sound.

[0085] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. An intelligent perception system for power operation and maintenance, comprising: a power data perception terminal, a power data storage terminal and a power calculation terminal, characterized in that: The power data sensing terminal is communicatively connected with the power data storage terminal and the power calculation terminal, wherein the power data sensing terminal includes: direct reading instrument equipment, power inspection drone equipment, power inspection robot equipment, and image acquisition equipment; The power data storage terminal is communicatively connected with the power calculation terminal, wherein the power calculation terminal includes: an algorithm server and an edge computing terminal; The direct reading instrument device is used to collect meter reading image sets corresponding to each power equipment meter, wherein the direct reading instrument device includes: a direct reading instrument and a direct reading instrument gateway; The electric power inspection drone equipment is used to conduct aerial inspections of inspection targets and collect inspection target image sets of each inspection target, wherein the electric power inspection drone equipment includes: drones and drone parking stations; The electric power inspection robot device is used to inspect the inspection targets and collect inspection target image sets of each inspection target on the ground, wherein the electric power inspection robot device includes: a robot, a base station, and a robot workstation; The image acquisition device is used to acquire a regional image set of each area to be detected, wherein the image acquisition device includes: a fixed visible light gun camera, a pan-tilt omnidirectional visible light ball camera, a fixed infrared gun camera, and a pan-tilt omnidirectional infrared ball camera; The algorithm server is communicatively connected with the edge computing terminal, the algorithm server is used to identify abnormal conditions in the power operation and maintenance image set, and the edge computing terminal is used to perform power operation and maintenance image preprocessing and power operation and maintenance image abnormality analysis of the power operation and maintenance image set, wherein the power operation and maintenance image set includes: a meter reading image set, a detection target image set, an inspection target image set, and a regional image set; The power data storage terminal is configured to: compress the power operation and maintenance image set to generate a compressed power operation and maintenance image set, and store the compressed power operation and maintenance image set in a hard disk recorder; Wherein, the algorithm server is configured as: Performing image recognition processing on each power operation and maintenance image in the power operation and maintenance image set to generate an identified power operation and maintenance image, thereby obtaining an identified power operation and maintenance image set; Anomaly detection processing is performed on the identified power operation and maintenance image set to generate anomaly detection power operation and maintenance image results.

2. The power operation and maintenance intelligent sensing system according to claim 1 is characterized in that: The power inspection drone equipment is configured to: In response to receiving the aerial patrol instruction, determining the current position of the patrol drone and the positions of each target to be detected, and planning a patrol path from the current position to the positions of each target to be detected; The drone is controlled to move along the patrol path, and the drone is controlled to collect and process images of each target to be detected to generate a detection target image, obtain a detection target image set, and send the collected detection target image set to the power data storage terminal and the power calculation terminal.

3. The power operation and maintenance intelligent sensing system according to claim 1 is characterized in that: The electric power inspection robot device is configured to: In response to receiving a ground inspection instruction, determining the current position of the inspection robot and the positions of each target to be inspected, and planning an inspection path from the current position to the positions of each target to be inspected; The robot is controlled to move along the inspection path, and is controlled to perform image acquisition and processing on each target to be inspected to generate an inspection target image, obtain an inspection target image set, and upload the inspection target image set to the power data storage terminal and the power calculation terminal through the base station.

4. The power operation and maintenance intelligent sensing system according to claim 1 is characterized in that: The fixed infrared camera is configured as follows: In response to receiving a face detection instruction, collecting a first detection area image corresponding to each first to-be-detected area to obtain a first detection area image set; Performing image preprocessing on each first detection area image in the first detection area image set to generate a preprocessed detection area image, thereby obtaining a preprocessed detection area image set; Performing face positioning processing on each preprocessed detection area image in the preprocessed detection area image set to generate a face positioning image, thereby obtaining a face positioning image set; Performing multi-scale detection processing on each face positioning image in the face positioning image set to generate a multi-scale detected face image, thereby obtaining a multi-scale detected face image set, wherein a face in the multi-scale detected face image has a bounding box; Performing face region cropping processing on each multi-scale detected face image in the multi-scale detected face image set to generate a region cropped face image, thereby obtaining a region cropped face image set; Performing face frame drawing processing on each region-cropped face image in the region-cropped face image set to generate a frame-drawn face image, thereby obtaining a frame-drawn face image set; Performing face image alignment processing on each frame-drawn face image in the frame-drawn face image set to generate an aligned face image, thereby obtaining an aligned face image set; Performing image denoising processing on each aligned face image in the aligned face image set to generate a denoised face image, thereby obtaining a denoised face image set; Performing image sharpening processing on each denoised face image in the denoised face image set to generate a sharpened face image, thereby obtaining a sharpened face image set.

5. The power operation and maintenance intelligent sensing system according to claim 1 is characterized in that: The edge computing terminal is configured to: Performing power operation and maintenance image preprocessing on the power operation and maintenance image set to generate a preprocessed power operation and maintenance image set; Based on the abnormal detection power operation and maintenance image result, performing power operation and maintenance image abnormality analysis processing on the pre-processed power operation and maintenance image set to generate an abnormal analysis power operation and maintenance image result; In response to determining that the abnormal power operation and maintenance image result meets the preset power operation and maintenance abnormal alarm condition, the associated alarm device is controlled to issue an abnormal alarm.

Images