A hanging object identification and early warning method, system, device and medium

Abstract

The application discloses a kind of hanging object identification early warning method, system, equipment and medium, method includes.The application identifies target hanging object and tracks, obtains the pixel coordinate information of target hanging object;Extract the non-rigid motion deformation feature of target hanging object, and construct deformation feature vector according to non-rigid motion deformation feature;Based on pixel coordinate information, extract the image feature of target hanging object, collect substation environmental sensor data, obtain the material environment parameter of target hanging object;Degree of threat is calculated, and early warning signal is output based on degree of threat.The application calculates the space variation characteristics of the optical flow field of pixel points in the target area, and through the classification and tracking of topological singular points, the non-rigid deformation feature of the target is quantified, which can distinguish the rigid translation of hanging object from dangerous non-rigid distortion and stretching;Combined with multi-source sensing data and material correction coefficient, generate dynamic safety discharge gap, and get rid of the limitation of fixed safety distance.

Description

Technical Field

[0001] This invention relates to the field of hanging object identification technology, and in particular to a hanging object identification and early warning method, system, device and medium. Background Technology

[0002] Computer vision and deep learning technologies are increasingly being applied in the field of foreign object monitoring for power equipment. Currently, systems for identifying hanging objects based on deep learning object detection algorithms have achieved static localization and classification of these objects, and multi-dimensional condition monitoring and risk assessment are becoming the mainstream development trend in this field. By analyzing substation monitoring videos frame by frame, the system can identify whether there are hanging objects such as kites or plastic bags in the work area and output their category and location bounding boxes.

[0003] However, existing technologies primarily rely on traditional convolutional neural networks for static target detection, with outputs focused on the bounding box and static category of the hanging object, failing to effectively perceive the internal non-rigid motion characteristics of the hanging object. For example, existing systems struggle to distinguish between a dry plastic bag merely shifting in the wind and a damp metal film undergoing violent twisting and stretching, resulting in a lack of assessment of the discharge risk caused by deformation. Furthermore, existing risk warning mechanisms typically rely on preset fixed safety distance thresholds, failing to fully consider the decisive influence of the hanging object's material and the surrounding environment on the discharge voltage and air breakdown gap.

[0004] Meanwhile, early warning and judgment mechanisms are often based solely on a snapshot of the current location, making it difficult to distinguish between rapidly approaching sudden threats and slowly approaching potential threats, leading to inaccurate assessments of the urgency of events.

[0005] Therefore, there is an urgent need for a substation floating object identification and early warning method that can comprehensively analyze the non-rigid deformation characteristics of the floating object, dynamically assess the safe discharge gap under the influence of materials and environment, and judge the trend of threat evolution to meet the needs of the field. Summary of the Invention

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a method, system, device and medium for identifying and warning of floating objects to solve the problems in the prior art that cannot effectively perceive the internal non-rigid motion characteristics of floating objects, and that fixed safety distance thresholds cannot fully consider the decisive influence of the material of the floating object itself and the on-site environment on the discharge voltage and air breakdown gap.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for identifying and issuing early warnings of hanging objects, comprising: Identify and track target hanging objects to obtain their pixel coordinate information; Extract the non-rigid motion deformation features of the target hanging object, and construct a deformation feature vector based on the non-rigid motion deformation features; Based on pixel coordinate information, multi-source sensing data values ​​of the target hanging object are extracted, and environmental sensor data of the substation are collected. In response to the acquisition of multi-source sensing data values ​​and environmental sensor data, material environmental parameters of the target hanging object are obtained. Extract the air breakdown discharge gap under the current environment and construct a feature library to characterize the mapping relationship between material type, environmental correction coefficient and air breakdown discharge gap; Based on the material type and environmental correction factor, the feature library is queried to determine the dynamic safe discharge gap under the current environment; Based on pixel coordinate information and deformation feature vector, predict the motion trajectory of the target hanging object, obtain the spatial distance between the target hanging object and the equipment based on the motion trajectory, and determine the bridging trend of the target hanging object through the motion trajectory. By combining spatial distance, dynamic safe discharge gap, and bridging trend, the threat level is calculated, and an early warning signal is output based on the threat level.

[0009] As a preferred embodiment of the floating object identification and early warning method of the present invention, the steps of identifying and tracking the target floating object and obtaining the pixel coordinate information of the target floating object include: Extract the moving foreground from the surveillance video and establish a device mask based on the sensor coordinates within the area; In response to the acquisition of the device mask, the moving foreground is filtered to obtain candidate regions; A pre-defined shape library is used to extract the contour geometric features of candidate regions. The contour geometric features are then matched with the parameters in the shape library to obtain the target geometric matching degree. A preset matching baseline value is used to filter areas where the geometric matching degree of the target exceeds the matching baseline value, and these areas are identified as target hanging objects. Extract the spatial position parameters of the target floating object and establish pixel coordinate information.

