Unmanned aerial vehicle-based overhead cable line loss intelligent detection system and method

Abstract

The invention discloses an aerial cable line loss intelligent detection system and method based on an unmanned aerial vehicle, and relates to the technical field of electric power inspection. The problems that a traditional line loss detection system lacks comprehensive perception ability to a complex environment and cannot adaptively switch a flight detection mode and a close-range attachment detection mode are solved. Comprising an unmanned aerial vehicle body, a cable acquisition module, a line loss intelligent detection module, a data processing module and an unmanned aerial vehicle control module, a flight detection mode and a hanging detection mode can be intelligently switched, a multi-mode sensor is arranged, and combined acquisition and analysis of cable images, sound waves and magnetic induction data are achieved; abnormal points are extracted through algorithms such as image edges, thermal aberration, spectrum features and magnetic field disturbance, adaptive correction is performed in combination with the environment drift degree for fusion judgment, line loss grade classification and multi-source detection report output are achieved, and the detection precision, safety and environment adaptive capacity are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power inspection, in particular to an overhead cable line loss intelligent detection system and method based on a UAV. BACKGROUND

[0002] With the large deployment of urban power grid overhead lines, the energy loss problem in the operation process of the lines is increasingly serious. When the overhead cable has a line loss problem, it will affect the power supply efficiency, and thus may become a potential hidden danger causing fire or accident. At present, the cable line loss detection method mainly relies on manual inspection, fixed camera equipment or infrared temperature measuring device, which has the following main problems:

[0003] 1. Low detection efficiency: manual inspection relies on personnel close-range operation, which is time-consuming and inefficient, and it is difficult to achieve large-scale and continuous detection;

[0004] 2. Insufficient accuracy: most traditional devices only rely on a single sensor to judge the cable state, lack multi-modal data support, are easily affected by environmental interference, and have false positives and false negatives;

[0005] 3. Poor safety: high-altitude operation has high safety risks, especially when operating in bad weather or near high-voltage lines;

[0006] 4. Weak adaptability: existing devices are mostly fixed-point laid, and cannot flexibly adapt to complex cable layout environments, making it difficult to adapt to different types of cables and site conditions;

[0007] In recent years, with the rapid development of UAV technology, infrared thermal imaging, electromagnetic sensing, laser radar and other sensing technologies, new solutions have been brought to line loss detection. However, most UAV inspection systems still face the following challenges:

[0008] Lack of comprehensive perception ability of factors such as wind speed, wind direction and humidity in complex environments, cannot adaptively switch between flight detection and close-range attachment detection modes, and a perfect multi-modal data fusion analysis and fault level classification mechanism has not been established.

[0009] Therefore, there is an urgent need for an overhead cable line loss intelligent detection system that integrates multi-modal sensing technology, has flight and underhanging dual-mode operation capability, can intelligently analyze environmental disturbances and implement abnormal level classification, to improve the intelligence and safety of power system operation. SUMMARY

[0010] In view of the above-mentioned problems, the present application provides an overhead cable line loss intelligent detection system and method based on a UAV.

[0011] The purpose of the application can be realized by the following technical solutions: an overhead cable line loss intelligent detection system based on a UAV, comprising: a UAV device, a cable acquisition module, a line loss intelligent detection module, a data processing module, and a UAV control module;

[0012] The UAV device comprises a UAV body, and the UAV body is provided with a visible light integrated infrared thermal imaging camera, a laser radar, and an electromagnetic field sensor, a slide rail support, a slide rail support, a movable slide rail support, a drive pulley, a pulley, a pulley, and a flight fan blade; and line loss detection data and environmental data are obtained by data acquisition by the UAV device;

[0013] The data processing module acquires the line loss detection data and the environmental data for lightweight processing to obtain execution standard data, which are respectively denoted as standard line loss detection data and standard environmental data, and are sent to the line loss intelligent detection module for detection;

[0014] The line loss intelligent detection module receives the standard line loss detection data and the standard environmental data for analysis: the standard line loss detection data are analyzed to obtain image abnormal points, sound wave abnormal points, and magnetic sensing abnormal points; the standard environmental data are analyzed to obtain a correction factor, a flight detection mode instruction, and a hanging detection mode instruction;

[0015] Based on the obtained image abnormal points, sound wave abnormal points, and magnetic sensing abnormal points, line loss abnormal points are classified by grade to obtain line loss abnormal point grades;

[0016] The UAV control module executes the flight detection mode instruction and the hanging detection mode instruction according to the generated flight detection mode instruction and the hanging detection mode instruction.

[0017] As a preferred embodiment of the application, the UAV device intelligently monitors the line loss detection process by the UAV body, and the detection process includes a flight detection mode and a hanging detection mode;

[0018] In the flight detection mode, image data of the overhead cable is acquired by the visible light integrated infrared thermal imaging camera during flight, sound wave data of the overhead cable is obtained by the laser radar, and magnetic sensing data of the overhead cable is detected by the electromagnetic field sensor;

[0019] In the hanging detection mode, the target cable is approached by the flight mode, and the pulley and the pulley mounted on the UAV body are hung on the overhead cable; the movable slide rail support is driven by the drive pulley to move, so that the UAV body is stably connected with the cable through the slide rail support and the slide rail support; data acquisition is performed by the visible light integrated infrared thermal imaging camera, the laser radar, and the electromagnetic field sensor.

