An unmanned aerial vehicle and AI vision-based power facility intelligent inspection and safety hazard identification method

Abstract

The application discloses a kind of power facilities intelligent inspection and safety hazard identification method based on unmanned aerial vehicle and AI vision, according to the real-time positioning and remote sensing data of unmanned aerial vehicle, adjust the environment perception mode of unmanned aerial vehicle;According to the perception live of unmanned aerial vehicle, change the inspection state of unmanned aerial vehicle;Environment image and shooting state feature generated during the inspection of unmanned aerial vehicle are acquired, to estimate the image acquisition interference of unmanned aerial vehicle;According to image acquisition interference, pre-process environment image;AI vision identification is carried out to environment image, and the construction feature of power facility is obtained;According to construction feature, the abnormal component of power facility is identified;According to the distribution of abnormal component in power facility, estimate the insulation deterioration trend of power facility, to identify the safety hazard state of power facility.Through the combination of unmanned aerial vehicle positioning and remote sensing data control inspection, also carry out interference exclusion and identification to environment image, identify abnormal position, then realize facility insulation performance prediction, improve inspection accuracy and reliability.

Description

Technical Field

[0001] This invention relates to the field of machine vision, and in particular to a method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision. Background Technology

[0002] Power facilities, such as high-voltage towers and lines, require regular inspection and maintenance. Given the high-risk and harsh environment inherent in power facility operations, drones are widely used for power line inspections. They can cover dangerous and complex terrain areas and offer significant flexibility and efficiency. Drones used for power line inspections typically acquire real-time operating data of power facilities through image capture and thermal infrared detection, and then use neural network models to process this data to quickly locate faults. However, existing drones are affected by their own internal and external factors during inspections, leading to interference in the acquired environmental images. This results in inaccurate reflection of the actual condition of the power facilities and an inability to accurately predict changes in the insulation performance of the facilities, reducing the accuracy of power facility inspections and the reliability of safety identification. Summary of the Invention

[0003] Considering that existing drones for power facility inspection cannot accurately match the inspection perception mode to the power facilities, nor can they eliminate interference from the collected environmental images, and that the inspection purpose is limited to surface damage and does not address safety hazards caused by deterioration in the insulation performance of the power facilities themselves, the accuracy and reliability of the inspections are reduced. In view of the above problems, this invention proposes a method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision, including:

[0004] Based on the real-time positioning and remote sensing data of the UAV, adjust the environmental perception mode of the UAV; based on the actual perception status of the UAV, change the inspection status of the UAV.

[0005] The system acquires environmental images and shooting status features generated during the UAV inspection, estimates image acquisition interference of the UAV based on the shooting status features, and preprocesses the environmental images based on the image acquisition interference.

[0006] AI visual recognition is used to obtain the structural features of the power facilities from the environmental images; based on the structural features, abnormal components of the power facilities are identified.

[0007] Based on the distribution of the abnormal components in the power facility, the insulation degradation trend of the power facility is estimated, thereby identifying the potential safety hazards of the power facility.

[0008] Optionally, the environmental perception mode of the UAV can be adjusted based on its real-time positioning and remote sensing data, including:

[0009] By comparing the real-time positioning information of the UAV with the remote sensing monitoring geographical boundary of the area where the UAV is currently located, remote sensing sub-data that spatially matches the UAV is extracted from the remote sensing data of the area.

[0010] Spatial features of environmental objects related to the UAV are extracted from the remote sensing sub-data; wherein, the spatial features of environmental objects refer to the spatial location and size of obstacles to the UAV and environmental objects that the UAV expects to detect; the environmental perception mode of the UAV is adjusted according to the spatial features of environmental objects; wherein, the environmental perception mode includes the type of perception means of the surrounding environment.

[0011] Optionally, the inspection status of the drone can be changed based on the drone's real-time perception, including:

[0012] The multimodal perception real-time data of the UAV is time-series aligned and target entity is identified to obtain the spatial distribution of obstacles and expected detection objects during the flight of the UAV.