[0010] As a preferred embodiment of the hanging object identification and early warning method of the present invention, the step of extracting the non-rigid motion deformation features of the target hanging object and constructing a deformation feature vector based on the non-rigid motion deformation features includes: Lock the target region corresponding to the pixel coordinate information, calculate the dense optical flow field vector of the pixels in the target region, and characterize the non-rigid motion state of the target hanging object; In response to the acquisition of dense optical flow field vectors, the spatial variation characteristics of the optical flow field of the corresponding pixel are calculated as the non-rigid motion deformation characteristics of the target hanging object; A zero-value deviation reference is set, and the spatial variation characteristics of the optical flow field are compared with the zero-value deviation reference to obtain the comparison results. Based on the comparison results, topological singularities in the dense optical flow field vectors are identified and classified. These topological singularities include source points, sink points, and saddle points. By statistically analyzing the spatial distribution and temporal changes of topological singularities, deformation feature vectors are constructed.

[0011] As a preferred embodiment of the hanging object identification and early warning method of the present invention, the step of obtaining the material environmental parameters of the target hanging object includes: Collect environmental sensor data; locate the image region corresponding to the pixel coordinate information, and calculate the average brightness value and matrix contrast value of the image region; The environmental sensor data, average brightness value, and matrix contrast value are combined to generate multi-source sensing data values. Multiple material feature ranges are preset, and based on multi-source sensing data values, average brightness values ​​and matrix contrast values ​​are extracted and compared with the preset multiple material feature ranges to classify the material category of the target hanging object. A preset environmental correction parameter library is established. Multi-source sensing data values ​​are called into the environmental correction parameter library for retrieval to determine the environmental correction coefficient.

[0012] As a preferred embodiment of the floating object identification and early warning method of the present invention, the step of determining the dynamic safe discharge gap under the current environment by querying a feature library based on the material type and environmental correction coefficient includes: A preset material reference discharge voltage table is used; the corresponding reference discharge voltage value can be obtained by querying the material reference discharge voltage table according to the material category. The environmental correction factor is used to calculate the corrected discharge voltage by applying the reference discharge voltage value to the environmental correction factor. A preset voltage gap conversion model data table is used. Based on the corrected discharge voltage, the voltage gap conversion model data table is queried to find the air breakdown discharge gap that matches the corrected discharge voltage. Material type, environmental correction factor and air breakdown discharge gap are stored in feature library to establish mapping relationship; Based on the material type and environmental correction factor, the feature database is queried to find the air breakdown discharge gap that matches the material type and environmental correction factor, and this gap is identified as the dynamic safe discharge gap.

[0013] As a preferred embodiment of the floating object identification and early warning method of the present invention, the step of obtaining the spatial distance between the target floating object and the substation equipment based on the movement trajectory, and determining the bridging trend of the target floating object through the movement trajectory includes: Calculate the position and velocity parameters of the target debris; By combining the deformation feature vector, the position and velocity parameters are predicted to obtain the trajectory of the target hanging object; Obtain the spatial coordinates of the device, calculate the distance between the motion trajectory and the spatial coordinates, and determine the bridging trend of the target hanging object based on the motion trajectory.

[0014] As a preferred embodiment of the hanging object identification and early warning method of the present invention, the step of calculating the threat level by combining spatial distance, dynamic safe discharge gap and bridging trend, and outputting an early warning signal based on the threat level includes: The spatial distance is scaled to obtain the actual physical distance; The threat ratio is obtained by comparing the actual physical distance with the dynamic safe discharge gap; Based on the threat ratio and bridging trend, determine the threat level and the rate of change of the threat level per unit time; Weights are assigned to threat levels and rates of change, and a weighted sum is performed to generate an early warning signal.

[0015] Secondly, the present invention provides a hanging object identification and early warning system, comprising: The visual perception and target locking module is used to identify and track the target hanging object and obtain the pixel coordinate information of the target hanging object; The deformation topology feature extraction module is used to extract the non-rigid motion deformation features of the target hanging object and construct a deformation feature vector based on the non-rigid motion deformation features. The environmental perception and material classification module is used to extract the image optical features of the target hanging object based on the pixel coordinate information, collect environmental sensor data of the substation, and obtain the material environmental parameters of the target hanging object in response to the acquisition of the image optical features and the environmental sensor data. The dynamic safety gap assessment module is used to extract the air breakdown discharge gap under the current environment, construct a feature library characterizing the mapping relationship between the material type, the environmental correction coefficient and the air breakdown discharge gap, and query the feature library according to the material type and the environmental correction coefficient to determine the dynamic safety discharge gap under the current environment. The threat evolution and early warning assessment module is used to predict the motion trajectory of the target hanging object based on the pixel coordinate information and the deformation feature vector, obtain the spatial distance between the target hanging object and the equipment based on the motion trajectory, determine the bridging trend of the target hanging object through the motion trajectory, calculate the threat level by combining the spatial distance, the dynamic safety discharge gap and the bridging trend, and output an early warning signal based on the threat level.

[0016] Thirdly, the present invention provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described method for identifying and warning of hanging objects.

[0017] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for identifying and warning of hanging objects.

[0018] Compared with existing technologies, the beneficial effects of this invention are as follows: By calculating the spatial variation characteristics of the optical flow field of pixels within the target area and classifying and tracking topological singularities, the non-rigid deformation characteristics of the target are quantified, enabling the differentiation between rigid translation of hanging objects and dangerous non-rigid twisting and stretching. Combined with multi-source sensing data and material correction coefficients, a dynamic safety discharge gap is generated, eliminating the limitation of fixed safety distances. This invention not only predicts the minimum spatial distance between hanging objects and equipment based on their motion trajectory but also introduces the judgment of bridging trends, combining threat levels and change rates for weighted fusion evaluation, effectively improving the accuracy and urgency assessment capabilities of substation hanging object early warning. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the overall process of the hanging object identification and early warning method according to an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram illustrating the construction of a deformation feature vector based on non-rigid motion deformation characteristics according to an embodiment of the present invention.