[0020] As a preferred embodiment of the present application, the process of analyzing the standard line loss detection data to obtain image abnormal points, sound wave abnormal points and magnetic induction abnormal points is as follows:

[0021] Analyzing the image data: preprocessing each frame of image collected to obtain processed image data; using image segmentation and edge detection algorithm to extract the outer contour boundary of the cable in the processed image data to obtain the projection area and edge contour feature of the cable in the image; then comparing the projection area and edge contour feature of one frame of image with the projection area and edge contour feature of the previous and next frames of image to obtain problem area image and problem edge contour image;

[0022] Analyzing the angle difference between each frame of image based on the area image and problem edge contour image to obtain image abnormal points;

[0023] Analyzing the sound wave data: removing image abnormal points based on the image data to obtain image normal frame sequence, extracting sound wave data sequence in the corresponding time period from the image normal frame sequence, analyzing the sound wave data sequence to obtain sound wave feature vector, using the sound wave feature vector in normal operation state to train standard template for comparison to obtain sound wave abnormal points;

[0024] Analyzing the magnetic induction data: identifying magnetic induction cutting points based on the magnetic induction data to obtain magnetic induction cutting points, constructing magnetic induction cutting surface from the magnetic induction cutting points, modeling the magnetic induction field around the magnetic induction cutting surface to obtain theoretical magnetic induction value, calculating the disturbance residual by difference between the actual collected magnetic induction value and the theoretical magnetic induction value, and if the module length of the disturbance residual is greater than the set magnetic disturbance threshold, judging as magnetic induction abnormal point.

[0025] As a preferred embodiment of the present application, the process of analyzing the angle difference between each frame of image is as follows:

[0026] Using line segment fitting algorithm to analyze and calculate the direction angle of the cable in the image; comparing the direction angles of the previous and next frames of image to obtain image angle offset value; based on the image angle offset value, correcting the image existing in the problem area image and problem edge contour image to obtain corrected angle image; comparing the corrected angle image with the original image, if the corrected area difference value is greater than the area threshold, performing image angle detection and recording as image abnormal point; if the corrected edge contour feature repetition rate is less than the edge threshold, performing image angle detection and recording as image abnormal point;

[0027] Then, color difference analysis is performed on each pre-processed image: the image obtained from the infrared thermal imaging channel contains temperature values corresponding to each pixel; the infrared thermal color difference is obtained by the temperature gradient change between pixels in the image area of each frame; if the infrared thermal color difference exceeds the set threshold, it is determined that there is a line loss hidden danger, which is recorded as an image abnormal point.

[0028] As a preferred embodiment of the present application, the process of analyzing based on the acoustic wave data sequence is as follows:

[0029] The acoustic wave original signal is band-pass filtered according to the acoustic wave data sequence to remove environmental noise and equipment mechanical interference, and a window function is used for framing to obtain each wave signal; each acoustic wave signal is subjected to short-time Fourier transform to obtain its time-frequency spectrum, and features are extracted from the time-frequency spectrum: including: main frequency f p , average energy spectrum E mean , spectrum centroid C s , spectrum spread S s , and spectrum skewness Sk s ; the extracted features are statistically formed into an acoustic wave feature vector: V i =[f p , E mean , C s , S s , Sk s ].

[0030] As a preferred embodiment of the present application, the process of using the standard template trained by the multi-segment acoustic wave feature vector in the normal running state for comparison is as follows:

[0031] The current acoustic wave feature vector sequence {V1, V2, …, V n} is dynamically time-warped (DTW) matched with the standard template T ref (t): Wherein, dist represents the Euclidean distance; if there is a local time period t k , the DTW cumulative distance gradient DTW(V, T ref ) is greater than the set threshold δ, or the spectral feature is lower than the set threshold ε i1 ; then the time point is determined as an acoustic wave abnormal point.

[0032] As a preferred embodiment of the present application, the process of analyzing the standard environment data to obtain the correction factor is as follows:

[0033] The environmental parameter data set under the historical normal flight condition is obtained, and a standard environment data model is established combined with the sensor output; the standard environment data model forms a reference distribution model in a multivariate regression manner, recorded as: E ref =f(W s , W d, T1, H), wherein: Ws is wind speed, Wd is wind direction, T1 is temperature, and H is humidity; the four types of environmental parameters are collected in real time in the current detection task to form an environmental vector E cur = (W s , W d , T1, H); the current environmental vector E cur is compared with the reference model E ref to obtain an environmental difference degree, and the environmental difference degree is defined as an environmental drift degree AE: The environmental drift degree AE is mapped to obtain a correction factor K adj .

[0034] As a preferred embodiment of the present application, the process of mapping the environmental drift degree AE is as follows:

[0035] When AE < ε1, it indicates that the environmental change is slight, and K adj = 1.0; when ε1≤ AE < ε2, it indicates that the environment is disturbed to a certain extent, and a linear function is used for weak correction: K adj = 1 + λ · (AE - ε1); wherein λ is an adjustment coefficient; when AE ≥ ε2, it indicates that the environmental disturbance is large, and K adj = 1.5; wherein ε1 and ε2 are both allowed disturbance threshold values.

