Bird prevention method and system for transformer substation based on radar detection and visible light recognition

Abstract

The application discloses a substation bird prevention method based on radar detection and visible light identification, which is applied to a substation bird prevention system comprising a radar detection unit and a visible light identification unit. The radar detection unit is used for detecting a target object in real time. If the target object is detected, the position and speed of the target object are calculated, and then the target object is identified by the visible light identification unit closest to the position of the target object. If the identification result of the target object is a bird, a fixed bird repelling device is started or an alarm is given and a patrol robot is started to expel the bird according to the speed of the target object. The radar detection unit is used for continuously detecting the target object in real time until the target object cannot be detected or the identification result of the target object is not a bird. The fixed bird repelling device is turned off or the alarm is given and the patrol robot is recovered. The application can accurately identify the bird activity in the substation and formulate an intelligent start-stop strategy of the bird repelling device, overcomes the shortcoming that a traditional expelling type bird repelling device is easily adapted by birds, and improves the bird prevention effect of the substation.

Description

Technical Field

[0001] The present invention relates to a bird-proofing technology for a substation, and in particular to a bird-proofing method and system for a substation based on radar detection and visible light recognition. Background Art

[0002] At present, substations still use traditional bird-repellent devices represented by isolation and expulsion types. The isolation type isolates key locations of substation equipment from birds by installing bird-proof barbs, bird-proof covers, insulating partitions, etc. to reduce the impact of bird activities, but it has disadvantages such as difficult installation and maintenance, easy adaptation by small birds, and entanglement by floating objects; the expulsion type drives away birds through interference such as rotating reflective lights, biological odors, ultrasound, electric shocks, and bionic natural enemies. Although it has the advantages of low price and easy installation, it has the disadvantages of waste of resources caused by long-term operation of the bird-repellent device and poor long-term bird prevention effect due to the easy adaptation of birds. Summary of the Invention

[0003] The technical problem to be solved by the present invention is as follows: In response to the above-mentioned problems of the prior art, a substation bird prevention method and system based on radar detection and visible light recognition is provided, which can accurately identify bird activities in the substation and formulate intelligent start and stop strategies for bird repellent devices, overcoming the shortcomings of traditional expulsion-type bird repellent devices that are easily adapted by birds, and improving the bird prevention effect of substations.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A substation bird prevention method based on radar detection and visible light recognition is applied to a substation bird prevention system including a radar detection unit and a visible light recognition unit. The method comprises the following steps:

[0006] The radar detection unit detects the target in real time. If a target is detected, the position and speed of the target are calculated, and then the visible light recognition unit closest to the target is used to identify the target.

[0007] If the target object is identified as a bird, the fixed bird repellent device is enabled to drive away the bird, and the radar detection unit continues to detect the target object in real time until the target object is no longer detected or the identification result of the target object is not a bird;

[0008] If the target object is identified as a bird and the speed of the target object is zero, an alarm is triggered and the patrol robot is enabled to drive away the bird, and the radar detection unit continues to detect the target object in real time until the target object is no longer detected or the identification result of the target object is not a bird;

[0009] If the target object cannot be detected or the identification result of the target object is not a bird, the fixed bird repellent device is turned off or the alarm is released and the inspection robot is recovered.

[0010] Furthermore, the radar detection unit includes four radars respectively arranged at the four corners of the substation, and the two radars on the same diagonal line cooperate with each other to form a group of bistatic radars. One radar in the bistatic radar acts as a transmitter to transmit signals, and the other radar acts as a receiver to receive signals. When the radar detection unit detects the target object in real time, it specifically obtains the signal transmitted by the transmitter and the signal received by the receiver in each group of bistatic radars in real time, and calculates the change value of the received signal of each receiver. If the change value corresponding to at least one group of bistatic radars is greater than the preset value, the target object is detected.

[0011] Furthermore, the change value of the received signal of each receiver specifically refers to the signal strength change value of the received signal of the receiver compared with the transmitted signal of the corresponding transmitter, or the time difference between the received signal of the receiver and the transmitted signal of the corresponding transmitter, or the frequency change value of the received signal of the receiver compared with the transmitted signal of the corresponding transmitter.