[0022] Figure 3 This is a flowchart illustrating the steps for obtaining the material and environmental parameters of a target floating object according to an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram illustrating the process of querying a feature library to determine the dynamic safe discharge gap under the current environment, as described in one embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] In some embodiments, a method for identifying and issuing early warnings of hanging objects is provided, including: S1, identify and track the target hanging object, and obtain the pixel coordinate information of the target hanging object.

[0026] S2, extract the non-rigid motion deformation features of the target hanging object, and construct a deformation feature vector based on the non-rigid motion deformation features.

[0027] S3 extracts multi-source sensing data values ​​of the target hanging object based on pixel coordinate information, collects environmental sensor data from the substation, and obtains material environmental parameters of the target hanging object in response to the acquisition of multi-source sensing data values ​​and environmental sensor data.

[0028] S4. Extract the air breakdown discharge gap under the current environment and construct a feature library that characterizes the mapping relationship between material type, environmental correction coefficient and air breakdown discharge gap.

[0029] S5. Based on the material type and environmental correction factor, query the feature library to determine the dynamic safe discharge gap under the current environment.

[0030] S6. Based on pixel coordinate information and deformation feature vector, predict the motion trajectory of the target hanging object, obtain the spatial distance between the target hanging object and the equipment based on the motion trajectory, and determine the bridging trend of the target hanging object through the motion trajectory.

[0031] S7 calculates the threat level by combining spatial distance, dynamic safe discharge gap, and bridging trend, and outputs a warning signal based on the threat level.

[0032] Current technologies primarily rely on traditional convolutional neural networks for static target detection, with outputs focused on the bounding boxes and static categories of hanging objects. They fail to effectively perceive the internal, non-rigid motion characteristics of these objects. For example, existing systems struggle to distinguish between a dry plastic bag merely shifting in the wind and a damp metal film undergoing violent twisting and stretching, resulting in a lack of assessment of discharge risks caused by deformation. Furthermore, existing risk warning mechanisms typically rely on preset, fixed safety distance thresholds, failing to adequately consider the decisive influence of the hanging object's material and the surrounding environment on discharge voltage and air breakdown gap.

[0033] Meanwhile, early warning and judgment mechanisms are often based solely on a snapshot of the current location, making it difficult to distinguish between rapidly approaching sudden threats and slowly approaching potential threats, leading to inaccurate assessments of the urgency of events.

[0034] This invention calculates the spatial variation characteristics of the optical flow field of pixels within a target area and quantifies the non-rigid deformation characteristics of the target through the classification and tracking of topological singularities. It can distinguish between the rigid translation of hanging objects and dangerous non-rigid twisting and stretching. Combining multi-source sensing data and material correction coefficients, it generates dynamic safety discharge gaps, overcoming the limitations of fixed safety distances. This invention not only predicts the minimum spatial distance between hanging objects and equipment based on their motion trajectories but also introduces the judgment of bridging trends. Combined with threat level and rate of change, it performs a weighted fusion assessment, effectively improving the accuracy and urgency assessment capabilities of substation hanging object early warning.

[0035] In some embodiments, refer to Figures 1-4 This paper provides a method for identifying and issuing early warnings of hanging objects, which includes the following steps: S1, identify and track the target hanging object, and obtain the pixel coordinate information of the target hanging object; The steps for implementing the first-level screening mechanism include S1.1 to S1.4.

[0036] S1.1 Extract the moving foreground from the surveillance video and establish a device mask based on the device layout coordinates within the area; Acquire the substation monitoring video sequence Fi. To accurately locate the area of ​​the hanging object and shield against interference from internal substation equipment, first obtain the preset coordinates of the current transformer, such as [(100,150),(300,450)], and the coordinates of the voltage transformer, such as [(400,150,(600,450)]. Based on these coordinates, set the corresponding pixels in the image to 1 and the rest to 0 to establish the device mask M. device .

[0037] For the high-voltage bushing and conductor busbar region, calculate the current frame F. t Compared to the previous frame F t-1 The absolute value of the grayscale difference at the corresponding pixel location, Dt. Set the motion detection threshold T. motion In this embodiment, T motion The value is set to 30, which is twice the average inter-frame noise level of the historical 1000 frames.

[0038] Assign a value of 1 to the positions of pixels in Dt that are greater than 30, and 0 to the rest, to obtain the moving foreground M. motion Call M motion and M device Perform a pixel-by-pixel logical AND operation on the two masks: R overlap =Mmotion ANDM device ; Identify R overlap The region with a pixel value of 1, i.e., the overlapping part of the foreground region and the mask, is the pixel set S of the abnormal motion region. abnormal .