[0036] As a preferred embodiment of the present application, the process of classifying the line loss abnormal points by levels is as follows:

[0037] The decision rule tree model is classified, and each aggregated abnormal point is constructed into a feature vector V fuse as an input node of the rule tree; the rule tree takes the feature vector of the fused abnormal point as input; whether the current abnormal cluster contains abnormal information from images, sound waves and magnetic sensing: if it is a three-mode abnormality, that is, images, sound waves and magnetic sensing simultaneously appear abnormal, it is determined that it has a highly credible line loss abnormal signal, and enters a high-confidence path;

[0038] If the number of modes is 1 or 2, it enters a low-mode path;

[0039] For the high-confidence path, the abnormal intensity values of the three modes are analyzed respectively:

[0040] When the abnormal intensity of the three modes all exceeds a set high threshold value, it is determined to be a level I line loss point, and if the intensity of part of the modes is lower than a set low threshold value, it is determined to be a level II line loss point;

[0041] For the low-mode path, the abnormal intensity values of the three modes are analyzed respectively:

[0042] When the anomaly only appears in part of the modal, further classification is carried out according to the maximum anomaly intensity and spatio-temporal consistency: if the anomaly intensity value of any modal exceeds the high threshold value, it is determined as a grade II line loss point; if the anomaly intensity value belongs to the moderate interval, and the anomaly points among the multi-modals differ by no more than 0.3 seconds in time and no more than 0.5 meters in spatial distance, it is a grade III line loss early warning point.

[0043] Another aspect of the present application provides an unmanned aerial vehicle-based overhead cable line loss intelligent detection method, comprising:

[0044] S1: collect environmental parameters after starting detection, calculate environmental drift degree and generate a correction factor, combine wind speed to determine to enter flight detection or under-hanging detection mode, and generate a self-stabilizing coefficient to improve attitude control stability;

[0045] S2: collect data in the corresponding mode: obtain visible light, infrared, laser radar and electromagnetic data in flight mode; collect close-range images, infrared, sound waves and electromagnetic information in under-hanging mode, and continuously monitor environmental parameters;

[0046] S3: standardize the collected data, enhance the image, extract the sound wave spectrum, and model the magnetic induction, to form standard detection data;

[0047] S4: analyze the standard data to identify abnormal points: image detection area difference and contour change, infrared identification of hot spots, sound wave matching spectrum deviation, and magnetic induction analysis of disturbance residual error;

[0048] S5: fuse the multi-modal abnormal points and output unified line loss identification results.

[0049] Compared with the prior art, the present application has the following advantages:

[0050] 1. The present application uses an unmanned aerial vehicle to combine flight detection and under-hanging detection modes, automatically selects the most suitable detection mode according to wind speed and environmental changes, maintains detection stability when the wind is strong or the environment is complex, can not only quickly patrol a wide range of cables, but also can closely and accurately detect key areas, has better adaptability and better detection effect.

[0051] 2. The present application uses images, infrared, sound waves and magnetic induction and other sensors to collect information, and uses intelligent analysis methods to judge whether there is a line loss problem, can more accurately find abnormal points, reduce missed detection and misjudgment, improves the overall detection efficiency and accuracy, and has higher practical value. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to facilitate understanding by those skilled in the art, the present application will be further described below with reference to the accompanying drawings.

[0053] Figure 1 The present application is a schematic diagram of the principle;

[0054] Figure 2 For the method steps of the present application;

[0055] Figure 3 For the schematic of the unmanned aerial vehicle of the present application Figure 1 ;

[0056] Figure 4 For the schematic of the unmanned aerial vehicle of the present application Figure 2 ;

[0057] Figure 5 For the schematic of the unmanned aerial vehicle of the present application Figure 3 ;

[0058] BRIEF DESCRIPTION OF DRAWINGS: 1, cable, 2, slide rail support one, 3, slide rail support two, 4, movable slide rail support, 5, drive pulley, 6, pulley one, 7, pulley two, 8, visible light integrated infrared thermal imaging camera, 9, flight fan, 10, unmanned aerial vehicle body. DETAILED DESCRIPTION

[0059] The technical solutions of the present application will be described clearly and completely below in conjunction with embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0060] It should be understood that the terms "comprise" and "include" used in the specification and claims of the present disclosure indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0061] It should also be understood that the terms used in the specification of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure. As used in the specification and claims of the present disclosure, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should be further understood that the term "and / or" used in the specification and claims of the present disclosure means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0062] Referring to Figure 1 , the present application provides an unmanned aerial vehicle-based overhead cable line loss intelligent detection system, which comprises an unmanned aerial vehicle device, a cable acquisition module, a line loss intelligent detection module, a data processing module, and an unmanned aerial vehicle control module.

[0063] Referring to the accompanying Figure 3 ,4 , 5 shown in FIG. 1, the embodiment provides a UAV device, which comprises a UAV body 10, and the UAV body 10 is provided with various functional components for data acquisition and detection, including a visible light integrated infrared thermal imaging camera 8, a laser radar and an electromagnetic field sensor, and is further provided with structural components for mounting and guiding, including a slide rail bracket one 2, a slide rail bracket two 3, a movable slide rail bracket 4, a drive pulley 5, a pulley one 6, a pulley two 7, and a plurality of flight fans 9 for flight propulsion. In the process of line loss detection, intelligent monitoring is performed by the UAV body 10, and the detection process further comprises a flight detection mode and a hanging detection mode.

[0064] In the embodiment, the UAV control module executes flight detection mode instructions and hanging detection mode instructions.

