Radar-based bird and unmanned aerial vehicle track target classification method and system

Abstract

The invention discloses a radar-based bird and unmanned aerial vehicle track target classification method and system, and relates to the technical field of flight target detection, and the method comprises the following steps: positioning a target position based on radar detection, and obtaining target track information based on the target position, the target track information being a target position of a target at each moment; according to the radar-based bird and unmanned aerial vehicle track target classification method and system, an unmanned aerial vehicle and a suspected bird target can be preliminarily screened out through the step S2, and whether the suspected bird target is a bird or not is judged again through the step S3 on the basis of analysis of echo signal strength, phase change characteristics and the like. The judgment accuracy of the crossing machine, the flapping-wing unmanned aerial vehicle and the flying bird is improved, and whether the target position of the flying bird emits electromagnetic waves or not can be judged through verification in the step S4, so that whether the target is the unmanned aerial vehicle or not is verified, and the judgment accuracy is further guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of flight target detection, and particularly relates to a method and system for classifying flight track targets of birds and unmanned aerial vehicles (UAVs) based on radar. Background Art

[0002] With the development of UAV technology, there are more and more types and models of UAVs on the market, such as aerial photography UAVs, racing drones, flapping-wing UAVs, etc., which to a certain extent increase the difficulty of UAV supervision.

[0003] The prior art with the publication number CN118884427A discloses a method for classifying rotary-wing UAVs and birds based on multi-domain feature and classifier fusion, including: first, using radar to observe rotary-wing UAVs and birds to obtain corresponding radar echo signals; then preprocessing the data to obtain the spectrum, time-frequency spectrum, and cepstrum of the echo signals; then performing multi-domain feature extraction, extracting features in the time domain, frequency domain, time-frequency domain, and cepstrum domain respectively; finally, based on Stacking, fusing multiple classifiers, inputting the multi-domain features into K-nearest neighbor, support vector machine, XGBoost model, and random forest respectively for combination and learning, and fusing the classification results of the four base classifiers as the input of the next-layer meta-classifier to obtain the final classification result. This prior art extracts various features in different domains, ensuring the comprehensiveness and diversity of the features, and uses Stacking to fuse multiple classifiers, making full use of the advantages of different classifiers and improving the accuracy of classifying rotary-wing UAVs and bird targets.

[0004] However, because racing drones among UAVs are small in size, fast in speed, and flexible in turning, and flapping-wing UAVs imitate the flight of birds by flapping their wings in terms of flight principle, which further increases the difficulty of recognition to a certain extent. Therefore, it is necessary to improve the recognition accuracy of UAVs such as racing drones and flapping-wing UAVs and birds, so that appropriate measures can be taken for flight targets according to the types of flight targets in special places. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for classifying flight track targets of birds and UAVs based on radar to solve the above deficiencies in the prior art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A method for classifying flight track targets of birds and UAVs based on radar, including the following steps: S1. Locate the target position based on radar detection, and obtain the target trajectory information based on the target position. The target trajectory information is the target position at each moment. Among them, the target velocity and target acceleration at each moment of the target can be calculated based on the target position at each moment of the target. The target position can be represented by (x, y, z), and the target velocity and target acceleration can be decomposed into three directions of the x-axis, y-axis, and z-axis. The x-axis, y-axis, and z-axis are perpendicular to each other in pairs; S2. Obtain the historical target trajectory information, set the first bird judgment label to be associated with the corresponding historical target trajectory information, and train the first machine learning model based on the first bird judgment label and the historical target trajectory information to obtain the first bird prediction model. The first bird prediction model is used to output the predicted first bird judgment label according to the input target trajectory information. The first bird judgment label includes a drone label and a suspected bird label; S3. Collect the echoes of various types of birds on the electromagnetic waves of each frequency emitted by the radar, extract the echo features, and set multiple second bird judgment labels. Associate the echo features with the corresponding second bird judgment labels, and train the second machine learning model based on the echo, echo features, and second bird judgment labels to obtain the second bird prediction model. The second bird prediction model is used to extract the echo features according to the input echo and output the second bird judgment label corresponding to the extracted echo features. The second bird judgment label includes a drone label and a bird label. Further, the bird label can include multiple bird category labels, which can indicate which specific bird the target belongs to; S4. Extract the electromagnetic waves other than the echo from the electromagnetic waves received by the radar to obtain interference electromagnetic waves, and use the time difference method to locate the target that emits the interference electromagnetic waves based on the time when each radar receives the interference electromagnetic waves to obtain the interference source position; S5. Analyze the electromagnetic waves received by the radar, extract the echo of the electromagnetic waves emitted by the radar, obtain the target trajectory information of the target corresponding to the echo, input the target trajectory information into the first bird prediction model. If the first bird prediction model outputs a suspected bird label, then input the echo into the second bird prediction model. If the second bird prediction model outputs a bird label, then judge whether there is an interference source position that coincides with the target position of the target. If so, judge that the target is a drone. If not, judge that the target is a bird.