[0039] S1.2, a preset shape library is used to extract the contour geometric features of the candidate region, and the contour geometric features are matched with the parameters in the shape library to obtain the target geometric matching degree; Based on the pixel set S of abnormal motion regions abnormal Traverse and statistically analyze the pixel set S of abnormal motion regions abnormal The total number of pixels contained is used to obtain the area value Area = 150 pixels; Extract the outermost pixel contour, count the total number of pixels in the contour, and obtain the perimeter value P. erimeter =45 pixels; Calculate the aspect ratio of the minimum bounding rectangle, with length L = 20 and width W = 10, resulting in elongation E = 2.0. The formula for calculating elongation is as follows: E=L / W=2.0.

[0040] Retrieve the shape parameters of the hanging object from the preset shape library. For example, shape 1 is a plastic bag, and the parameter set DB1 for shape 1 is {Area}. db1 =160,P db1 =50,E db1 =2.2}.

[0041] The calculated parameters are compared one by one with the parameters in the library to calculate the relative deviation of each parameter, specifically the area deviation Dev. Area =|150-160| / 160=0.0625; Perimeter deviation Dev Perim =∣45-50∣ / 50=0.1; Elongation deviation Dev Elong =|2.0-2.2| / 2.2=0.091. Weights are set based on feature stability: area features are the most stable, so the weight W is... Area =0.4; Elongation weight W Elong =0.3; the perimeter feature is the most variable, weight W Perim =0.3; then calculate the weighted combination matching degree using the following formula: MatchScore = 1 - (W) Area ×Dev Area +W Perim ×Dev Perim +W Elong ×Dev Elong ); Substituting the data above, we can conclude that: MatchScore=1-(0.4×0.0625+0.3×0.1+0.3×0.091)=0.9177.

[0042] S1.3, preset matching degree benchmark value, filter areas where the geometric matching degree of the target exceeds the matching degree benchmark value, and identify them as target hanging objects; extract the spatial position parameters of the target hanging objects and establish pixel coordinate information; Set the matching baseline value T match The value is 0.85. The target geometric fit of 0.9177 calculated in the previous step is compared with the baseline value T. match Comparing with 0.85, since 0.9177 is greater than 0.85, the region is determined to match the characteristics of the target hanging object. The spatial location information of all pixels in this target region is extracted to establish the pixel coordinate set S of the hanging object. p .

[0043] S2, extract the non-rigid motion deformation features of the target hanging object, and construct a deformation feature vector based on the non-rigid motion deformation features.

[0044] The execution steps include S2.1 to S2.3, and the execution steps of S2.1 to S2.3 are as follows: Figure 2 As shown.

[0045] S2.1, lock the target region corresponding to the pixel coordinate information, calculate the dense optical flow field vector of the pixels in the target region, and characterize the non-rigid motion state of the target hanging object.

[0046] Lock the pixel coordinate set S of the hanging object p For example, {[120,160],[120,161],…}, in consecutive video frames F t With F t+1 The corresponding target region R is defined. The dense optical flow vectors of all pixels (x, y) within R are calculated. , where u is the horizontal displacement and v is the vertical displacement. For example, the vector of point (121, 161) is (2.5, 1.0).

[0047] S2.2, in response to the acquisition of dense optical flow field vectors, calculates the spatial variation characteristics of the optical flow field of the corresponding pixel point, as the non-rigid motion deformation characteristics of the target hanging object.

[0048] For the complete optical flow field vector data obtained in S2.1, the spatial variation characteristics of the optical flow field are calculated point by point. The formula for calculating the divergence parameter D(x,y) is as follows: D(x,y)= u / x+ v / y; The formula for calculating the curl parameter C(x,y) is: C(x,y)= v / x- u / y.

[0049] At this point, the vector (2.5, 1.0) for the aforementioned point (121, 161) is calculated, yielding D(121, 161) = 0.5 and the curl parameter C(121, 161) = -0.2. The D(x, y) and C(x, y) parameter values ​​for all pixels within region R are then summarized to establish the optical flow field divergence curl values.

[0050] S2.3, set a zero-value deviation reference, compare the spatial variation characteristics of the optical flow field with the zero-value deviation reference, and obtain the comparison results; based on the comparison results, identify and classify the topological singularities in the dense optical flow field vectors, including source points, sink points and saddle points; and statistically analyze the spatial distribution of the topological singularities.

[0051] Set zero value deviation from the reference The value is 0.05. The D(x,y) and C(x,y) values ​​for each pixel are... Comparison and judgment: If D(x,y)>0.05 and |C(x,y)|<0.05, it is identified as the source point; If D(x,y)<-0.05 and |C(x,y)|<0.05, it is identified as a sink point; If |D(x,y)| < 0.05 and |C(x,y)| < 0.05, and the neighborhood flow field satisfies the saddle point shape, it is identified as a saddle point.

[0052] Count the number of the three types of topological singularities in the current frame Ft, the source point Nsrc,t=2, the sink point Nsnt,t=1, and the saddle point Nsad,t=0, and generate the spatial location set Lsrc,t={(130,170),…}, thus obtaining the distribution of topological singularities.

[0053] S2.4. Track the relative distance and number evolution of topological singularities to construct deformation feature vectors.

[0054] Based on the above topological singular point distribution, the spatial locations of all singular points within a continuous time frame can be tracked, for example, Lsrc,t={(130,170),(135,172)} and . The following formula is used to calculate The Euclidean distance between all pairwise singular points within a frame is given by the formula: ; For example, the distance between points (130, 170) and (140, 180) is: ; All distance values ​​are aggregated into a set of Euclidean distances, Dt={14.14,…}.

