Power transmission line external damage identification method and device based on SSD and Kalman filtering

Abstract

The invention provides a power transmission line external damage tracking method and device based on SSD and Kalman filtering. The method comprises the following steps: acquiring a to-be-monitored video image sequence; inputting the sequence into a pre-trained SSD model, and identifying each external force damage source; respectively constructing a corresponding motion prediction model for each identified external force damage source through a Kalman filtering algorithm; obtaining the motion state of the corresponding external force damage source through the motion detection model, setting a search area based on the mass center position of the motion state, predicting the mass center position of the external force damage source in the next frame of image, and achieving the tracking of the external force damage source. According to the method, the SSD model and the Kalman filtering are combined, the motion detection model is constructed, the search area is set according to the motion state of the external force damage source, full-range search does not need to be carried out on the whole image, and the calculation amount can be reduced; and meanwhile, comparison with a preset database is not needed, so that the retrieval process is saved.

Description

Technical Field

[0001] The present invention relates to the field of image recognition and tracking, and in particular to a transmission line external failure tracking method, device and computer-readable storage medium based on SSD and Kalman filtering. Background Art

[0002] With the development of economy and technology, the demand for electric power resources is increasing. Maintaining the safe, stable and healthy operation of the power system is crucial to the development of various industries. However, the construction of infrastructure such as roads and houses needs to be carried out within the transmission line corridor, and most of the external damage to the transmission line is caused by large-scale machinery and equipment that are illegally constructed.

[0003] At present, the main way to protect transmission lines from external damage is through manual inspections, such as signing safety agreements, issuing safety notices, and setting up safety signs. However, this approach has a long cycle and cannot monitor the transmission line corridors for construction activities around the clock. On the other hand, with the development of monitoring technology, existing technologies also use computer vision, deep learning and other technologies to identify different types of external force damage sources for transmission lines. However, this technology can only determine the detection results by comparing with sample data in a preset database, which requires a large amount of data retrieval, calculation and comparison analysis in the database, and there is a technical problem of large amount of calculation. Summary of the invention

[0004] The present invention provides a transmission line external damage tracking method, device and computer-readable storage medium based on SSD and Kalman filtering, which can predict the center of mass position of the external force damage source in the next frame image within a set search area without the need for a full image search, so as to solve the technical problem of high computational complexity in the prior art of external damage identification.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a transmission line external failure tracking method based on SSD and Kalman filtering, comprising:

[0006] Acquire a video image sequence to be monitored;

[0007] Inputting the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence;

[0008] Through the Kalman filter algorithm, a corresponding motion prediction model is constructed for each identified external force damage source; and through the motion detection model, the motion state of the corresponding external force damage source is obtained, and then the search area is set based on the centroid position of the motion state, and the centroid position of the external force damage source in the next frame image is predicted through the search area to achieve tracking of the external force damage source; wherein, the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model.

[0009] As a preferred solution, the search area is set based on the centroid position of the motion state, specifically:

[0010] The centroid of the motion state is set as the center of the search area, and the number of image pixels corresponding to the external force damage source of the motion state is used as the area of ​​the search area, so as to generate the corresponding search area.

[0011] As a preferred solution, the center of mass of the motion state is obtained by horizontal and vertical projection of a binary image corresponding to the external force damage source;

[0012] The horizontal projection H(y) is:

[0013]

[0014] Wherein, g(x,y) represents the gray value of the pixel (x,y) in the binary image. H(y) is a one-dimensional array, the length of which is the height of the binary image;

[0015] The projection in the vertical direction is:

[0016]

[0017] Wherein, H(x) is a one-dimensional array, the array length is the height of the binary image, and the size of the binary image is M×N.

[0018] As a preferred solution, the area of ​​the search area is:

[0019]

[0020] Among them, B(x,y) is the binary image of the external force damage source, O i To detect the target, A i is the area of ​​the detection target.

[0021] As a preferred solution, the observation equation of the motion detection model is:

[0022] y(t)=H(t)x(t)+v(t);

[0023] Among them, y(t) is the observation vector, H(t) is the observation matrix, and v(t) is the observation error.