[0012] Furthermore, when calculating the position and velocity of the target object, it specifically includes:

[0013] Each bistatic radar group calculates the three-dimensional coordinates of the target based on the time difference between the receiver's received signal and the transmitter's transmitted signal. At the same time, each bistatic radar group calculates the target's velocity based on the frequency change between the receiver's received signal and the corresponding transmitter's transmitted signal, as well as the target's three-dimensional coordinates.

[0014] The three-dimensional coordinate points of the target object calculated by the two sets of bistatic radars are connected into a line segment, the middle coordinate point of the line segment is taken as the final coordinate point of the target object, and the average of the velocity values ​​of the target object calculated by the two sets of bistatic radars is taken as the final velocity of the target object.

[0015] Furthermore, before the target is recognized by the visible light recognition unit closest to the target, the method further includes: converting the three-dimensional coordinates of the target position into a two-dimensional image plane, as expressed as follows:

[0016]

[0017] Among them, u and v are two-dimensional image coordinates, u0 and v0 are the offsets of the camera optical axis, and f x and f y are the image width and height transformation ratios, R 3×3 is the rotation matrix, T 3×1 is the translation matrix, I 1×1 is the identity matrix, 0 1×3 is a zero matrix, X r 、Y r 、Z rare the X-axis coordinate, Y-axis coordinate, and Z-axis coordinate of the target object’s position, respectively.

[0018] Furthermore, when the target object is identified by the visible light recognition unit closest to the target object, the process specifically includes:

[0019] Selecting a visible light recognition unit closest to the target object, controlling the visible light recognition unit to face the target object, and then capturing images at specified time intervals;

[0020] The image captured by the visible light recognition unit is obtained and sharpened, and then the size of the sharpened image is unified and input into the detection network of the target detection model; the detection network performs predictions at different scales to obtain feature maps of different sizes to detect targets of different sizes; each feature map is divided into a corresponding number of grids according to different sizes, and each grid predicts a specified number of bounding boxes, each of which includes coordinate information of the bounding box, a confidence level of the bounding box, and a probability that the target in the bounding box is a bird; the confidence value of the bounding box is calculated, and invalid bounding boxes are filtered out through non-maximum suppression to obtain the bounding box with the highest confidence level to obtain a recognition result.

[0021] Furthermore, the formula for sharpening is:

[0022]

[0023] Among them, x is the grayscale of the center pixel in the neighborhood of the image, y is the average grayscale of other pixels in the neighborhood of the image, the function f(x, y) is the Laplace operator, and the function g(x, y) is the Laplace formula after image enhancement.

[0024] Furthermore, before the target object is identified by the visible light recognition unit closest to the target object, the method further includes:

[0025] Obtain images captured by all visible light recognition units within a specified time period, where the probability of the recognition result being a bird is greater than a preset threshold, add data labels to the images and add them to a bird atlas database, and use the bird atlas database to train the target detection model.

[0026] The present invention also provides a substation bird protection system, comprising:

[0027] The radar detection unit is used to detect the target object in real time. If the target object is detected, the position and speed of the target object are calculated;

[0028] a visible light recognition unit for identifying a target object based on its position, specifically, a position toward the target object, and then capturing images at specified time intervals, acquiring the captured images and performing sharpening processing, and then resizing the sharpened images and inputting them into a detection network of a target detection model to obtain a recognition result;

[0029] The substation bird pest multi-dimensional prevention and control unit is used to enable a fixed bird-repellent device to drive away birds when the target object is identified as a bird. It is also used to alarm and enable a patrol robot to drive away birds when the target object is identified as a bird and the speed of the target object is zero. It is also used to shut down the fixed bird-repellent device or cancel the alarm and recycle the patrol robot when no target object is detected or the target object is not identified as a bird.

[0030] Furthermore, it also includes:

[0031] The radar detection and visible light recognition fusion unit is used to convert the three-dimensional coordinates of the target's position into a two-dimensional image plane;

[0032] A data storage unit is used to obtain images captured by all visible light recognition units within a specified time period, wherein the probability of the images being identified as birds is greater than a preset threshold, and to add data labels to the images and then add them to the bird atlas database;

[0033] A data processing unit is used to train a target detection model using the bird atlas database.

[0034] Compared with the prior art, the advantages of the present invention are:

[0035] The present invention utilizes a combination of a radar detection unit and a visible light recognition unit. This integrated detection and recognition approach provides more accurate target detection and recognition, especially in dynamic environments. The radar detection unit detects the target's position and velocity in real time, while the visible light recognition unit accurately identifies the target after the radar determines its position.