[0055] Simultaneously obtain the total number of singularities N in the current frame. t =3 and the total number of singularities N in the next frame t+1 =4, calculate the rate of change of the total number of singularities per unit time: ΔNt=N t+1 -N t =1; The average of the set of Euclidean distances =15.2 is combined with the rate of change of quantity ΔNt=1 to establish the deformation topological feature vector Vt=[15.2,1].

[0056] S3 extracts multi-source sensing data values ​​of the target hanging object based on pixel coordinate information, collects environmental sensor data from the substation, and obtains material environmental parameters of the target hanging object in response to the acquisition of multi-source sensing data values ​​and environmental sensor data.

[0057] The execution steps include S3.1 to S3.3, and the execution steps of S3.1 to S3.3 are as follows: Figure 3 As shown.

[0058] S3.1: Collect environmental sensor data, locate the image region corresponding to the pixel coordinate information, and calculate the average brightness value and matrix contrast value of the image region.

[0059] Obtain the real-time reading H=75% from the humidity sensor in the substation, and the real-time reading P=101.3 kPa from the barometric pressure sensor. Lock the image region R. float Extracting image region R float It calculates all pixels and converts the RGB value of each pixel to a grayscale value G(x,y).

[0060] Calculate the sum of gray values ​​∑G(x,y)=185000, divide by the total number of pixels N=1000, and obtain the average brightness value L. avg =185. Set the gray-level co-occurrence matrix (GLCM), with calculation parameters of step size 1 and angle 0 degrees, and calculate R. float The matrix contrast value C of the region glcm The calculation process is described in the following formula, where... and This represents the grayscale value of a pair of adjacent pixels in an image. For example, i is the grayscale value of the center pixel, and j is the grayscale value of the adjacent pixel. ; S3.2, combine environmental sensor data, average brightness value and matrix contrast value to generate multi-source sensing data value.

[0061] Humidity reading H=75, air pressure reading P=101.3, average brightness value L avg =185, Matrix contrast value C glcm =2.5 are combined to generate multi-source sensing data value Vmulti=[75,101.3,185,2.5].

[0062] S3.3, preset multiple material feature ranges, extract average brightness value and matrix contrast value based on multi-source sensing data values, compare with the preset multiple material feature ranges, and classify the material category of the target hanging object; preset environmental correction parameter library, call the multi-source sensing data values ​​to enter the environmental correction parameter library to perform retrieval, and determine the environmental correction coefficient.

[0063] Based on the average brightness value Lavg=185 extracted from Vmulti and the matrix contrast value Cglcm=2.5, a preset material feature range library is obtained: Material 1 is a plastic bag, and the range of material 1 (Range 1) is {L}. avg ∈[150,200],C glcm ∈[2.0,3.0]}; Material 2 is a metal foil, and the range of material 2 is {L}. avg ∈[210,250],C glcm ∈[0.5,1.5]}.

[0064] L avg =185 and C glcm =2.5 is compared with the above range. Since both parameters meet the characteristic range of material 1, the target material is classified as plastic bag.

[0065] Subsequently, the H=75 and P=101.3 values ​​from the multi-source sensing data were retrieved to search the environmental correction parameter library.

[0066] The default environmental correction rule is: if the material is a plastic bag and H>70%, the humidity correction factor K is... H =1.1; if P>100kPa, the pressure correction factor K P =0.95. After retrieval, based on the multi-source sensing data values ​​H=75 and P=101.3, the environmental correction coefficient was determined to be K. H =1.1,K P =0.95.

[0067] S4, the steps of extracting the air breakdown discharge gap under the current environment, constructing a feature library characterizing the mapping relationship between material type, environmental correction coefficient and air breakdown discharge gap, and determining the dynamic safe discharge gap under the current environment include:

[0068] The execution steps include S4.1 to S4.4, and the execution steps of S4.1 to S4.4 are as follows: Figure 4 As shown.

[0069] S4.1. Preset material reference discharge voltage table, and obtain the corresponding reference discharge voltage value by querying the material reference discharge voltage table according to the material category.

[0070] Based on the target material category (plastic bag), the preset material reference discharge voltage table is consulted. The material reference discharge voltage table presets the reference discharge voltage V corresponding to the plastic bag. base =150kV, the reference discharge voltage V corresponding to the metal foil base =50kV. Retrieve the reference discharge voltage value V corresponding to the current material of the plastic bag. base =150kV.

[0071] S4.2, call the environmental correction factor, calculate the corrected discharge voltage by combining the reference discharge voltage value with the environmental correction factor.

[0072] Call the environmental correction factor K obtained in S3.3 H =1.1 and K P =0.95. Calculate the corrected discharge voltage using the following formula: V corr =V base ×K H ×K P ; Substituting specific numerical values, the calculation process is as follows: V corr =150×1.1×0.95=156.75kV; The corrected discharge voltage was calculated to be 156.75 kV.

[0073] S4.3, preset voltage gap conversion model data table, based on the corrected discharge voltage, query the voltage gap conversion model data table, and retrieve the air breakdown discharge gap that matches the corrected discharge voltage.

[0074] Based on the corrected discharge voltage V corr =156.75kV, consult the preset voltage gap conversion model data table. The relationship between the preset voltage level and the gap in the voltage gap conversion model data table is as follows: The gap is 1.0m for voltage range [100, 150) kV, 1.5m for voltage range [150, 200) kV, and 2.0m for voltage range [200, 250) kV.