[0024] As a preferred solution, the SSD model includes VGG-Base, Extra-Layers and Pred-Layers;

[0025] Among them, the VGG-Base is used to extract the features of the image, the Extra-Layers is used to obtain the feature map according to the extracted features; the Pred-Layers is used to predict the feature map to obtain the coordinates, confidence and category information of the prediction box.

[0026] As a preferred solution, the scale parameter of the feature map is:

[0027]

[0028] Among them, S k represents the scale parameter of the kth feature map, S min represents the minimum value of the scale parameter, S max represents the maximum value of the scale parameter, and m represents the number of feature maps.

[0029] As a preferred solution, the training process of the SSD model includes:

[0030] Acquire several sample images with external force damage sources;

[0031] The LabelImg tool is used to label the external force damage sources in the image data, and an XML-type image dataset is obtained, and the type information of each external force damage source is obtained at the same time; among them, the types of external force damage sources include mechanical type, foreign body type, fire type, landslide type and human type;

[0032] The image data set is randomly divided into a training set, a validation set and a test set according to a preset ratio;

[0033] A basic model based on the SSD algorithm is constructed through the training set, and the basic model is trained through the training set, the test set and the validation set. When the basic model meets the preset convergence conditions, the trained SSD model is obtained.

[0034] As a preferred solution, the motion prediction model is specifically:

[0035]

[0036] Where X(t) represents the system state at time t, x and y represent the horizontal and vertical coordinates of the center of mass position of the detection target of the motion detection model, respectively, and v x and v yThey represent the speed of the detection target in the x-axis direction and the y-axis direction per unit time respectively, and ω(t) is the estimation error.

[0037] Correspondingly, the embodiment of the present invention also provides a transmission line external damage tracking device based on SSD and Kalman filtering, including an acquisition module, an identification module and a tracking module; wherein,

[0038] The acquisition module is used to acquire the video image sequence to be monitored;

[0039] The recognition module is used to input the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence;

[0040] The tracking module is used to construct a corresponding motion prediction model for each identified external force damage source through the Kalman filter algorithm; and obtain the motion state of the corresponding external force damage source through the motion detection model, and then set a search area based on the centroid position of the motion state, and predict the centroid position of the external force damage source in the next frame image through the search area to achieve tracking of the external force damage source; wherein the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model.

[0041] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0042] The embodiment of the present invention provides a transmission line external damage tracking method, device and computer-readable storage medium based on SSD and Kalman filtering. The transmission line external damage tracking method includes: obtaining a video image sequence to be monitored; inputting the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence; using the Kalman filtering algorithm, constructing a corresponding motion prediction model for each identified external force damage source; and obtaining the motion state of the corresponding external force damage source through the motion detection model, and then setting a search area based on the centroid position of the motion state, predicting the centroid position of the external force damage source in the next frame image through the search area, and realizing tracking of the external force damage source; wherein the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model. By implementing the embodiments of the present application, combining the SSD model and the Kalman filter, a motion detection model is constructed and a search area is set according to the motion state of the external force destruction source. There is no need to perform a full-range search of the entire image as in the existing technical solutions, which can reduce the amount of calculation. At the same time, there is no need to compare and analyze with a preset database, which saves the process of retrieving the database, can further reduce the amount of calculation, and improve the overall tracking performance. In addition, the state variables are the speed and position of the external force destruction source, which can make the identification and tracking results more in line with the actual situation and improve the accuracy of the prediction of the target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 : A flow chart of an embodiment of a transmission line external failure tracking method provided by the present invention based on SSD and Kalman filtering.

[0044] Figure 2 : It is a structural schematic diagram of an embodiment of a transmission line external break tracking device provided by the present invention based on SSD and Kalman filtering. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Embodiment 1

[0047] Please refer to Figure 1 , a transmission line external failure tracking method based on SSD and Kalman filtering provided by an embodiment of the present invention, comprising steps S1 to S3; wherein,

[0048] Step S1, obtaining a video image sequence to be monitored.