[0013] Based on the spatial distribution and the static and dynamic attributes of the UAV, the inspection flight trajectory of the UAV is changed; wherein, the static and dynamic attributes refer to the variable range of the UAV's external dimensions and flight attitude angles.

[0014] Optionally, environmental images and shooting status features generated during the UAV inspection are acquired, and image acquisition interference of the UAV is estimated based on the shooting status features, including:

[0015] The system acquires environmental image data streams and shooting status feature data streams generated during the UAV inspection; wherein, the shooting status feature data streams include ambient light change data streams and shooting shake change data streams during the shooting of environmental images.

[0016] Based on the shooting state feature data stream, the exposure interference and image capture jitter interference of the UAV are estimated.

[0017] Optionally, the environmental image is preprocessed according to the image acquisition interference, including:

[0018] Based on the occurrence time intervals of exposure interference and jitter interference in the image acquisition frame, the potentially interfered image frames of the environmental image are identified.

[0019] Based on the intensity of exposure interference and jitter interference in the image acquisition frame, interference removal and repair preprocessing are performed on the potentially interfered image frames.

[0020] Optionally, based on the occurrence time intervals of the exposure interference and the jitter interference of the image acquisition frame, the potentially interfered image frames of the environmental image are identified, including:

[0021] Retrieve the occurrence time intervals of exposure interference and jitter interference in the image acquisition frame;

[0022] An exposure interference intensity function E(t) is set for each time point of image acquisition based on the time interval of exposure interference occurrence in the image acquisition scene. The structure of the exposure interference intensity function E(t) is as follows:

[0023] ;

[0024] Where E(t) represents the exposure interference intensity function; t xs and t xe These represent the start and end times of the time interval corresponding to the occurrence of exposure interference, respectively; t represents the image acquisition time; δ represents the gradient width of the exposure interference edge, used to characterize the total transition time from none to peak interference and then from peak interference back to none.

[0025] The jitter intensity function J(t) for each time point of image acquisition is set using the time interval of the image acquisition jitter interference occurrence. The structure of the jitter intensity function J(t) is as follows:

[0026] ;

[0027] Where J(t) represents the jitter interference intensity function; t ys and t ye These represent the start and end times of the time interval corresponding to the occurrence of image acquisition jitter interference, respectively; k represents the damping coefficient, with a value range of [0.2, 2.0].

[0028] The joint interference index I(t) for each time point of image acquisition is set using the exposure interference intensity function E(t) of the image acquisition image exposure interference and the jitter interference intensity function J(t) of the image acquisition image jitter interference. The joint interference index I(t) for each time point is obtained using the following formula:

[0029] I(t)=[E(t)+J(t)+α×E(t)×J(t)]×(1+α) -1

[0030] Where I(t) represents the joint interference index corresponding to each time point of image acquisition; α represents the coordination coefficient, which takes a value range of (0, 0.5] and is used to enhance the interference weight of the area where the interference occurs simultaneously.

[0031] The joint interference index I(t) corresponding to each time point of image acquisition is compared with a preset exponential threshold. When the joint interference index I(t) corresponding to each time point of image acquisition exceeds the preset exponential threshold, the image frame corresponding to that time point t is marked as a potentially interfered image frame. The exponential threshold is set by the following formula:

[0032] Th(t)=Th0×exp[-β×E(t) g ×J(t) g ]

[0033] Where Th(t) represents the exponential threshold; Th0 represents the preset initial threshold; β represents the adjustment coefficient, with a value range of (0, 1); E(t) g and J(t) g These represent the average values ​​of exposure interference intensity and jitter interference intensity at all time points during image acquisition, respectively.

[0034] The consecutive potentially interfered image frames are integrated to form a potentially interfered image frame segment.

[0035] Optionally, AI visual recognition is performed on the environmental image to obtain the structural features of the power facilities, including:

[0036] AI visual recognition is performed on the environmental image to obtain pixel contour features and pixel color features;

[0037] Cluster analysis is performed on the pixel contour features and the pixel chromaticity features to obtain the structural features of the power facility; wherein, the structural features include the structural features of the insulating components of the power facility.