[0036] The present invention takes immediate action upon bird recognition, activating fixed bird repellent devices or alarms and enabling patrol robots to repel the birds. This immediate response mechanism improves bird repellent efficiency. Furthermore, the system possesses self-monitoring and alarm-disarming capabilities. If the radar fails to detect a target or the target is identified as a different bird, the system automatically deactivates the fixed bird repellent device or deactivates the alarm and retracts the patrol robot. This adaptive mechanism reduces unnecessary bird repellent actions and enhances the system's intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of the structure of a substation bird prevention system according to an embodiment of the present invention.

[0038] Figure 2 Flowchart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The present invention will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the scope of protection of the present invention is not limited thereby.

[0040] Example 1

[0041] This embodiment proposes a substation bird prevention method based on radar detection and visible light recognition. When applied to the substation bird prevention system, it can more accurately locate and identify bird activities in the substation, formulate intelligent start and stop strategies for fixed bird repellent devices, and ultimately achieve the effect of improving the quality and effectiveness of substation bird prevention.

[0042] like Figure 1 As shown, the substation bird prevention system of this embodiment includes a radar detection unit, a visible light recognition unit, a radar detection and visible light recognition fusion unit, a data storage unit, a data processing unit, and a substation bird damage multi-dimensional prevention and control unit. The functional modules are described as follows:

[0043] Radar detection unit: Based on millimeter-wave radar detection technology, it can obtain the three-dimensional spatial coordinates and speed of bird activities in the substation in real time;

[0044] Visible light recognition unit: Based on the YOLOv3 target detection algorithm, it can sharpen and identify bird images in the radar positioning area;

[0045] Radar detection and visible light recognition fusion unit: Based on coordinate fusion technology, it can convert the three-dimensional coordinates of bird point clouds detected by radar into a two-dimensional image plane, improving the accuracy of visible light target recognition in different environments;

[0046] Data storage unit: mainly used to record and store bird activity data and bird maps of the substation;

[0047] Data processing unit: Based on deep learning algorithms, it uses stored bird atlases to continuously train bird target image recognition models to improve target detection efficiency;

[0048] Substation bird pest multi-dimensional prevention and control unit: Based on the substation bird activity data, it intelligently starts and stops the fixed bird-repellent devices and bird-repellent robots in the substation.

[0049] like Figure 2 As shown, based on the substation bird prevention system of this embodiment, the method of this embodiment includes the following steps:

[0050] S1) detecting the target object in real time by using a radar detection unit;

[0051] S2) If no target object is detected, the substation bird pest multi-dimensional prevention and control unit maintains the alarm disabled and shuts down the fixed bird repellent device or patrol robot. If a target object is detected, the radar detection unit calculates the target object's position and speed, and then the visible light recognition unit closest to the target object identifies the target object;

[0052] S3) If the identification result of the target object is not a bird, the substation bird damage multi-dimensional prevention and control unit maintains the alarm state while turning off the fixed bird repellent device or the inspection robot;

[0053] If the target object is identified as a bird, the substation bird pest multi-dimensional prevention and control unit will enable the fixed bird repellent device to drive away the bird. At the same time, the radar detection unit will continue to detect the target object in real time until the target object is no longer detected or the target object is detected but the identification result of the target object is not a bird.

[0054] If the identification result of the target object is a bird and the speed of the target object is zero, the substation bird damage multi-dimensional prevention and control unit will alarm and enable the inspection robot to expel the bird. At the same time, the radar detection unit will continue to detect the target object in real time until the target object is no longer detected or the target object is detected but the identification result of the target object is not a bird.

[0055] The following describes each step in detail in conjunction with relevant functional modules.

[0056] In this embodiment, the radar detection unit includes four radars respectively arranged at the four corners of the substation, and two radars on the same diagonal line cooperate with each other to form a group of bistatic radars. Therefore, the radar detection unit includes two groups of bistatic radars.

[0057] In a bistatic radar, one radar acts as a transmitter to transmit signals, and the other radar acts as a receiver to receive signals. In step S1 of this embodiment, when detecting a target object in real time using the radar detection unit, specifically, the signals transmitted by the transmitter and the signals received by the receiver in each group of bistatic radars are obtained in real time, and the change value of the received signal of each receiver is calculated. If the change value corresponding to at least one group of bistatic radars is greater than a preset value, the target object is detected.