[0049] Step S2: input the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence.

[0050] Step S3, through the Kalman filter algorithm, a corresponding motion prediction model is constructed for each identified external force damage source; and through the motion detection model, the motion state of the corresponding external force damage source is obtained, and then the search area is set based on the center of mass position of the motion state, and the center of mass position of the external force damage source in the next frame image is predicted through the search area to achieve tracking of the external force damage source; wherein, the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model.

[0051] In this embodiment, before performing target detection and target tracking, a target detection model needs to be trained first. This embodiment adopts a neural network model based on SSD. SSD, the full name of which is Single Shot MultiBox Detector, is a target detection algorithm proposed by Wei Liu at ECCV 2016. It is currently one of the main detection frameworks, with a running speed comparable to YOLO and a detection accuracy comparable to Faster RCNN.

[0052] The training process of the SSD model used in this embodiment includes:

[0053] Acquire several sample images with external force damage sources;

[0054] The LabelImg tool is used to label the external force damage sources in the image data, and an XML-type image dataset is obtained, and the type information of each external force damage source is obtained at the same time; among them, the types of external force damage sources include mechanical type, foreign body type, fire type, landslide type and human type;

[0055] The image dataset is randomly divided into a training set, a validation set and a test set according to a preset ratio (preferably 8:1:1);

[0056] A basic model based on the SSD algorithm is constructed through the training set, and the basic model is trained through the training set, the test set and the validation set. When the basic model meets the preset convergence conditions, the trained SSD model is obtained.

[0057] The training set and the validation set are used to train the basic model, find the optimal model parameters, or determine the stopping point of the training. Then, the performance of the model can be verified by the validation set, and whether it meets the convergence requirements can be judged according to the error, etc. If it meets the requirements, the trained SSD model is obtained. The training of the SSD model can be applied before step S2 or before step S1. After training it, the trained SSD model can be called at any time.

[0058] Furthermore, the SSD model includes VGG-Base, Extra-Layers and Pred-Layers.

[0059] Among them, the VGG-Base is used to extract the features of the image, and the Extra-Layers is used to further process the features extracted by VGG, increase the model's receptive field of the image, and obtain feature maps at the same time, so that the feature maps obtained by Extra-layers carry more abstract information. The feature maps to be predicted in this embodiment are composed of six feature maps, corresponding to different scales. The Pred-Layers is used to predict the feature maps to obtain the coordinates, confidence and category information of the prediction box.

[0060] Preferably, the scale parameter of the feature map is:

[0061]

[0062] Among them, S k represents the scale parameter of the kth feature map, S min represents the minimum value of the scale parameter, preferably 0.2, S max Represents the maximum value of the scale parameter, preferably 0.9, and m represents the number of feature maps. Through the above step S2, images containing external force damage sources can be screened out, or identification can be performed for a real-time video image sequence. That is, the external force damage source in the latest image is identified, so that the state and position of the next frame can be predicted in step S3.

[0063] For the above step S3, this embodiment is a motion prediction model established based on the Kalman filter algorithm. When multiple external force damage sources are identified in the same image, a motion prediction model is constructed for each external force damage source. When only one is identified, only one needs to be constructed. The state variables can be the speed and position of the detection target (for example, the external force damage source at the next moment). Assuming that the state variables at a certain moment are the speed and position of the target, during the tracking process, since the time interval between two adjacent frames of images is very short, the target motion state changes little, so it can be assumed that the target is in uniform motion within a unit time interval. Then the motion system is a linear state model:

[0064] X(t)=A(Δt)X(t-Δt)+ω(t-Δt);

[0065] Among them, X(t) represents the system state at time t, A(Δt) represents the state matrix within time Δt, and ω(t) represents the prediction error.