[0065] In the flight detection mode, the UAV body 10 generates lift by the flight fans 9 to realize air flight, and in the flight process, cable image data is acquired by the visible light integrated infrared thermal imaging camera 8, cable spatial position and shape data are collected by the laser radar to obtain sound wave data, and magnetic induction data is detected by the electromagnetic field sensor to obtain electromagnetic waves around the cable, so as to realize multi-dimensional perception and line loss detection of the overhead cable 1 in the air; the collected cable image data, sound wave data and magnetic induction data are collectively referred to as line loss detection data.

[0066] In the hanging detection mode, the UAV device first approaches the target cable 1 by flight, and is mounted on the cable 1 by the pulley one 6 and the pulley two 7 provided on the UAV body 10; the drive pulley 5 drives the movable slide rail bracket 4 to act, so that the UAV body 10 is stably connected and locked with the cable by the slide rail bracket two 3, and is thus hung on the cable 1 in a stable attached state; in this mounting state, the camera 8, the laser radar and the electromagnetic field sensor still work cooperatively to collect cable image data, sound wave data and magnetic induction data at a close distance and with high precision, which are collectively referred to as line loss detection data.

[0067] The UAV device is also used for collecting environmental data, which includes but is not limited to wind speed, wind direction, humidity and temperature; the humidity and temperature are collected by a temperature and humidity sensor carried by the UAV device.

[0068] In the example, the specific process of collecting wind speed and wind direction is as follows:

[0069] When the unmanned aerial vehicle is in flight, the flight fan blades 9 are continuously operated to provide lift and attitude control; the flight control unit is also provided in the unmanned aerial vehicle device, the flight control unit obtains the speed change information of the flight fan blades 9 in real time through the built-in fan blade motor speed sensor, analyzes the attitude data and flight speed parameters of the unmanned aerial vehicle body 10; when external wind acts on the unmanned aerial vehicle, the dynamic change of the speed of the flight fan blades 9 is caused; by comparing the speed difference, power compensation amplitude and attitude adjustment behavior of each flight fan blade 9 at the same time point, the direction and strength of the output wind disturbance are calculated, and the wind speed and wind direction in the external environment are calculated.

[0070] The attitude data includes the pitch angle, roll angle and heading angle.

[0071] It should be noted that: the flight control unit can also combine the spatial attitude information provided by the inertial measurement unit (IMU) to process the fan disturbance data through the wind disturbance model and the dynamic response compensation algorithm to realize intelligent estimation of the real-time wind speed value and the wind direction angle. This process does not require additional independent anemometer or wind direction sensor, but relies on the natural response behavior of the flight fan blades 9 and the data fusion of the flight control system.

[0072] The specific process of processing the fan disturbance data through the wind disturbance model and the dynamic response compensation algorithm to realize intelligent estimation of the real-time wind speed value and the wind direction angle is as follows:

[0073] The flight control unit obtains the real-time flight parameters of the unmanned aerial vehicle body 10 in the flight process, including the speed data of each flight fan blade 9 output by the fan blade motor speed sensor, the attitude data output by the inertial measurement unit, and the flight speed parameter.

[0074] The real-time flight parameters are input into the preset wind disturbance model, which is used to establish the relationship between the external wind and the attitude and dynamic response of the unmanned aerial vehicle body 10; by comparing the speed difference, output power deviation and attitude correction amplitude of each flight fan blade 9 at the same time point, the dynamic response characteristic parameters under the action of external disturbance are obtained; then, the dynamic response compensation algorithm is called to fuse and calculate the dynamic response characteristic parameters and the attitude data provided by the IMU, to extract the effective disturbance component caused by the external wind by compensating the self-vibration, flight noise and inertial drift error of the unmanned aerial vehicle, and to decompose to obtain the component data in the horizontal and vertical directions.

[0075] Based on the component data in the horizontal and vertical directions, the wind speed and wind direction angle are output by the wind disturbance model according to the mechanical mapping relationship, and the results are transmitted to the data processing module in real time.

[0076] The cable acquisition module acquires line loss detection data and environmental data collected by the unmanned aerial vehicle device and sends them to the data processing module.

[0077] The data processing module acquires the collected line loss detection data and environmental data for processing to obtain unified execution standard data, which are respectively denoted as standard line loss detection data and standard environmental data, and sends them to the line loss intelligent detection module for detection.

[0078] After the line loss intelligent detection module receives the standard line loss detection data and the standard environmental data, it analyzes them, specifically as follows.

[0079] The standard line loss detection data are analyzed as follows.

[0080] The image data are analyzed as follows: the collected each frame of image is preprocessed to obtain processed image data; the image segmentation and edge detection algorithm is used to extract the outer contour boundary of the cable in the processed image data to obtain the projection area and edge contour feature of the cable in the image.

[0081] The projection area and edge contour feature of one frame of image obtained are compared with the projection area and edge contour feature of the image of each frame before and after it to obtain the projection area difference value and the edge contour feature repetition rate, which are compared with the pre-set area difference value and edge contour feature repetition rate operation threshold value; the area difference value operation threshold value is denoted as the area operation threshold value; the edge contour feature repetition rate operation threshold value is denoted as the edge operation threshold value; if the area difference value is greater than the area operation threshold value, image angle detection is performed, which is denoted as the problem area image; if the edge contour feature repetition rate is less than the edge operation threshold value, image angle detection is performed, which is denoted as the problem edge image.

[0082] It should be noted that the projection area is represented by the number of pixels; the preprocessing process includes image denoising, histogram equalization, illumination standardization and other operations.