[0058] In this embodiment, the change value of the received signal of each receiver specifically refers to the signal strength change value of the received signal of the receiver compared with the transmission signal of the corresponding transmitter, or the time difference between the received signal of the receiver and the transmission signal of the corresponding transmitter, or the frequency change value of the received signal of the receiver compared with the transmission signal of the corresponding transmitter. Specifically:

[0059] Regarding signal strength variations, when a transmitted signal encounters an object, a portion of the signal is reflected back to the receiver. Due to the presence of the object, the reflected signal is typically weaker than the directly transmitted signal. The receiver can then detect the presence of the object by detecting the change in signal strength.

[0060] The time difference of a signal is a time delay between reaching a target and returning to the receiver. This delay is the round-trip time it takes for the signal to travel from the transmitter to the target and then to the receiver. By accurately measuring this time difference, the distance between the target and the transmitter and receiver can be calculated.

[0061] Regarding signal frequency variations, if the target is moving, the received signal frequency will change, which is called the Doppler effect. By analyzing the frequency changes, the speed and direction of movement of the target can be determined.

[0062] In summary, the presence of the target object can be detected through the above signal changes.

[0063] In step S2 of this embodiment, when calculating the position and velocity of the target, the target is located by combining the time it takes for the target's reflected signal to reach the radars at different bases. Simultaneously, the target's velocity is obtained by analyzing the difference frequency between the transmitted and received signals. Specifically, this includes:

[0064] S101) Each group of bistatic radars calculates the three-dimensional coordinates of the target object based on the time difference between the received signal of the receiver and the transmitted signal of the transmitter;

[0065] Specifically, for a set of bistatic radars, transmitter A is located at known coordinates (x A ,y A ,z A ), receiver B is located at the known coordinates (x B ,y B ,z B ), when a target is detected, the signal from transmitter A reflects off the target and is received by receiver B. The total time from when the measurement signal is sent from transmitter A to when the reflected signal is received by receiver B is t. Using the signal propagation speed v (e.g., the speed of light), the total distance of the signal path is calculated as: d = v * t.

[0066] Assuming the three-dimensional coordinates of the target object are (x, y, z), establish the equation related to the distance:

[0067]

[0068] Among them, x A 、y A 、z AThey represent the X-axis, Y-axis, and Z-axis coordinates of the transmitter A position, respectively. B 、y B 、z B They represent the X-axis, Y-axis, and Z-axis coordinates of the position of receiver B, respectively. The equation is a nonlinear equation and can be solved by numerical methods (x, y, z). In this embodiment, the signal receiving angle of the receiver and the intervals of the X-axis, Y-axis, and Z-axis coordinates of the target position are added as constraints. Receiver B records the angle of signal incidence, which may include the azimuth angle θ (the angle on the horizontal plane) and the pitch angle φ (the angle on the vertical plane).

[0069] The direction vector of the received signal can be expressed as v = (cosφ·cosθ,cosφ·sinθ,sinφ), and the vector between the target and the receiver is (xx B ,yy B ,zz B ), this vector should be in the same direction as the direction vector v of the received signal, which can be represented by the dot product: (xx B )cosφ·cosθ+(yy B )cosφ·sinθ+(zz B )sinφ=||xx B ,yy B ,zz B ||.

[0070] The constraints are combined with the aforementioned path equation for numerical calculation. The angle information can help eliminate certain uncertainties, so that the equation can be solved uniquely.

[0071] S102) Each bistatic radar group calculates the speed of the target object based on a frequency change between a received signal of the receiver and a transmitted signal of the corresponding transmitter, and the three-dimensional coordinates of the target object;

[0072] Specifically, through step S101, the three-dimensional coordinates of the target object's position can be obtained as (x, y, z), and the Doppler shift measured by receiver B is Δf. The relationship between the Doppler shift and the target radial velocity is:

[0073]

[0074] Among them, v TA is the radial velocity component of the target relative to the transmitter A, v TB is the radial velocity component of the target relative to the receiver B, f0 is the frequency of the transmitted signal, and c is the speed of light.

[0075] Since the position of the target is known, the radial velocity component can be determined through geometric relationships. The direction vector of the target relative to the transmitter A is (xx A ,yy A ,zz A ), the direction vector of the target relative to the receiver B is (xx B ,yy B ,zz B ).