[0066] The system state and the estimated error are represented by a four-dimensional vector including the target position and size and their changes, as shown in the following formula:

[0067]

[0068] The above formula establishes the state equation of the motion model, where x and y represent the horizontal and vertical coordinates of the center of mass of the detection target, and v x 、v y Represents the speed of the target in the x-axis and y-axis directions per unit time. The system state includes the four-dimensional vector of the target position and size and its changes to represent the system state and estimation error, as shown in the following formula:

[0069]

[0070] The above formula establishes the state equation of the motion model, where x and y represent the horizontal and vertical coordinates of the center of mass of the detection target, and v x 、v y represents the speed of the target in the x-axis and y-axis directions per unit time, and ω(t) is the estimation error.

[0071] In the process of monitoring the video sequence, the predicted value of the running model can be continuously corrected through the images of different frames, and the center of mass and area of ​​the detection target can be calculated, and the center of mass and area can be used as characteristic values. For example, the center of mass of the motion state is set as the center of the search area, and the number of image pixels corresponding to the external force damage source of the motion state is used as the area of ​​the search area, thereby generating the corresponding search area. After entering the next frame of the image, the center of mass position that best matches the detection target area is found in the search area, that is, the prediction of the next frame position of the external force damage source is realized, and the tracking function is realized.

[0072] As a preferred implementation, the centroid of the target can be obtained by projecting the binary image in the horizontal and vertical directions. Assuming the size of the binary image is M×N, its projection in the horizontal direction is:

[0073]

[0074] Wherein, g(x,y) represents the gray value of the pixel (x,y) in the binary image. H(y) is a one-dimensional array, the length of which is the height of the binary image;

[0075] The projection in the vertical direction is:

[0076]

[0077] Where H(x) is a one-dimensional array whose length is the height of the binary image.

[0078] As for the area of ​​the search region, it can be defined as the detection area of ​​the target as the characteristic value of the motion state. For example, the number of image pixels belonging to the detection target is taken as the area of ​​the target. i Area A i The calculation formula is as follows:

[0079]

[0080] Where B(x,y) is a binary image.

[0081] As a preferred implementation, the Kalman filter algorithm is used to predict the system state through the observed value. It can be assumed that there is a linear relationship between the system state and the observed value, then:

[0082] y(t)=H(t)x(t)+v(t)

[0083] Where y(t) is the observation vector, H(t) is the observation matrix, and v(t) is the observation error. By implementing the embodiment of the present application, a motion detection model is constructed for the video (sequence) data to be monitored by combining the SSD model and Kalman filtering, and external force damage sources of transmission lines in surveillance cameras, for example, can be identified and tracked. This method can accurately predict the position of the target object, which not only reduces the impact of noise on the results, but also effectively narrows the search range of feature extraction by constructing a search area without searching the entire range of the image, reducing the amount of calculation, and making the recognition and tracking results more in line with the actual situation.

[0084] Accordingly, refer to Figure 2 The embodiment of the present invention also provides a transmission line external damage tracking device based on SSD and Kalman filtering, including an acquisition module 101, an identification module 102 and a tracking module 103; wherein,

[0085] The acquisition module 101 is used to acquire a video image sequence to be monitored;

[0086] The identification module 102 is used to input the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence;

[0087] The tracking module 103 is used to construct a corresponding motion prediction model for each identified external force damage source through the Kalman filter algorithm; and obtain the motion state of the corresponding external force damage source through the motion detection model, and then set a search area based on the center of mass position of the motion state, and predict the center of mass position of the external force damage source in the next frame image through the search area to achieve tracking of the external force damage source; wherein the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model.

[0088] As a preferred implementation, the tracking module sets a search area based on the centroid position of the motion state, specifically:

[0089] The tracking module sets the centroid of the motion state as the center of the search area, and takes the number of image pixels corresponding to the external force damage source of the motion state as the area of ​​the search area, thereby generating a corresponding search area.

[0090] As a preferred implementation, the center of mass of the motion state is obtained by horizontal and vertical projection of a binary image corresponding to the external force damage source;

[0091] The horizontal projection H(y) is:

[0092]

[0093] Wherein, g(x,y) represents the gray value of the pixel (x,y) in the binary image. H(y) is a one-dimensional array, the length of which is the height of the binary image;

[0094] The projection in the vertical direction is:

[0095]

[0096] Wherein, H(x) is a one-dimensional array, the array length is the height of the binary image, and the size of the binary image is M×N.