[0038] Optionally, based on the aforementioned structural features, the abnormal components of the power facility are identified, including:

[0039] The damaged structural features of the insulating components of the power facility are extracted from the structural features, and the extent of the exposed parts of the insulating components is estimated based on the damaged structural features.

[0040] Based on the location range, identify whether the insulating component is an abnormal component.

[0041] Optionally, based on the distribution of the abnormal components in the power facility, the insulation degradation trend of the power facility is estimated, thereby identifying the safety hazard status of the power facility, including:

[0042] Based on the distribution location of the abnormal components in the power facility and their direct exposure area to the external environment, the intrusion rate of the power facility by external substances is estimated.

[0043] Based on the intrusion speed, the insulation degradation trend of the power facility is estimated; wherein the insulation degradation trend includes at least the time when the power facility experiences an insulation failure event;

[0044] Based on the insulation degradation trend and the operating condition of the power facility, identify the safety hazard status of the power facility; wherein, the safety hazard status includes the time when the power facility experiences a preset arc discharge event.

[0045] The beneficial effects of the above-mentioned technical solutions provided in the embodiments of the present invention include at least the following:

[0046] This invention provides a method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision. The method adjusts the drone's environmental perception mode based on real-time drone positioning and remote sensing data; changes the drone's inspection status according to its perception; acquires environmental images and shooting status features generated during drone inspection to estimate image acquisition interference; preprocesses the environmental images based on the image acquisition interference; performs AI vision recognition on the environmental images to obtain the structural features of the power facilities; identifies abnormal components of the power facilities based on the structural features; and estimates the insulation degradation trend of the power facilities based on the distribution of abnormal components, thereby identifying the safety hazard status of the power facilities. By combining drone positioning and remote sensing data to control the inspection, and also performing interference elimination and identification on the environmental images to identify abnormal parts, the method achieves facility insulation performance prediction, improving the accuracy and reliability of the inspection.

[0047] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0048] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0049] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0050] Figure 1 This is a flowchart illustrating a method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision, provided in an embodiment of the present invention. Detailed Implementation

[0051] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0052] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," "far," "near," "front," and "rear," etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0053] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0054] Please see Figure 1 As shown in the figure, an embodiment of this application provides a method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision. This method includes:

[0055] Adjust the drone's environmental perception mode based on the drone's real-time positioning and remote sensing data; change the drone's inspection status based on the drone's actual perception status.

[0056] Acquire environmental images and shooting status features generated during UAV inspections; estimate image acquisition interference from the UAV based on the shooting status features; and preprocess the environmental images based on the image acquisition interference.

[0057] AI visual recognition is used to identify the structural features of power facilities from environmental images; based on these features, abnormal components of the power facilities are identified.

[0058] Based on the distribution of abnormal components in power facilities, the insulation degradation trend of power facilities can be estimated, thereby identifying potential safety hazards in power facilities.

[0059] The beneficial effects of the above embodiments are as follows: This intelligent inspection and safety hazard identification method for power facilities based on UAVs and AI vision combines UAV positioning and remote sensing data to control the inspection, and also performs interference elimination and identification on environmental images, identifies abnormal parts, and then realizes the prediction of facility insulation performance, thereby improving the accuracy and reliability of the inspection.

[0060] In another embodiment, adjusting the UAV's environmental perception mode based on the UAV's real-time positioning and remote sensing data includes:

[0061] By comparing the real-time positioning information of the UAV with the remote sensing monitoring geographical boundary of the area where the UAV is currently located, remote sensing sub-data that spatially matches the UAV is extracted from the remote sensing data of the area.

[0062] Spatial features of environmental objects related to the UAV are extracted from remote sensing sub-data. These features refer to the spatial location and size of environmental objects that pose obstacles to the UAV and that the UAV expects to detect. Based on these spatial features, the UAV's environmental perception mode is adjusted. The environmental perception mode includes the type of perception method used to perceive the surrounding environment.