[0076] Assume that the velocity vector of the target is v = (v x ,v y ,v z ),but:

[0077]

[0078] Substituting the above radial velocity formula into the Doppler shift formula, we get:

[0079]

[0080] The Doppler frequency shift Δf is measured by the receiver B, and assuming that the target object moves at a uniform speed as a constraint, the velocity vector v of the target object can be solved by numerical methods. x ,v y ,v z ).

[0081] In another method, the three-dimensional coordinates (x, y, z) of the target object's position at the current moment and the next moment obtained from step S101 can be used to calculate the displacement vector of the target object within one clock interval, and the displacement vector is divided by the clock interval to obtain the velocity vector v of the target object = (v x ,v y ,v z ).

[0082] S103) After positioning and measuring the target object's velocity based on the arrival time of each bistatic radar set, the target object's final three-dimensional coordinates and velocity are averaged from the measurements of the two bistatic radar sets. Specifically, the target object's three-dimensional coordinates calculated by the two bistatic radar sets are connected to form a line segment, the middle coordinate point of the line segment is taken as the target object's final coordinate point, and the target object's final velocity is averaged from the velocity values ​​calculated by the two bistatic radar sets.

[0083] In step S2 of this embodiment, before the visible light recognition unit closest to the target object recognizes the target object, the method further includes: converting the three-dimensional coordinates of the target object's position to a two-dimensional image plane by a radar detection and visible light recognition fusion unit. The radar detection and visible light recognition fusion unit uses the intersection of the two diagonals of the substation horizontal plane as the two-dimensional coordinate origin and the three-dimensional coordinate origin, and converts the three-dimensional coordinates of the point cloud obtained by the radar detection unit to the two-dimensional image plane through a coordinate fusion technology. The expression is as follows:

[0084]

[0085] Among them, u and v are two-dimensional image coordinates, u0 and v0 are the offsets of the camera optical axis, and f x and f y are the image width and height transformation ratios, R 3×3 is the rotation matrix, T 3×1 is the translation matrix, I 1×1 is the identity matrix, 0 1×3 is a zero matrix, X r 、Y r , Z r are the X-axis coordinate, Y-axis coordinate, and Z-axis coordinate of the target object’s position, respectively.

[0086] In step S2 of this embodiment, when the target object is identified by the visible light recognition unit closest to the target object, the following steps are specifically performed:

[0087] S201) selecting a visible light recognition unit closest to the target object;

[0088] In this embodiment, the visible light recognition unit includes all industrial video cameras installed outdoors in the substation. These industrial video cameras are pre-marked with corresponding coordinates. Therefore, in step S2 of this embodiment, before the target object is identified by the visible light recognition unit closest to the target object, the real-time distance between each visible light recognition unit and the target object is calculated according to the coordinates of each visible light recognition unit and the coordinates of the target object's position, and the visible light recognition unit with the smallest distance is selected as the visible light recognition unit closest to the target object.

[0089] S202) controlling the position of the visible light recognition unit toward the target object, and then capturing images at specified time intervals;

[0090] In this embodiment, based on the two-dimensional coordinates obtained by the radar detection and visible light recognition fusion unit, the camera closest to the two-dimensional coordinate point is controlled to shoot an image toward the coordinate point every 10 minutes.

[0091] S203) The image captured by the visible light recognition unit is obtained and sharpened. In this embodiment, the Laplace algorithm is used for sharpening, and the formula is as follows:

[0092]

[0093] Among them, x is the grayscale of the center pixel in the neighborhood of the image, y is the average grayscale of other pixels in the neighborhood of the image, the function f(x, y) is the Laplace operator, and the function g(x, y) is the Laplace formula after image enhancement.

[0094] S204) unifying the size of the sharpened image and inputting it into the detection network of the target detection model to obtain a recognition result;

[0095] In this embodiment, the target detection model uses the YOLOv3 image recognition model. After the sharpened pre-processed image is scaled to 416×416 and input into its detection network, the recognition process is as follows:

[0096] The detection network makes predictions at three different scales, obtaining feature maps of three different sizes: 13×13, 26×26, and 52×52, to detect objects of different sizes.