[0083] The angle difference between each frame of image of the image angle is analyzed as follows: the line segment fitting algorithm is used to analyze and calculate the direction angle of the cable in the image; the direction angles of the image of each frame before and after are compared to obtain the image angle offset value; the image existing in the problem area image and the problem edge image is corrected based on the image angle offset value to obtain a corrected angle image; the corrected angle image is compared with the original image; if the corrected area difference value is greater than the area operation threshold value, it is denoted as an image abnormal point; if the corrected edge contour feature repetition rate is less than the edge operation threshold value, it is denoted as an image abnormal point.

[0084] Then color difference analysis is performed on each pre-processed image: the image obtained from the infrared thermal imaging channel contains temperature values corresponding to each pixel; the temperature gradient change between pixels in the image area of each frame before and after is used to determine whether there are hot spots, uneven temperature differences, abnormal heating and other phenomena in the area through the infrared thermal color difference ΔT; the thermal color difference calculation method is: ΔT=T max -T min ; wherein T max and T min represent the highest and lowest temperature values in the cable area, and if the thermal color difference ΔT exceeds the set threshold, it is determined that there is a line loss hidden danger, which is recorded as an image abnormal point;

[0085] The sound wave data is analyzed: the image abnormal points are removed from the image data to obtain image normal data; the image normal frames sequence is obtained according to the collection frames corresponding to the image normal data, and the sound wave data sequence in the corresponding time period is extracted from the image normal frames sequence, recorded as S={s1, s2,..., s n};

[0086] The sound wave original signal is band-pass filtered according to the sound wave data sequence, and the environmental noise and equipment mechanical interference are removed, and each wave signal S i (t) is obtained by using a window function for framing; each sound wave signal S i (t) is transformed by short-time Fourier transform (STFT) to obtain the corresponding time-frequency domain spectrum, and then features are extracted from the time-frequency domain spectrum, including the main frequency f p , the average energy spectrum E mean , the spectrum centroid C s , the spectrum spread S s and the spectrum skewness Sk s ; the extracted features are statistically formed into a sound wave feature vector, namely: V i =[f p , E mean , C s , S s , Sk s ];

[0087] Then the standard template T ref (t) is trained using multiple sound wave feature vectors in the normal operating state;

[0088] The current sound wave feature vector sequence {V1, V2,..., V n} is matched with the standard template T ref (t) by dynamic time warping (DTW): wherein dist represents the Euclidean distance; V j is the feature vector corresponding to the standard template;

[0089] If the local time period tk , DTW(V, T ref ) > δ, or DTW(V, T ref ) < ε i1 , the time point is determined as an acoustic anomaly point; and the acoustic anomaly point is marked on the time axis.

[0090] It should be noted that: ε i1 ∈ [f p ’, E mean ’, C s ’, S s ’, Sk s ’], wherein f p ’, E mean ’, C s ’, S s ’ and Sk s ’ are threshold values of the main frequency, the average energy spectrum, the spectrum centroid, the spectrum spread and the spectrum skewness respectively;

[0091] Analyzing the magnetic induction data: the electromagnetic field sensor carried by the unmanned aerial vehicle device collects spatial magnetic field data at a constant frequency during the inspection process, including: the magnetic induction intensity vector The magnetic induction direction θ B : the angle between the magnetic force line and the cable axis, the flight path direction The magnetic induction cutting direction: refers to The vector angle θ c between and ;

[0092] Based on the magnetic induction cutting direction, the magnetic induction cutting point is identified; the vector angle between the magnetic induction direction and the flight direction is calculated in real time When the magnetic induction direction and the flight direction θ c > angle threshold θ th , it is defined as a magnetic induction cutting point: P cut = {(x i1 , y i1 , z i1 , t i1 ) | θ c (x i1 , y i1 , z i1 , t i1 ) > θ th}; a plurality of magnetic induction cutting points form a magnetic induction cutting surface, that is, a spatial section of the magnetic flux change: S cut = SurfaceFit(P cut );

[0093] According to the obtained magnetic induction cutting surface as a reference base surface, the magnetic induction field around the magnetic induction cutting surface is modeled: let the space-time coordinates of any sampling point be The magnetic induction value is Then the theoretical magnetic induction value of the point is analyzed by multivariate kernel regression Expression: Wherein: The space-time coordinates of the historical sampling point are represented; The magnetic induction vector measured at the sampling point is represented; The kernel function is represented, and the Gaussian kernel function is used: h is the kernel function bandwidth parameter, which controls the smoothness of local fitting;

[0094] The actual collected magnetic induction value is compared with the theoretical magnetic induction value to perform difference analysis, and the disturbance residual is calculated: If the length of the disturbance residual is greater than the set magnetic disturbance threshold, that is, Then it is judged as a magnetic induction anomaly point; wherein σ b is the set magnetic disturbance threshold;

[0095] The specific process of analyzing the standard environment data is as follows:

[0096] The environmental parameter data set under the historical normal flight condition is obtained, and the standard environment data model is established combined with the sensor output; the standard environment data model forms a reference distribution model in a multivariate regression manner, denoted as: E ref =f(W s s, W d d, T1, H), wherein: Ws is the wind speed, Wd is the wind direction, T1 is the temperature, and H is the humidity; in the current detection task, the above four types of environmental parameters are collected in real time to form the current environmental vector: E cur =(W s ’s, W d ’d, T1’, H’); then the current environmental vector E cur is compared with the reference distribution model E ref to obtain the environmental difference degree, and the environmental difference degree is defined as the environmental drift degree ΔE, that is:

[0097] The correction factor K adj is obtained based on the environmental drift degree ΔE, and is specifically:

[0098] When ΔE<ε1, it indicates that the environmental change is slight, K adj =1.0, indicating that no correction is needed;

[0099] When ε1≤ΔE<ε2, it indicates that the environment is disturbed to a certain extent, and a linear function is used for weak correction: K adj=1+λ·(ΔE-ε1);where λ is an adjustment coefficient, which is set by the technical staff through subjective assignment method;

[0100] When ΔE≥ε2, it means that the environmental disturbance is large, K adj =1.5, to ensure the stability of the identification system;

[0101] wherein ε1 and ε2 are both allowed disturbance thresholds, and ε1 is greater than ε2;

[0102] Finally, the correction factor K adj is combined with the magnetic disturbance threshold to obtain the corrected magnetic disturbance threshold σ b ’; the magnetic sensing anomaly point is re-determined, and if the module length of the disturbance residual is greater than the corrected magnetic disturbance threshold, that is, then it is judged as a magnetic sensing anomaly point.

[0103] According to the correction factor K adj and the real-time wind speed and direction, flight detection mode instructions and underhang detection mode instructions are generated;

[0104] The process of generating flight detection mode instructions is: if the environment is stable, that is, ΔE<ε, where ε is the flight allowed disturbance threshold, then enter the flight detection mode;

[0105] The process of generating underhang detection mode instructions is: if the wind speed W s >4.0 m / s, switch to underhang mode; then generate a self-stabilizing coefficient according to the wind speed and direction: where χ is the modulation parameter and φ is the cable direction angle.

[0106] Based on the obtained image anomaly points, acoustic anomaly points and magnetic sensing anomaly points, the line loss anomaly points are classified by grade, and the specific process is:

[0107] The image anomaly points, acoustic anomaly points and magnetic sensing anomaly points are collectively referred to as aggregated anomaly points; classification is performed through a decision rule tree model; each aggregated anomaly point is constructed as a feature vector V fuse as a rule tree input node: wherein n modal is the number of triggered anomaly modes, δ t is the maximum time difference between anomaly points of each mode, and δ s is the maximum spatial distance between anomaly points of each mode;

[0108] The rule tree takes the feature vector of the fused anomaly points as input, and makes hierarchical judgments based on the number of anomaly modes, the intensity of each type of sensing, and the time and spatial consistency between anomalies. The decision process is as follows:

[0109] Whether the current anomaly cluster contains anomaly information from image, acoustic and magnetic sensing three modalities:

[0110] If it is a three-mode abnormality, that is, the image, sound wave and magnetic field are abnormal at the same time, and the three modes are covered, it is determined that the line loss abnormal signal has high credibility, and enters the high credibility path;

[0111] If all the three modes are normal, no processing is needed;

[0112] Other conditions, that is, if the number of modes is 1 or 2, enter the low mode path, and further distinguish the abnormal strength and consistency.

[0113] For the high credibility path, the abnormal strength values of the three modes are analyzed respectively:

[0114] When the abnormal strength of the three is more than the set high threshold, it is determined that the line loss point is level I, that is, the fault position is serious and has high consistency;

[0115] When the abnormal strength of the three is not more than the set high threshold, it is normal;

[0116] Other conditions, it is determined that the line loss point is level II, indicating that it has moderate risk but needs to be continuously monitored.

[0117] For the low mode path, the abnormal strength values of the three modes are analyzed respectively:

[0118] When the abnormality only appears in part of the mode, according to the maximum abnormal strength and the time and space consistency, further classification is carried out:

[0119] If the abnormal strength value of any mode is more than the high threshold, it is determined that the line loss point is level II;

[0120] If the abnormal strength value belongs to the moderate interval, and the abnormal points among the multiple modes differ by not more than 0.3 seconds in time and not more than 0.5 meters in space distance, it is determined that the line loss early warning point is level III; otherwise, it is determined that the abnormal point is invalid, and does not enter the final report.

[0121] The level of each output line loss abnormal point is output in an electronic manner, and the abnormal point report has a multi-source detection report of positioning coordinates, risk level and sensing type.

[0122] Please refer to Figure 2 The present application provides a kind of overhead cable line loss intelligent detection method based on unmanned aerial vehicle, comprising:

[0123] S1: After the UAV initiates detection, it collects environmental parameters such as wind speed, wind direction, temperature, and humidity in the current detection environment to form the current environment vector. This vector is then compared with a standard environment model constructed under historical normal flight conditions to calculate the environmental drift. An environmental correction factor is generated based on the environmental drift, and the detection mode is determined in conjunction with the current wind speed value: if the environmental change is slight and the environmental drift is below the flight disturbance threshold, the UAV is controlled to enter the flight detection mode; if the wind speed exceeds the set threshold, it switches to the under-mounted detection mode. Simultaneously, a self-stabilizing coefficient is generated based on the wind speed and cable direction angle to improve the attitude control stability in the under-mounted mode.

[0124] S2: Based on the above mode determination results, the UAV enters the corresponding detection mode and carries out data collection tasks;

[0125] In flight detection mode, the UAV maintains stable flight in the air by using flight blades, uses integrated visible light and infrared thermal imaging cameras to acquire cable surface temperature image information, scans the spatial position and shape of the cable by LiDAR, and uses electromagnetic field sensors to capture magnetic field change data around the cable.