[0063] The beneficial effects of the above embodiments: High-voltage power towers or lines are usually installed in complex terrain areas such as mountains or jungles. To ensure the safe and stable execution of inspections, drones need to predict potential obstacles and take appropriate obstacle avoidance measures during inspection flights. Existing drones are generally equipped with cameras and radar sensors, which can detect the presence of surrounding objects through optical imaging and radar monitoring. However, optical imaging and radar monitoring have limited effective detection ranges in complex terrain areas and cannot detect environmental objects at greater distances in advance. It is understood that the aforementioned environmental objects can be, but are not limited to, obstacles encountered by the drone during inspections and targets to be monitored (i.e., power facilities). To enable the drone to form a basic understanding of the presence of objects in its new environmental space, it is necessary to combine the pre-formed global remote sensing monitoring results of the area to form a preliminary spatial distribution representation of environmental objects, providing effective guidance for drone inspection flights. Specifically, by comparing the drone's real-time positioning information with the remote sensing monitoring geographical boundary of the area where the drone is currently located, remote sensing sub-data matching the drone is extracted. The aforementioned remote sensing sub-data can be, but is not limited to, remote sensing sub-data within a preset radius centered on the drone's location. Then, object contour recognition is performed on the aforementioned remote sensing sub-data to obtain the spatial location and size of environmental objects that obstruct the drone's flight and the environmental objects that the drone wants to detect. It is understandable that the spatial characteristics of environmental objects related to the drone directly reflect the relative positional relationship between the drone and these objects. Based on these spatial characteristics, the drone's perception methods for its surroundings are adjusted. For example, when environmental objects around the drone are within a preset distance range, a combination of optical and thermal imaging can be used for environmental perception; when environmental objects are outside the preset distance range, a combination of optical imaging and radar detection can be used. By adjusting the perception methods, all environmental objects encountered during the inspection can be comprehensively and without omission, providing a basis for subsequently determining the drone's inspection flight path.

[0064] In another embodiment, the inspection status of the drone is changed based on the drone's real-time perception data, including:

[0065] By performing time-series alignment and target entity recognition on the multimodal perception real-time data of the UAV, the spatial distribution of obstacles and expected detection objects during the UAV's flight can be obtained;

[0066] The inspection flight trajectory of the UAV is changed according to the spatial distribution and the static and dynamic attributes of the UAV; where static and dynamic attributes refer to the variable range of the UAV's external dimensions and flight attitude angles.

[0067] The beneficial effects of the above embodiments are as follows: The UAV is equipped with various environmental perception devices such as visible light cameras, thermal infrared cameras, and radar sensors. When the UAV is in the corresponding environmental perception mode, at least some of the environmental perception devices inside the UAV work synchronously to form multimodal perception real-time data. The aforementioned multimodal perception real-time data may include, but is not limited to, image data and radar sensing data, which reflect the presence of environmental objects around the UAV from different modal levels. Considering that obstacles and targets exist simultaneously around the UAV, and that the UAV needs to avoid obstacles and detect targets during flight inspection, it is necessary to distinguish between obstacles and targets in order for the UAV to perform its inspection tasks normally. Specifically, the multimodal perception real-time data of the UAV is time-series aligned (i.e., the acquisition time stamps of all modal perception real-time data are timestamped) and target entity identified to obtain the spatial distribution of obstacles and expected detection objects during the UAV's flight. Understandably, in order to avoid obstacles and detect the status of targets during inspection flights, UAVs need to calibrate an inspection flight trajectory that is suitable for obstacle avoidance and accurate target detection in the spatial distribution of obstacles and objects to be detected, based on the UAV's external dimensions and variable range of flight attitude angles, so as to achieve efficient and reliable flight control of the UAV.

[0068] In another embodiment, environmental images and shooting status features generated during UAV inspection are acquired, and image acquisition interference of the UAV is estimated based on the shooting status features, including:

[0069] Acquire environmental image data streams and shooting status feature data streams generated during UAV inspections; among which, the shooting status feature data streams include data streams of changes in ambient light and shooting jitter during the capture of environmental images;

[0070] Based on the shooting status characteristic data stream, estimate the exposure interference and image jitter interference of the image acquisition screen of the UAV.