[0097] As a preferred implementation, the area of ​​the search area is:

[0098]

[0099] Among them, B(x,y) is the binary image of the external force damage source, O i To detect the target, A i is the area of ​​the detection target.

[0100] As a preferred implementation, the observation equation of the motion detection model is:

[0101] y(t)=H(t)x(t)+v(t);

[0102] Among them, y(t) is the observation vector, H(t) is the observation matrix, and v(t) is the observation error.

[0103] As a preferred implementation, the SSD model includes VGG-Base, Extra-Layers and Pred-Layers;

[0104] Among them, the VGG-Base is used to extract the features of the image, the Extra-Layers is used to obtain the feature map according to the extracted features; the Pred-Layers is used to predict the feature map to obtain the coordinates, confidence and category information of the prediction box.

[0105] As a preferred implementation, the scale parameter of the feature map is:

[0106]

[0107] Among them, S k represents the scale parameter of the kth feature map, S min represents the minimum value of the scale parameter, S max represents the maximum value of the scale parameter, and m represents the number of feature maps.

[0108] As a preferred implementation, the training process of the SSD model includes:

[0109] Acquire several sample images with external force damage sources;

[0110] The LabelImg tool is used to label the external force damage sources in the image data, and an XML-type image dataset is obtained, and the type information of each external force damage source is obtained at the same time; among them, the types of external force damage sources include mechanical type, foreign body type, fire type, landslide type and human type;

[0111] The image data set is randomly divided into a training set, a validation set and a test set according to a preset ratio;

[0112] A basic model based on the SSD algorithm is constructed through the training set, and the basic model is trained through the training set, the test set and the validation set. When the basic model meets the preset convergence conditions, the trained SSD model is obtained.

[0113] As a preferred implementation, the motion prediction model is specifically:

[0114]

[0115] Where X(t) represents the system state at time t, x and y represent the horizontal and vertical coordinates of the center of mass position of the detection target of the motion detection model, respectively, and v x and v y They represent the speed of the detection target in the x-axis direction and the y-axis direction per unit time respectively, and ω(t) is the estimation error.

[0116] Correspondingly, an embodiment of the present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, the transmission line external failure tracking method based on SSD and Kalman filtering is implemented.

[0117] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal, and various interfaces and lines are used to connect various parts of the entire terminal.

[0118] The memory can be used to store the computer program, and the processor realizes various functions of the terminal by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0119] Correspondingly, an embodiment of the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the transmission line external failure tracking method based on SSD and Kalman filtering.

[0120] Wherein, if the module integrated in the transmission line external break tracking device based on SSD and Kalman filtering is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0121] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0122] The embodiment of the present invention provides a transmission line external damage tracking method, device and computer-readable storage medium based on SSD and Kalman filtering. The transmission line external damage tracking method includes: obtaining a video image sequence to be monitored; inputting the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence; using the Kalman filtering algorithm, constructing a corresponding motion prediction model for each identified external force damage source; and obtaining the motion state of the corresponding external force damage source through the motion detection model, and then setting a search area based on the centroid position of the motion state, predicting the centroid position of the external force damage source in the next frame image through the search area, and realizing tracking of the external force damage source; wherein the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model. By implementing the embodiments of the present application, combining the SSD model and the Kalman filter, a motion detection model is constructed and a search area is set according to the motion state of the external force destruction source. There is no need to perform a full-range search of the entire image as in the existing technical solutions, which can reduce the amount of calculation. At the same time, there is no need to compare and analyze with a preset database, which saves the process of retrieving the database, can further reduce the amount of calculation, and improve the overall tracking performance. In addition, the state variables are the speed and position of the external force destruction source, which can make the identification and tracking results more in line with the actual situation and improve the accuracy of the prediction of the target object.