[0097] Each feature map is divided into a corresponding number of grids according to different sizes. The specific number is: the 13×13 feature map is divided into 169 grids; the 26×26 feature map is divided into 676 grids; the 52×52 feature map is divided into 2704 grids;

[0098] Each grid predicts a specified number of bounding boxes, which in this embodiment is 3. Each bounding box includes: 4 bounding box coordinate information, which in this embodiment is the center coordinates, width and height of the target; 1 bounding box confidence, which indicates the probability that the target exists in the grid; 1 bird information, specifically the probability that the target in the bounding box is a bird;

[0099] The confidence value of the frame is calculated, invalid frames are filtered out through non-maximum suppression, and the frame with the highest confidence is obtained as the target frame. The recognition result is obtained based on the bird information of the target frame. If the probability value of the bird information is greater than a preset threshold, the target object is recognized as a bird; otherwise, it is not a bird. If there is a target object in the image that is recognized as a bird, the image is marked as a bird atlas and uploaded to the data storage unit together with the corresponding target object position and speed data.

[0100] In this embodiment, the data recorded by the data storage unit specifically includes: statistical data on bird nests in substations, statistical data on the locations of bird activities in substations, duration data on bird activities in substations, frequency data on bird activities in substations, and species of birds active in substations, and the bird atlas recorded by the data storage unit forms a bird atlas database.

[0101] Specifically, substation bird nest statistics include the time the nest was discovered, the name of the substation where the nest is located, and the name of the equipment where the nest is located. Substation bird activity location statistics include the name of the substation involved in the bird activity and the name of the equipment involved in the bird activity. Substation bird activity duration data includes the total duration of birds' stay at the substation, that is, the total duration of time the target object was detected and the target was a bird. Substation bird activity frequency data includes the total number of substation bird activities in different months, quarters, and years, that is, the total number of times the target object was detected and the target was a bird. Substation bird activity species include data such as the bird name.

[0102] In this embodiment, the bird atlas database contains all bird atlases taken since the system was put into operation, and is updated every quarter and uploaded to the data processing unit.

[0103] The data processing unit uses the bird atlas database updated quarterly to regularly conduct image recognition model training for the visible light recognition unit.

[0104] Therefore, in step S2 of this embodiment, before the target object is recognized by the visible light recognition unit closest to the target object, the following steps are further included:

[0105] The data storage unit obtains images captured by all visible light recognition units within a specified time period, where the probability of the images being identified as birds is greater than a preset threshold. The images are then labeled with data and added to a bird atlas database. The data processing unit then uses the bird atlas database to train a target detection model. Model training specifically involves deep learning training of the YOLOv3 image recognition model for bird target image recognition using the substation bird atlas recorded in the data storage unit. The trained and updated image recognition model is then synchronously updated in the visible light recognition unit. The convolutional neural network of the YOLOv3 image recognition model includes an input layer, a convolution layer, a pooling layer, an activation layer, a fully connected layer, and an output layer.

[0106] In this embodiment, the substation bird pest multi-dimensional prevention and control unit uses the monitored bird activity data to issue real-time alarms, intelligently start and stop the substation's fixed bird repellent device, and automatically navigate and recover the inspection robot. In step S3 of this embodiment, the substation bird pest multi-dimensional prevention and control unit issues an alarm and enables the fixed bird repellent device or inspection robot to repel birds, and the substation bird pest multi-dimensional prevention and control unit cancels the alarm and shuts down the fixed bird repellent device or inspection robot, specifically including:

[0107] When the visible light recognition unit detects bird activity and the radar detection unit detects the bird's speed as zero, the substation bird pest multi-dimensional prevention and control unit automatically sends a real-time message to the substation control center, alerting operators and maintenance personnel of bird intrusion and presence. When the visible light recognition unit no longer detects bird activity, the multi-dimensional prevention and control unit automatically sends a clear alarm message to the substation control center, alerting operators and maintenance personnel that bird activity has ceased.

[0108] When the visible light recognition unit detects bird activity, the substation's multi-dimensional bird pest control unit automatically activates the substation's fixed bird repellent devices, which include acoustic and ultrasonic devices. When the visible light recognition unit no longer detects bird activity, the multi-dimensional bird pest control unit automatically deactivates the fixed bird repellent devices.

[0109] When the visible light recognition unit detects bird activity and the radar detection unit detects the bird's speed as zero, the inspection robot automatically navigates to the two-dimensional coordinates of the substation bird activity, obtained by the radar detection and visible light recognition fusion unit. The robot then activates its built-in laser bird repellent to repel the bird. When the visible light recognition unit no longer detects bird activity, the inspection robot automatically navigates back to its original location.