[0071] The beneficial effects of the above embodiments are as follows: The UAV determines the structural state of power facilities (e.g., whether structural damage has occurred) through visual recognition. Therefore, the environmental images generated during UAV inspections directly reflect the structural condition of power facilities and other targets. Considering that ambient light intensity and camera shake during UAV inspections can affect environmental images, in order to accurately identify the actual structural condition of power facilities from environmental images, the environmental image data stream and shooting state feature data stream generated during UAV inspections are acquired simultaneously. The ambient light intensity and the amplitude and direction of camera shake during UAV image capture are correlated with the shooting state features. Furthermore, based on the ambient light change data stream and camera shake change data stream during environmental image capture, exposure interference and camera shake interference in the UAV's image acquisition are estimated, providing a basis for subsequent environmental image correction.

[0072] In another embodiment, environmental image preprocessing based on image acquisition interference includes:

[0073] Based on the time intervals of exposure interference and jitter interference in the image acquisition frame, the potentially interfered image frames of the environmental image are identified.

[0074] Based on the intensity of exposure interference and jitter interference in the image acquisition frame, interference removal and repair preprocessing are performed on potentially interfered image frames.

[0075] The beneficial effects of the above embodiments are as follows: The UAV is not affected by ambient light interference and camera shake interference throughout the entire inspection and shooting process. Instead, it is affected by ambient light interference and / or camera shake interference during certain periods. To avoid increasing workload by correcting the entire environmental image, the potentially interfered image frames are first identified based on the time intervals of exposure interference and camera shake interference in the image acquisition frame. For example, these time intervals can be mapped to the shooting timeline of the environmental image to determine the timestamps of the potentially interfered image frames. These potentially interfered image frames may include several consecutively distributed image frames. Then, based on the intensity of exposure interference and camera shake interference in the image acquisition frame, interference removal and repair preprocessing are performed on the potentially interfered image frames. For example, a convolutional neural network is used to perform noise reduction processing on each potentially interfered image frame.

[0076] In another embodiment, based on the occurrence time intervals of the exposure interference and the jitter interference of the image acquisition frame, the potentially interfered image frames of the environmental image are identified, including:

[0077] Retrieve the occurrence time intervals of exposure interference and jitter interference in the image acquisition frame;

[0078] An exposure interference intensity function E(t) is set for each time point of image acquisition based on the time interval of exposure interference occurrence in the image acquisition scene. The structure of the exposure interference intensity function E(t) is as follows:

[0079] ;

[0080] Where E(t) represents the exposure interference intensity function; t xs and t xeThese represent the start and end times of the exposure interference occurrence time interval, respectively; t represents the image acquisition time; δ represents the width of the exposure interference edge gradient, which is used to characterize the total transition time from no interference to peak value and then from peak value back to no interference. Typically, the width of the exposure interference edge gradient is 1 / 4 or 1 / 3 of the exposure interference occurrence time interval corresponding to the time length.

[0081] The jitter intensity function J(t) for each time point of image acquisition is set using the time interval of the image acquisition jitter interference occurrence. The structure of the jitter intensity function J(t) is as follows:

[0082] ;

[0083] Where J(t) represents the jitter interference intensity function; t ys and t ye These represent the start and end times of the time interval corresponding to the occurrence of image acquisition jitter interference, respectively; k represents the damping coefficient, with a value range of [0.2, 2.0], which is preset according to the drone model or gimbal model and is used to control the speed of jitter attenuation;

[0084] The joint interference index I(t) for each time point of image acquisition is set using the exposure interference intensity function E(t) of the image acquisition image exposure interference and the jitter interference intensity function J(t) of the image acquisition image jitter interference. The joint interference index I(t) for each time point is obtained using the following formula:

[0085] I(t)=[E(t)+J(t)+α×E(t)×J(t)]×(1+α) -1

[0086] Where I(t) represents the joint interference index corresponding to each time point of image acquisition; α represents the coordination coefficient, with a value range of (0, 0.5], which is used to enhance the interference weight of the area where the interference occurs simultaneously. This formula introduces an interaction term on the basis of simple linear superposition, so that the interference degree of the overlapping area is reasonably amplified.