[0123] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A transmission line external failure tracking method based on SSD and Kalman filtering, characterized in that: include: Acquire a video image sequence to be monitored; Inputting the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence; Through the Kalman filter algorithm, a corresponding motion prediction model is constructed for each identified external force damage source; and through the motion detection model, the motion state of the corresponding external force damage source is obtained, and then the search area is set based on the centroid position of the motion state, and the centroid position of the external force damage source in the next frame image is predicted through the search area to achieve tracking of the external force damage source; wherein, the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model.

2. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in claim 1, characterized in that: The setting of the search area based on the center of mass position of the motion state is specifically: The centroid of the motion state is set as the center of the search area, and the number of image pixels corresponding to the external force damage source of the motion state is used as the area of ​​the search area, so as to generate the corresponding search area.

3. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in claim 2, characterized in that: The center of mass of the motion state is obtained by projecting the binary image corresponding to the external force damage source in the horizontal direction and the vertical direction; The horizontal projection H(y) is: Wherein, g(x,y) represents the gray value of the pixel (x,y) in the binary image. H(y) is a one-dimensional array, the length of which is the height of the binary image; The projection in the vertical direction is: Wherein, H(x) is a one-dimensional array, the array length is the height of the binary image, and the size of the binary image is M×N.

4. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in claim 3, characterized in that: The area of ​​the search area is: Among them, B(x,y) is the binary image of the external force damage source, O i To detect the target, A i is the area of ​​the detection target.

5. The transmission line external failure tracking method based on SSD and Kalman filtering as claimed in claim 1, characterized in that: The observation equation of the motion detection model is: y(t)=H(t)x(t)+v(t); Among them, y(t) is the observation vector, H(t) is the observation matrix, and v(t) is the observation error.

6. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in claim 1, characterized in that: The SSD model includes VGG-Base, Extra-Layers and Pred-Layers; Among them, the VGG-Base is used to extract the features of the image, the Extra-Layers is used to obtain the feature map according to the extracted features; the Pred-Layers is used to predict the feature map to obtain the coordinates, confidence and category information of the prediction box.

7. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in claim 6, characterized in that: The scale parameter of the feature map is: Among them, S k represents the scale parameter of the kth feature map, S min represents the minimum value of the scale parameter, S max represents the maximum value of the scale parameter, and m represents the number of feature maps.

8. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in any one of claims 1 to 7, characterized in that: The training process of the SSD model includes: Acquire several sample images with external force damage sources; The LabelImg tool is used to label the external force damage sources in the image data, and an XML-type image dataset is obtained, and the type information of each external force damage source is obtained at the same time; among them, the types of external force damage sources include mechanical type, foreign body type, fire type, landslide type and human type; The image data set is randomly divided into a training set, a validation set and a test set according to a preset ratio; A basic model based on the SSD algorithm is constructed through the training set, and the basic model is trained through the training set, the test set and the validation set. When the basic model meets the preset convergence conditions, the trained SSD model is obtained.

9. A transmission line external failure tracking method based on SSD and Kalman filtering as claimed in any one of claims 1 to 7, characterized in that: The motion prediction model is specifically: Where X(t) represents the system state at time t, x and y represent the horizontal and vertical coordinates of the center of mass position of the detection target of the motion detection model, respectively, and v x and v y They represent the speed of the detection target in the x-axis direction and the y-axis direction per unit time respectively, and ω(t) is the estimation error.

10. A transmission line external damage tracking device based on SSD and Kalman filtering, characterized in that: It includes an acquisition module, an identification module and a tracking module; wherein, The acquisition module is used to acquire the video image sequence to be monitored; The recognition module is used to input the video image sequence into a pre-trained SSD model to identify each external force damage source in each image in the video image sequence; The tracking module is used to construct a corresponding motion prediction model for each identified external force damage source through the Kalman filter algorithm; and obtain the motion state of the corresponding external force damage source through the motion detection model, and then set a search area based on the centroid position of the motion state, and predict the centroid position of the external force damage source in the next frame image through the search area to achieve tracking of the external force damage source; wherein the state variables of the motion prediction model are the speed and position of the external force damage source, and the motion prediction model is a linear state model.

Images