[0126] In the under-mounted detection mode, the drone connects to the cable via a rail and pulley assembly, allowing it to hang securely on the cable. It then collects images, infrared, acoustic, and electromagnetic information in close proximity to the cable, enabling high-precision localized line loss detection. Simultaneously, the flying fan blades and environmental sensors acquire real-time data on wind speed, direction, and humidity.

[0127] S3: The raw detection data and environmental data collected by the UAV are transmitted to the data processing module for formatting and standardization, generating standard line loss detection data and standard environmental data. Image data undergoes preprocessing operations such as denoising, histogram equalization, and illumination correction to improve image contrast and structural clarity. Acoustic signals are bandpass filtered to eliminate background noise, converted into a time-spectrum graph through short-time Fourier transform, and features such as dominant frequency, energy spectrum, and spectral centroid are extracted. Magnetic flux cross-sections are constructed based on magnetic flux cutting points, and kernel regression modeling is performed on the magnetic field space to calculate theoretical magnetic flux density. After the above processing is completed, the standardized data is sent to the intelligent line loss detection module for further analysis.

[0128] S4: The intelligent line loss detection module analyzes standardized data and identifies anomalies; the image analysis module judges abnormal areas in the image by the difference in cable projection area and edge contour repetition rate between consecutive frames, and corrects the image angle offset; the infrared thermal imaging analysis module identifies hot spots and abnormal heating points by calculating the pixel temperature difference within the area; the acoustic wave analysis module extracts features from the sound signal and performs dynamic time warping matching with the preset template to identify spectral deviations or time-domain abnormal segments; the magnetic induction analysis module identifies magnetic induction anomalies by comparing the disturbance residual between the actual magnetic induction value and the theoretical magnetic induction value, combined with the correction of the magnetic disturbance threshold; all anomalies are marked in the temporal and spatial dimensions.

[0129] S5: Perform temporal and spatial fusion matching on image anomalies, acoustic anomalies, and magnetic anomalies, and combine the multimodal perception results to form a unified line loss identification result.

[0130] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An unmanned aerial vehicle (UAV)-based overhead cable line loss intelligent detection system, comprising: The unmanned aerial vehicle device, the cable acquisition module, the line loss intelligent detection module, the data processing module and the unmanned aerial vehicle control module; characterized in that: the unmanned aerial vehicle device comprises an unmanned aerial vehicle body (10), the unmanned aerial vehicle body (10) is provided with a visible light integrated infrared thermal imaging camera (8), a laser radar and an electromagnetic field sensor, the unmanned aerial vehicle body (10) is further provided with a sliding rail support one (2), a sliding rail support two (3), a movable sliding rail support (4), a driving pulley (5), a pulley one (6), a pulley two (7) and flight fan blades (9); line loss detection data and environmental data are obtained through data acquisition of the unmanned aerial vehicle device; The data processing module processes the line loss detection data and the environmental data to obtain standard line loss detection data and standard environmental data, and sends them to the line loss intelligent detection module for detection; The line loss intelligent detection module analyzes the standard line loss detection data and the standard environmental data after receiving them, including: analyzing the standard line loss detection data to obtain image abnormal points, sound wave abnormal points and magnetic sensing abnormal points; analyzing the standard environmental data to obtain correction factors, flight detection mode instructions and under-hanging detection mode instructions; based on the image abnormal points, the sound wave abnormal points and the magnetic sensing abnormal points, classifying the line loss abnormal points to obtain line loss abnormal point levels; The unmanned aerial vehicle control module is used to execute the flight detection mode instructions and the under-hanging detection mode instructions. 2.The unmanned aerial vehicle (UAV) -based overhead cable line loss intelligent detection system of claim 1, wherein, The state of the unmanned aerial vehicle body (10) during line loss detection includes a flight detection mode and an under-hanging detection mode; In the flight detection mode: during flight, the visible light integrated infrared thermal imaging camera (8) acquires image data of the overhead cable (1), the laser radar acquires sound wave data of the overhead cable (1), and the electromagnetic field sensor acquires magnetic sensing data of the overhead cable (1); In the under-hanging detection mode: the target cable is approached by flight, and the pulley one (6) and the pulley two (7) provided on the unmanned aerial vehicle body (10) are mounted on the overhead cable (1); the driving pulley (5) drives the movable sliding rail support (4) to move, so that the unmanned aerial vehicle body (10) is stably connected to the cable through the sliding rail support one (2) and the sliding rail support two (3); the visible light integrated infrared thermal imaging camera (8), the laser radar and the electromagnetic field sensor acquire data. 3.The unmanned aerial vehicle (UAV) -based overhead cable line loss intelligent detection system of claim 1, wherein, The process of analyzing the standard line loss detection data to obtain image abnormal points, sound wave abnormal points and magnetic sensing abnormal points is as follows: Image data analysis: each frame of image data collected is preprocessed to obtain processed image data; the image segmentation and edge detection algorithm is used to extract the outer contour boundary of the cable in the processed image data, to obtain the projection area and edge contour features of the cable in the image; the projection area and edge contour features of one frame of image data are compared with those of the previous and subsequent frames of image data, to obtain a problem area image and a problem edge contour image; Based on the problem area image and the problem edge contour image, the angle difference between each frame of image data is analyzed to obtain image abnormal points; The process of analyzing the angle difference between each frame of image is as follows: The direction angle of the cable in the image is obtained by using a line segment fitting algorithm; the direction angles of the images of the previous and subsequent frames are compared to obtain an image angle offset value; the image angle offset value is used to correct the images existing in the problem area image and the problem edge contour image, to obtain a corrected angle image; the corrected angle image is compared with the original image, if the difference value of the corrected area is greater than the area threshold, image angle detection is performed, and the image abnormal point is recorded; if the repeated rate of the corrected edge contour feature is less than the edge threshold, image angle detection is performed, and the image abnormal point is recorded; 4. The unmanned aerial vehicle based overhead cable line loss intelligent detection system according to claim 3, characterized in that, The pre-processed image of each frame is then analyzed for color difference: the image obtained through the infrared thermal imaging channel contains temperature values corresponding to each pixel; The infrared thermal color difference is obtained by the temperature gradient change between the pixels in the image area of each frame; if the infrared thermal color difference exceeds the set threshold, it is determined that there is a line loss hidden danger, and the image abnormal point is recorded. The process of analyzing the sound wave data sequence is as follows: The sound wave original signal is band-pass filtered according to the sound wave data sequence, and the environmental noise and equipment mechanical interference are removed; a window function is used for framing to obtain each sound wave signal; the short-time Fourier transform is used on each sound wave signal to obtain its time-frequency domain spectrum; the features are extracted from the time-frequency domain spectrum, including the main frequency, the average energy spectrum, the spectrum centroid, the spectrum dispersion degree and the spectrum skewness, and the sound wave feature vector is formed by statistics.