[0087] The joint interference index I(t) corresponding to each time point of image acquisition is compared with a preset exponential threshold. When the joint interference index I(t) corresponding to each time point of image acquisition exceeds the preset exponential threshold, the image frame corresponding to that time point t is marked as a potentially interfered image frame. The exponential threshold is set by the following formula:

[0088] Th(t)=Th0×exp[-β×E(t) g ×J(t) g ]

[0089] Where Th(t) represents the exponential threshold; Th0 represents the preset initial threshold; β represents the adjustment coefficient, with a value range of (0, 1); E(t) g and J(t) g These represent the average values ​​of exposure interference intensity and jitter interference intensity at all time points during image acquisition, respectively.

[0090] The consecutive potentially interfered image frames are integrated to form a potentially interfered image frame segment.

[0091] The beneficial effects of the above embodiments are as follows: The above technical solutions achieve multi-dimensional technical improvements in actual operation, effectively ensuring the reliability and efficiency of intelligent inspection of power facilities. Regarding interference quantification accuracy, the piecewise linearly varying exposure interference intensity function controls the step error of interference intensity within a negligible range. Compared to the traditional step model, this effectively reduces the intensity fitting error at the beginning and end of the interference, more accurately matching the physical gradual change process of actual exposure interference. The damped sinusoidal jitter interference intensity function achieves precise control of jitter attenuation speed through the damping coefficient, effectively improving the time-domain fitting degree of jitter intensity under different gimbal models, significantly outperforming the fitting effect of the fixed amplitude model. In terms of joint interference evaluation performance, the normalized joint interference index introduces an interaction term and uses (1+α)... -1 The constraints effectively enhance the stability of the weighting ratio in the interference overlap region, avoiding both the underestimation of collaborative interference by a single linear superposition model and the numerical distortion caused by excessive amplification of interaction terms. Compared with the traditional linear superposition model, this effectively improves the accuracy of joint interference assessment. Furthermore, regarding the robustness of interference judgment, the dynamic exponential threshold Th(t) adopts an exponential decay model, correlating the global average interference intensity with the initial threshold. In scenarios with high overall interference levels, this effectively improves the accuracy and stability of threshold adaptive adjustment and its matching degree with the actual situation of UAV image acquisition, thereby effectively improving the detection rate of potentially interfered frames. In low-interference scenarios, the threshold remains stable, reducing the false alarm rate. Through precise quantification, reasonable weighting, and dynamic adjustment, the comprehensive performance indicators of interference identification are significantly optimized, providing high-quality image preprocessing support for subsequent AI visual hazard identification and ensuring the efficiency and safety of power facility inspection.

[0092] In another embodiment, AI visual recognition is performed on the environmental image to obtain the structural features of the power facility, including:

[0093] AI visual recognition is used to obtain pixel contour features and pixel color features from environmental images;

[0094] Cluster analysis is performed on pixel contour features and pixel chromaticity features to obtain the structural features of power facilities; among which, structural features include the structural features of the insulating components of power facilities.

[0095] The beneficial effects of the above embodiments are as follows: In practical operation, convolutional neural networks can be used to perform AI visual recognition on environmental images to obtain pixel contour features and pixel chromaticity features within each image frame of the environmental image. Then, cluster analysis is performed on the pixel contour features and the pixel chromaticity features to obtain the structural features of the insulating components of power facilities; wherein, the structural features of the insulating components may include, but are not limited to, the surface contour features and surface material coverage features of the insulating components.

[0096] In another embodiment, identifying abnormal components of the power facility based on structural features includes:

[0097] Damage structural features of insulation components of power facilities are extracted from structural features, and the extent of exposed parts of insulation components is estimated based on the damage structural features.

[0098] Based on the location and scope, identify whether the insulating component is an abnormal component.

[0099] The beneficial effects of the above embodiments are as follows: In actual operation, the damaged structural features of the insulating components of power facilities are extracted from the structural features. These damaged structural features are then compared with the external structural features of the insulating components in an intact state to estimate the extent of exposure of the insulating components, such as the exposed area caused by damage to the surface coating. It is then determined whether the area of ​​this area exceeds a preset area threshold. If so, the insulating component is determined to be an abnormal component; otherwise, it is determined not to be an abnormal component.