[0075] By comparing the values ​​of 156.75kV with each voltage range, it was determined that the corrected discharge voltage of 156.75kV falls within the range of [150, 200)kV. The gap data of 1.5m corresponding to this voltage range was retrieved and extracted as the air breakdown discharge gap under the current environment.

[0076] Material category: plastic bags; Environmental correction factor K H =1.1,K P =0.95 and the gas breakdown discharge gap of 1.5m are used as a mapping record and stored in the feature library to establish a mapping relationship.

[0077] S4.4. Based on the material type and environmental correction factor, query the feature library to determine the dynamic safe discharge gap under the current environment.

[0078] In real-time detection applications, the system, based on the currently identified material category as plastic bag, uses an environmental correction factor K. H =1.1,K P =0.95, used as a combined index, is directly searched in the feature library constructed in step S4 to retrieve the matching air breakdown discharge gap of 1.5m.

[0079] The search result was determined as the dynamic safety discharge gap D under the current environment. base =1.5m.

[0080] S5 predicts the motion trajectory of the target hanging object based on pixel coordinate information and deformation feature vector, obtains the spatial distance between the target hanging object and the equipment based on the motion trajectory, and judges the bridging trend of the target hanging object through the motion trajectory.

[0081] The execution steps include S5.1 to S5.2.

[0082] S5.1 Calculate the position and velocity parameters of the target hanging object; combine the deformation feature vector to predict the position and velocity parameters and obtain the motion trajectory of the target hanging object.

[0083] Call the pixel coordinate set S of the hanging object p Calculate the pixel coordinate set S of the hanging object p The average of all pixel coordinates within the range, x i and y i Let x and y represent the x and y coordinates of the i-th pixel within the target area, respectively. ; The following example illustrates how to obtain the centroid position C of the current frame. t =(120,160), combined with the centroid position C of the previous frame t-1 =(118,159), calculate the instantaneous velocity V. t =C t -C t-1 =(2,1).

[0084] Combining the rate of change of the number of singular points in the deformation topological eigenvector Vt=[15.2,1], Vt[1]=1, the velocity adjustment coefficient k=0.1 is set. The predicted velocity, V t Represents instantaneous speed: .

[0085] From the centroid position C of the current frame t Initially, regarding the position of the center of mass and V... pred Perform iterative deduction, P1=C t +V pred =(122.2,161.1); P2=P1+V pred =(124.4,162.2); P3=P2+vpred=(126.6,163.3).

[0086] The final motion trajectory, T, is generated. pred ={P1,P2,P3}.

[0087] S5.2, obtain the spatial coordinates of the device, calculate the distance between the motion trajectory and the spatial coordinates, and determine the bridging trend of the target hanging object based on the motion trajectory.

[0088] Obtain the fixed spatial coordinates {150, 175} of the disconnector contact TA, and obtain the fixed spatial coordinates {155, 165} of the circuit breaker outgoing terminal TB. Call the predicted motion trajectory T. pred Calculate the points P on the trajectory i The Euclidean distances to TA and TB are calculated, and the minimum value d among all calculated distances is selected. min =25.7.

[0089] Set the bridging threshold D cross For a distance of 40 pixels, check whether the spatial coordinates of the predicted motion trajectory end point P3 and the aforementioned disconnector switch contact TA and circuit breaker outgoing terminal TB are both less than 40 pixels.

[0090] Calculating d(P3,TA) yields 25.7, where 25.7 < 40. Calculating d(P3,TB) yields 28.5, where 28.5 < 40. Both conditions are met, therefore the target hanging object is determined to have a bridging tendency T.rend =1.

[0091] S6 calculates the threat level by combining spatial distance, dynamic safe discharge gap, and bridging trend, and outputs a warning signal based on the threat level.

[0092] The execution steps include S6.1 to S6.3.

[0093] S6.1, perform scale conversion on the spatial distance to obtain the actual physical distance; compare the actual physical distance with the dynamic safety discharge gap to obtain the threat ratio.

[0094] Set the pixel-to-meter conversion factor K p2m =0.05m / pixel, which can reduce the minimum spatial distance d min =25.7 pixels converted to actual physical distance, K p2m This setting is based on the known fact that a 1-meter-long insulator string occupies 20 pixels in the image during camera calibration. d min-meter =d min ×K p2m =25.7×0.05=1.285m.

[0095] The actual physical distance is 1.285m, and the dynamic safety discharge gap D is... base To compare with a height of 1.5m, calculate the threat ratio using the following formula: R T =D base / d min-meter =1.5 / 1.285≈1.17.

[0096] S6.2, based on the threat ratio and bridging trend, determine the threat level and the rate of change of the threat level per unit time.

[0097] Based on the threat ratio R T Threat Level Classification: R T ≤1.0 is level 0, 1.0 <R T ≤1.5 is level 1, R T A score of 1.5 indicates a level of 2.

[0098] The current threat ratio of 1.17 is in the range (1.0, 1.5), so the initial threat is determined to be L. t Level 1. Combined with the bridging trend T rend =1, calculate the final threat level L final =L t +T rend =1+1=2.