5. The unmanned aerial vehicle based overhead cable line loss intelligent detection system according to claim 3, characterized in that, The process of using the multi-segment sound wave feature vector in the normal running state to train the standard template for comparison is as follows: The process of analyzing the standard environment data to obtain the correction factor is as follows:

6. The unmanned aerial vehicle based overhead cable line loss intelligent detection system according to claim 5, characterized in that, The environmental parameter data set under the historical normal flight condition is obtained, and a standard environment data model is established combined with the sensor output; the standard environment data model forms a reference distribution model in a multivariate regression manner; four types of environmental parameters are collected in real time in the current detection task to form an environmental vector at the current time; the current environmental vector is compared with the reference model to obtain the environmental difference degree, which is defined as the environmental drift degree; the environmental drift degree is mapped to obtain the correction factor. The current acoustic wave feature vector sequence is dynamically time warping matched with the standard template to obtain a DTW cumulative distance gradient, if there is a local time period with the DTW cumulative distance gradient greater than a set threshold δ, or the DTW cumulative distance gradient is lower than a set threshold ε i1， Then the time point is determined as an acoustic wave abnormal point. 7.The unmanned aerial vehicle (UAV) -based overhead cable line loss intelligent detection system of claim 1, wherein, The process of classifying the line loss abnormal points by grade is as follows: ​ 8.The unmanned aerial vehicle (UAV) -based overhead cable line loss intelligent detection system of claim 7, wherein, ​ The decision rule tree model is classified, each aggregated anomaly point is constructed into a feature vector as a rule tree input node, the rule tree is inputted with the feature vector of the fused anomaly point, whether the current anomaly cluster contains anomaly information from three modalities of images, sound waves and magnetic sensing: if it is a three-modality anomaly, i.e., images, sound waves and magnetic sensing simultaneously appear abnormal, it is determined to have a highly reliable line loss anomaly signal and enters a high-confidence path; if the number of abnormal modalities is one or two, it enters a low-modality path: if the abnormal intensity of any one or two exceeds a set high threshold, it is determined to be a level II line loss point, if the abnormal intensity value belongs to a moderate interval, and the anomaly points among the multiple modalities differ by no more than 0.3 seconds in time and no more than 0.5 meters in spatial distance, it is determined to be a level III line loss early warning point; For the high-confidence path, the abnormal intensity values of the three modalities are analyzed respectively: when the abnormal intensity values of the three modalities all exceed a set high threshold, it is determined to be a level I line loss point; When the abnormal intensity values of the three modalities all do not exceed a set high threshold, it is normal; otherwise, it is determined to be a level II line loss point, indicating moderate risk but requiring continuous monitoring.

9. An unmanned aerial vehicle-based overhead cable line loss intelligent detection method, characterized in that The unmanned aerial vehicle-based overhead cable line loss intelligent detection system according to any one of claims 1-8 comprises: S1: after starting detection, environmental parameters are collected, environmental drift is calculated and a correction factor is generated, a flight detection or a hanging detection mode is entered in combination with wind speed, and a self-stabilization coefficient is generated to improve attitude control stability; S2: data is collected in the corresponding mode: visible light, infrared, laser radar and electromagnetic data are obtained in the flight mode; short-distance images, infrared, sound waves and electromagnetic information are collected in the hanging mode, and environmental parameters are continuously monitored; S3: collected data is standardized, image enhancement, sound wave spectrum extraction and magnetic sensing modeling are performed, and standard detection data are formed; S4: standard data are analyzed to identify anomaly points: image detection area difference and contour change, infrared identification of hot spots, sound wave matching of spectral deviation and magnetic sensing analysis of disturbance residual; S5: multiple-modality anomaly points are fused, and a unified line loss identification result is outputted.

Images