[0100] In another embodiment, based on the distribution of abnormal components within the power facility, the insulation degradation trend of the power facility is estimated, thereby identifying potential safety hazards in the power facility, including:

[0101] Based on the distribution location of abnormal components in the power facility and their direct exposure area to the external environment, the intrusion rate of external substances into the power facility is estimated.

[0102] Based on the intrusion speed, estimate the insulation degradation trend of the power facility; wherein the insulation degradation trend includes at least the time when the power facility experiences an insulation failure event;

[0103] Based on the insulation degradation trend and the operating condition of the power facilities, identify the safety hazard status of the power facilities; among which, the safety hazard status includes the time when the power facilities will experience a preset arc discharge event.

[0104] The beneficial effects of the above embodiments are as follows: In actual operation, based on the distribution location of abnormal components in the power facility and their direct exposure area to the external environment, the rate of intrusion of external substances into the power facility is estimated, such as the amount of external moisture intrusion per unit time. Based on this rate of intrusion, the cumulative amount of external substances intruding into the power facility is determined, thereby estimating the time when an insulation failure event occurs. Furthermore, based on the time of the insulation failure event and the operating conditions of the power facility (such as the power transmission parameters), the time when a preset arc discharge event occurs is identified, providing a reliable basis for subsequently changing the operating state of the power facility in advance to prevent the escalation of danger.

[0105] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision, characterized in that, include: The environmental perception mode of the UAV is adjusted based on the UAV's real-time positioning and remote sensing data. Based on the real-time perception of the drone, change the inspection status of the drone; The system acquires environmental images and shooting status features generated during the UAV inspection, and estimates the image acquisition interference of the UAV based on the shooting status features. Preprocess the environmental image based on the image acquisition interference; AI visual recognition is used to obtain the structural features of the power facilities from the environmental images; based on the structural features, abnormal components of the power facilities are identified. Based on the distribution of the abnormal components in the power facility, the insulation degradation trend of the power facility is estimated, thereby identifying the potential safety hazards of the power facility.

2. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 1, characterized in that: Based on the real-time positioning and remote sensing data of the UAV, adjust the environmental perception mode of the UAV, including: By comparing the real-time positioning information of the UAV with the remote sensing monitoring geographical boundary of the area where the UAV is currently located, remote sensing sub-data that spatially matches the UAV is extracted from the remote sensing data of the area. Spatial features of environmental objects related to the UAV are extracted from the remote sensing sub-data; wherein, the spatial features of environmental objects refer to the spatial location and size of obstacles to the UAV and environmental objects that the UAV expects to detect; the environmental perception mode of the UAV is adjusted according to the spatial features of environmental objects; wherein, the environmental perception mode includes the type of perception means of the surrounding environment.

3. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 2, characterized in that: Based on the real-time perception data of the drone, the inspection status of the drone is changed, including: The multimodal perception real-time data of the UAV is time-series aligned and target entity is identified to obtain the spatial distribution of obstacles and expected detection objects during the flight of the UAV. Based on the spatial distribution and the static and dynamic attributes of the UAV, the inspection flight trajectory of the UAV is changed; wherein, the static and dynamic attributes refer to the variable range of the UAV's external dimensions and flight attitude angles.

4. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 1, characterized in that: Acquire environmental images and shooting status features generated during the UAV inspection, and estimate the image acquisition interference of the UAV based on the shooting status features, including: The system acquires environmental image data streams and shooting status feature data streams generated during the UAV inspection; wherein, the shooting status feature data streams include ambient light change data streams and shooting shake change data streams during the shooting of environmental images. Based on the shooting state feature data stream, the exposure interference and image capture jitter interference of the UAV are estimated.

5. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 4, characterized in that: Based on the image acquisition interference, the environmental image is preprocessed, including: Based on the occurrence time intervals of exposure interference and jitter interference in the image acquisition frame, the potentially interfered image frames of the environmental image are identified. Based on the intensity of exposure interference and jitter interference in the image acquisition frame, interference removal and repair preprocessing are performed on the potentially interfered image frames.

6. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 5, characterized in that: Based on the occurrence time intervals of exposure interference and jitter interference in the image acquisition frame, the potentially interfered image frames of the environmental image are identified, including: Retrieve the occurrence time intervals of exposure interference and jitter interference in the image acquisition frame; An exposure interference intensity function E(t) is set for each time point of image acquisition based on the time interval of exposure interference occurrence in the image acquisition scene. The structure of the exposure interference intensity function E(t) is as follows: ； Where E(t) represents the exposure interference intensity function; t xs and t xe These represent the start and end times of the time interval corresponding to the occurrence of exposure interference, respectively; t represents the image acquisition time; δ represents the gradient width of the exposure interference edge, used to characterize the total transition time from none to peak interference and then from peak interference back to none. The jitter intensity function J(t) for each time point of image acquisition is set using the time interval of the image acquisition jitter interference occurrence. The structure of the jitter intensity function J(t) is as follows: ； Where J(t) represents the jitter interference intensity function; t ys and t ye These represent the start and end times of the time interval corresponding to the occurrence of image acquisition jitter interference, respectively; k represents the damping coefficient, with a value range of [0.2, 2.0]. The joint interference index I(t) for each time point of image acquisition is set using the exposure interference intensity function E(t) of the image acquisition image exposure interference and the jitter interference intensity function J(t) of the image acquisition image jitter interference. The joint interference index I(t) for each time point is obtained using the following formula: I(t)=[E(t)+J(t)+α×E(t)×J(t)]×(1+α) -1 Where I(t) represents the joint interference index corresponding to each time point of image acquisition; α represents the coordination coefficient, which takes a value range of (0, 0.5] and is used to enhance the interference weight of the area where the interference occurs simultaneously. The joint interference index I(t) corresponding to each time point of image acquisition is compared with a preset exponential threshold. When the joint interference index I(t) corresponding to each time point of image acquisition exceeds the preset exponential threshold, the image frame corresponding to that time point t is marked as a potentially interfered image frame. The exponential threshold is set by the following formula: Th(t)=Th0×exp[-β×E(t)] g ×J(t) g ] Where Th(t) represents the exponential threshold; Th0 represents the preset initial threshold; β represents the adjustment coefficient, with a value range of (0, 1); E(t) g and J(t) g These represent the average values ​​of exposure interference intensity and jitter interference intensity at all time points during image acquisition, respectively. The consecutive potentially interfered image frames are integrated to form a potentially interfered image frame segment.

7. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 1, characterized in that: AI visual recognition is performed on the environmental image to obtain the structural features of the power facilities, including: AI visual recognition is performed on the environmental image to obtain pixel contour features and pixel color features; Cluster analysis is performed on the pixel contour features and the pixel chromaticity features to obtain the structural features of the power facility; wherein, the structural features include the structural features of the insulating components of the power facility.

8. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 6, characterized in that: Based on the aforementioned structural features, the abnormal components of the power facility are identified, including: The damaged structural features of the insulating components of the power facility are extracted from the structural features, and the extent of the exposed parts of the insulating components is estimated based on the damaged structural features. Based on the location range, identify whether the insulating component is an abnormal component.

9. The method for intelligent inspection and safety hazard identification of power facilities based on drones and AI vision as described in claim 7, characterized in that: Based on the distribution of the abnormal components in the power facility, the insulation degradation trend of the power facility is estimated, thereby identifying the safety hazard status of the power facility, including: Based on the distribution location of the abnormal components in the power facility and their direct exposure area to the external environment, the intrusion rate of the power facility by external substances is estimated. Based on the intrusion speed, the insulation degradation trend of the power facility is estimated; wherein the insulation degradation trend includes at least the time when the power facility experiences an insulation failure event; Based on the insulation degradation trend and the operating condition of the power facility, identify the safety hazard status of the power facility; wherein, the safety hazard status includes the time when the power facility experiences a preset arc discharge event.

Images