[0099] Get the final level L from the previous moment. final,t-1=1, calculate the rate of change ΔL of the threat level per unit time. t =L final -L final,t-1 =2-1=1, thus obtaining the threat level value and the rate of change [2,1].

[0100] S6.3 assigns weights to threat levels and change rates, performs weighted summation, and generates early warning signals.

[0101] For threat level value L final =2 and the rate of change ΔLt=1, respectively, set the first weighting coefficient W L =0.7 and the second weighting coefficient W ΔL =0.3. The warning priority index is calculated using the above parameters, and the formula is: P riority =L final ×W L +Δ Lt ×W ΔL； Substituting the values ​​yields a specific example: P riority =2×0.7+1×0.3=1.7.

[0102] The system uses the final calculated warning priority index P riority The magnitude of the value dynamically triggers the output of a corresponding level of warning signal.

[0103] The preset warning threshold is 0.8 for the first warning, 1.4 for the second, and 1.8 for the third. The system determines the warning priority index P. riority The threshold range to be entered: When P riority When the value is less than 0.8, the system determines it to be in a safe state and does not trigger an alert. When 0.8≤P riority When the value is less than 1.4, the system determines it as a Level 1 warning, triggers a pop-up notification on the remote monitoring terminal, and sends a notification message to the maintenance personnel. When 1.4≤P riority When the value is less than 1.8, the system determines it as a level 2 warning, triggers the audible and visual alarms at the substation site and the central control room, locks and tracks the target, and starts automatic video recording and saving. When P riority When the value is ≥1.8, the system determines it as a Level 3 severe warning, triggers the emergency linkage interface, disconnects the circuit breaker control circuit that has a tendency to cross-connect with the hanging object, and implements forced ventilation or dehumidification linkage. Substituting into the specific calculation example of this embodiment, P... riority=1.7 Since 1.4≤1.7<1.8, the system confirms it as a level two warning, immediately triggering the audible and visual alarms at the substation site and the central control room, locking the pixel coordinate set of the hanging object onto the monitoring screen, and starting key tracking recording.

[0104] This embodiment also provides a hanging object identification and early warning system, including: The visual perception and target locking module is used to identify and track target hanging objects and obtain the pixel coordinate information of the target hanging objects; The deformation topology feature extraction module is used to extract the non-rigid motion deformation features of the target hanging object and construct a deformation feature vector based on the non-rigid motion deformation features. The environmental perception and material classification module is used to extract the image optical features of the target hanging object based on pixel coordinate information, collect environmental sensor data of the substation, and obtain the material environmental parameters of the target hanging object in response to the acquisition of image optical features and environmental sensor data. The dynamic safety gap assessment module is used to extract the air breakdown discharge gap under the current environment, construct a feature library that characterizes the mapping relationship between material type, environmental correction factor and air breakdown discharge gap, and query the feature library according to material type and environmental correction factor to determine the dynamic safety discharge gap under the current environment. The threat evolution and early warning assessment module is used to predict the movement trajectory of the target hanging object based on pixel coordinate information and the deformation feature vector, obtain the spatial distance between the target hanging object and the equipment based on the movement trajectory, determine the bridging trend of the target hanging object through the movement trajectory, calculate the threat level by combining the spatial distance, dynamic safety discharge gap and bridging trend, and output an early warning signal based on the threat level.

[0105] This embodiment also provides an electronic device suitable for identifying and warning of hanging objects, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for identifying and warning of hanging objects as proposed in the above embodiment.

[0106] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the method for identifying and warning of hanging objects as proposed in the above embodiments.

[0107] The storage medium proposed in this embodiment and the method for identifying and warning of hanging objects proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0108] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0109] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying and issuing early warnings of hanging objects, characterized in that, include: Identify and track target hanging objects to obtain their pixel coordinate information; Extract the non-rigid motion deformation features of the target hanging object, and construct a deformation feature vector based on the non-rigid motion deformation features; Image features of the target hanging object are extracted based on pixel coordinate information, and environmental sensor data of the substation are collected. In response to the acquisition of image features and environmental sensor data, material environmental parameters of the target hanging object are obtained. Extract the air breakdown discharge gap under the current environment and construct a feature library to characterize the mapping relationship between material type, environmental correction coefficient and air breakdown discharge gap; Based on the material type and environmental correction factor, the feature library is queried to determine the dynamic safe discharge gap under the current environment; Based on pixel coordinate information and deformation feature vector, predict the motion trajectory of the target hanging object, obtain the spatial distance between the target hanging object and the equipment based on the motion trajectory, and determine the bridging trend of the target hanging object through the motion trajectory. By combining spatial distance, dynamic safe discharge gap, and bridging trend, the threat level is calculated, and an early warning signal is output based on the threat level.

2. The method for identifying and warning of hanging objects as described in claim 1, characterized in that, The steps for identifying and tracking target hanging objects and obtaining their pixel coordinate information include: Extract moving foreground from surveillance video and establish device masks based on the coordinates of devices deployed within the area; In response to the acquisition of the device mask, the moving foreground is filtered to obtain candidate regions; A pre-defined shape library is used to extract the contour geometric features of candidate regions. The contour geometric features are then matched with the parameters in the shape library to obtain the target geometric matching degree. A preset matching baseline value is used to filter areas where the geometric matching degree of the target exceeds the matching baseline value, and these areas are identified as target hanging objects. Extract the spatial position parameters of the target floating object and establish pixel coordinate information.