[0110] Example 2

[0111] This embodiment provides a substation bird prevention system, including:

[0112] The radar detection unit is used to detect the target object in real time. If the target object is detected, the position and speed of the target object are calculated;

[0113] a visible light recognition unit for identifying a target object based on its position, specifically, a position toward the target object, and then capturing images at specified time intervals, acquiring the captured images and performing sharpening processing, and then resizing the sharpened images and inputting them into a detection network of a target detection model to obtain a recognition result;

[0114] The radar detection and visible light recognition fusion unit is used to convert the three-dimensional coordinates of the target's position into a two-dimensional image plane;

[0115] A data storage unit is used to obtain images captured by all visible light recognition units within a specified time period, wherein the probability of the images being identified as birds is greater than a preset threshold, and to add data labels to the images and then add them to the bird atlas database;

[0116] a data processing unit, configured to train a target detection model using the bird atlas database;

[0117] The substation bird pest multi-dimensional prevention and control unit is used to enable a fixed bird-repellent device to drive away birds when the target object is identified as a bird. It is also used to alarm and enable a patrol robot to drive away birds when the target object is identified as a bird and the speed of the target object is zero. It is also used to shut down the fixed bird-repellent device or cancel the alarm and recycle the patrol robot when no target object is detected or the target object is not identified as a bird.

[0118] In summary, the present invention proposes a substation bird prevention method and system based on radar detection and visible light recognition. Based on millimeter-wave radar detection technology, the three-dimensional spatial coordinates and speed of bird activities in the substation are obtained in real time; based on coordinate fusion technology, the three-dimensional coordinates detected by the radar can be converted to a two-dimensional image plane, improving the accuracy of visible light target recognition in different environments; visible light recognition is based on the YOLOv3 target detection algorithm, which can perform sharp image recognition of bird information in the radar positioning area; based on the image recognition results, the substation bird activity data and bird atlas are recorded and stored, and based on a deep learning algorithm, the stored bird atlas is regularly used to continuously train the bird target image recognition model to improve target detection efficiency; finally, based on the substation bird activity data and bird image recognition results, the substation fixed bird repellent device and bird repellent robot are intelligently started and stopped. The present invention can meet the accuracy of substation bird target recognition in different environments, reasonably formulate the intelligent start-stop strategy of the substation bird repellent device, and overcome the defect that traditional bird repellents are easily adapted by birds.

[0119] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A bird prevention method for substations based on radar detection and visible light recognition, characterized in that: Applied to a substation bird prevention system comprising a radar detection unit and a visible light recognition unit, the method comprises the following steps: The radar detection unit detects the target in real time. If a target is detected, the position and speed of the target are calculated, and then the visible light recognition unit closest to the target is used to identify the target. If the target object is identified as a bird, the fixed bird repellent device is activated to drive away the bird, and the radar detection unit continues to detect the target object in real time until the target object is no longer detected or the target object is not identified as a bird; If the target object is identified as a bird and the speed of the target object is zero, an alarm is sounded and the inspection robot is activated to drive away the bird, and the radar detection unit continues to detect the target object in real time until the target object is no longer detected or the identification result of the target object is not a bird; If the target object cannot be detected or the identification result of the target object is not a bird, the fixed bird repellent device will be turned off or the alarm will be released and the inspection robot will be recovered; The fixed bird-repelling device includes a sound bird-repelling device, and the inspection robot is equipped with a laser bird-repelling device.

2. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 1 is characterized in that: The radar detection unit includes four radars respectively arranged at the four corners of the substation, and two radars on the same diagonal line cooperate with each other to form a group of bistatic radars. One radar in the bistatic radar acts as a transmitter to transmit signals, and the other radar acts as a receiver to receive signals. When the radar detection unit detects the target object in real time, it specifically obtains the signal transmitted by the transmitter and the signal received by the receiver in each group of bistatic radars in real time, and calculates the change value of the received signal of each receiver. If the change value corresponding to at least one group of bistatic radars is greater than a preset value, the target object is detected.

3. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 2 is characterized in that: The change value of the received signal of each receiver specifically refers to the signal strength change value of the received signal of the receiver compared with the transmission signal of the corresponding transmitter, or the time difference between the received signal of the receiver and the transmission signal of the corresponding transmitter, or the frequency change value of the received signal of the receiver compared with the transmission signal of the corresponding transmitter.

4. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 2, characterized in that: Calculating the position and velocity of a target object includes: Each bistatic radar group calculates the three-dimensional coordinates of the target based on the time difference between the receiver's received signal and the transmitter's transmitted signal. At the same time, each bistatic radar group calculates the target's velocity based on the frequency change between the receiver's received signal and the corresponding transmitter's transmitted signal, as well as the target's three-dimensional coordinates. The three-dimensional coordinate points of the target object calculated by the two sets of bistatic radars are connected into a line segment, the middle coordinate point of the line segment is taken as the final coordinate point of the target object, and the average of the velocity values ​​of the target object calculated by the two sets of bistatic radars is taken as the final velocity of the target object.

5. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 1, characterized in that: Before the target object is recognized by the visible light recognition unit closest to the target object, the method further includes: converting the three-dimensional coordinates of the target object's position into a two-dimensional image plane, as expressed as follows: Among them, u and v are two-dimensional image coordinates, u0 and v0 are the offsets of the camera optical axis, and f x and f y are the image width and height transformation ratios, R 3×3 is the rotation matrix, T 3×1 is the translation matrix, I 1×1 is the identity matrix, 0 1×3 is a zero matrix, X r 、Y r 、Z r are the X-axis coordinate, Y-axis coordinate, and Z-axis coordinate of the target object’s position, respectively.

6. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 1, characterized in that: When identifying a target object through the visible light recognition unit closest to the target object, the following steps are specifically performed: Selecting a visible light recognition unit closest to the target object, controlling the visible light recognition unit to face the target object, and then capturing images at specified time intervals; The image captured by the visible light recognition unit is obtained and sharpened, and then the size of the sharpened image is unified and input into the detection network of the target detection model; the detection network performs predictions at different scales to obtain feature maps of different sizes to detect targets of different sizes; each feature map is divided into a corresponding number of grids according to different sizes, and each grid predicts a specified number of bounding boxes, each of which includes coordinate information of the bounding box, a confidence level of the bounding box, and a probability that the target in the bounding box is a bird; the confidence value of the bounding box is calculated, and invalid bounding boxes are filtered out through non-maximum suppression to obtain the bounding box with the highest confidence level to obtain a recognition result.

7. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 6, characterized in that: The formula for sharpening is: Among them, x is the grayscale of the center pixel in the neighborhood of the image, y is the average grayscale of other pixels in the neighborhood of the image, the function f(x, y) is the Laplace operator, and the function g(x, y) is the Laplace formula after image enhancement.

8. The method for preventing birds from substations based on radar detection and visible light recognition according to claim 6, characterized in that: Before the target is recognized by the visible light recognition unit closest to the target, the method further includes: Obtain images captured by all visible light recognition units within a specified time period, where the probability of the recognition result being a bird is greater than a preset threshold, add data labels to the images and add them to a bird atlas database, and use the bird atlas database to train the target detection model.

9. A substation bird prevention system for implementing the substation bird prevention method based on radar detection and visible light recognition according to any one of claims 1 to 8, characterized in that: include: The radar detection unit is used to detect the target object in real time. If the target object is detected, the position and speed of the target object are calculated; a visible light recognition unit for identifying a target object based on its position, specifically, a position toward the target object, and then capturing images at specified time intervals, acquiring the captured images and performing sharpening processing, and then resizing the sharpened images and inputting them into a detection network of a target detection model to obtain a recognition result; The substation bird pest multi-dimensional prevention and control unit is used to activate the fixed bird-repelling device to drive away the birds when the target object is identified as a bird. It is also used to alarm and activate the inspection robot to drive away the birds when the target object is identified as a bird and the speed of the target object is zero. It is also used to shut down the fixed bird-repelling device or cancel the alarm and recycle the inspection robot when the target object cannot be detected or the identification result of the target object is not a bird.

10. The substation bird protection system according to claim 9, characterized in that: Also includes: The radar detection and visible light recognition fusion unit is used to convert the three-dimensional coordinates of the target's position into a two-dimensional image plane; A data storage unit is used to obtain images captured by all visible light recognition units within a specified time period, wherein the probability of the images being identified as birds is greater than a preset threshold, and to add data labels to the images and then add them to the bird atlas database; A data processing unit is used to train a target detection model using the bird atlas database.

Images