3. The method for identifying and warning of hanging objects as described in claim 1 or 2, characterized in that, The steps for extracting the non-rigid motion deformation features of the target hanging object and constructing a deformation feature vector based on these features include: Lock the target region corresponding to the pixel coordinate information, calculate the dense optical flow field vector of the pixels in the target region, and characterize the non-rigid motion state of the target hanging object; In response to the acquisition of dense optical flow field vectors, the spatial variation characteristics of the optical flow field of the corresponding pixel are calculated as the non-rigid motion deformation characteristics of the target hanging object; A zero-value deviation reference is set, and the spatial variation characteristics of the optical flow field are compared with the zero-value deviation reference to obtain the comparison results. Based on the comparison results, topological singularities in the dense optical flow field vectors are identified and classified. The spatial distribution and temporal changes of topological singularities are statistically analyzed, the relative distances between topological singularities and the evolution of the number of topological singularities are tracked, and deformation feature vectors are constructed.

4. The method for identifying and warning of hanging objects as described in claim 3, characterized in that, The steps to obtain the material and environmental parameters of the target hanging object include: Collect environmental sensor data, locate the image region corresponding to pixel coordinate information, and calculate the average brightness value and matrix contrast value of the image region. The environmental sensor data, average brightness value, and matrix contrast value are combined to generate multi-source sensing data values. Multiple material feature ranges are preset, and based on multi-source sensing data values, average brightness values ​​and matrix contrast values ​​are extracted and compared with the preset multiple material feature ranges to classify the material category of the target hanging object. A preset environmental correction parameter library is established. Multi-source sensing data values ​​are called into the environmental correction parameter library for retrieval to determine the environmental correction coefficient.

5. The method for identifying and warning of hanging objects as described in claim 4, characterized in that, The steps for determining the dynamic safe discharge gap under the current environment by querying the feature library based on the material type and environmental correction factor include: A preset material reference discharge voltage table is used to obtain the corresponding reference discharge voltage value by querying the material reference discharge voltage table according to the material category. The environmental correction factor is used to calculate the corrected discharge voltage by applying the reference discharge voltage value to the environmental correction factor. A preset voltage gap conversion model data table is used. Based on the corrected discharge voltage, the voltage gap conversion model data table is queried to find the air breakdown discharge gap that matches the corrected discharge voltage. In response to the acquisition of material type, environmental correction factor and air breakdown discharge gap, a mapping relationship is established and a feature library is constructed; Based on the material type and environmental correction factor, the feature database is queried to find the air breakdown discharge gap that matches the material type and environmental correction factor, and this gap is identified as the dynamic safe discharge gap.

6. The method for identifying and warning of hanging objects as described in claim 5, characterized in that, The steps of obtaining the spatial distance between the target suspended object and the substation equipment based on the motion trajectory, and determining the bridging trend of the target suspended object based on the motion trajectory, include: Calculate the position and velocity parameters of the target hanging object; By combining the deformation feature vector, the position and velocity parameters are predicted to obtain the trajectory of the target hanging object; Obtain the spatial coordinates of the device, calculate the distance between the motion trajectory and the spatial coordinates, and determine the bridging trend of the target hanging object based on the motion trajectory.

7. The method for identifying and warning of hanging objects as described in claim 6, characterized in that, The steps for calculating the threat level and outputting a warning signal based on the threat level, by combining spatial distance, dynamic safe discharge gap, and bridging trend, include: The spatial distance is scaled to obtain the actual physical distance; The threat ratio is obtained by comparing the actual physical distance with the dynamic safe discharge gap; Based on the threat ratio and bridging trend, determine the threat level and the rate of change of the threat level per unit time; Weights are assigned to threat levels and rates of change, and a weighted sum is calculated to generate an early warning signal.

8. A system for identifying and warning of hanging objects, using the method for identifying and warning of hanging objects as described in claim 7, characterized in that, include: The visual perception and target locking module is used to identify and track target hanging objects and obtain the pixel coordinate information of the target hanging objects; The deformation topology feature extraction module is used to extract the non-rigid motion deformation features of the target hanging object and construct a deformation feature vector based on the non-rigid motion deformation features. The environmental perception and material classification module is used to extract the image optical features of the target hanging object based on pixel coordinate information, collect environmental sensor data of the substation, and obtain the material environmental parameters of the target hanging object in response to the acquisition of image optical features and environmental sensor data. The dynamic safety gap assessment module is used to extract the air breakdown discharge gap under the current environment, construct a feature library that characterizes the mapping relationship between material type, environmental correction factor and air breakdown discharge gap, and query the feature library according to material type and environmental correction factor to determine the dynamic safety discharge gap under the current environment. The threat evolution and early warning assessment module is used to predict the movement trajectory of the target hanging object based on pixel coordinate information and the deformation feature vector, obtain the spatial distance between the target hanging object and the equipment based on the movement trajectory, determine the bridging trend of the target hanging object through the movement trajectory, calculate the threat level by combining the spatial distance, dynamic safety discharge gap and bridging trend, and output an early warning signal based on the threat level.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the hanging object identification and early warning method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the hanging object identification and early warning method according to any one of claims 1 to 7.

Images