[0007] Further, the S2 includes the following steps: Obtain the historical trajectory information and set the first bird judgment label. The first bird judgment label includes a drone label and a suspected bird label; Screen out the historical trajectory information of the determined drones from the historical trajectory information and associate it with the drone label; After filtering out the historical trajectory information of the determined drone, associate the historical trajectory information with the suspected bird tag; Select the first machine learning model, use the historical trajectory information as the input of the first machine learning model, extract the features of the historical trajectory information, associate the first bird judgment tag associated with the historical trajectory information with the features of the historical trajectory information, and use the first bird judgment tag as the output of the first machine learning model. Train the first machine learning model to obtain the first bird prediction model. Among them, the features of the historical trajectory information of the determined drone should be significantly different from the features of the bird's trajectory information, reducing the possibility of misidentification, so that the first bird prediction model will only identify the target whose trajectory information conforms to the features of the historical trajectory information of the determined drone as a drone, improving the accuracy of identifying drones, and the targets identified as drones can be separated first, reducing the workload of subsequent identification and classification work.

[0008] Further, the S3 includes the following steps: Obtain the electromagnetic waves emitted by the radar when detecting various types of birds at various distances from the radar and the corresponding received echoes; Based on the emission time of the electromagnetic wave and the reception time of the echo corresponding to the electromagnetic wave, calculate the distance between the target and the radar when the electromagnetic wave touches the target and generates an echo, and obtain the target distance; Obtain the signal intensity of the electromagnetic wave and the corresponding echo, and calculate the signal intensity at which the echo decays relative to the corresponding electromagnetic wave to obtain the signal attenuation intensity; Classify the signal attenuation intensity and the target distance based on the type of bird to obtain multiple bird groups, and then classify the signal attenuation intensity and the target distance based on the frequency of the electromagnetic wave and the corresponding echo in each bird group to obtain multiple frequency groups; Conduct a regression analysis on the relationship between the signal attenuation intensity and the corresponding target distance of each frequency group to obtain a relationship model between the signal attenuation intensity and the target distance, and obtain the error of the relationship model between the signal attenuation intensity and the target distance to obtain the first error threshold. The relationship model between the signal attenuation intensity and the target distance is used to output a predicted value of the signal attenuation intensity based on the input target distance; Set a judgment formula, which is used to judge whether the absolute value of the difference between the predicted value of the signal attenuation intensity corresponding to the same target distance in each frequency group and the signal attenuation intensity is less than the first error threshold, and associate the judgment result with the corresponding second bird judgment tag. Specifically, associate the result of yes with the bird machine tag, and associate the result of no with the drone tag; Integrate all the signal attenuation intensity and target distance relationship models and judgment formulas to obtain a second bird prediction model. The second bird prediction model is used to, based on the input echo, obtain the frequency of the echo, retrieve the corresponding electromagnetic wave, calculate the target distance and the signal attenuation intensity, input the calculated target distance into the signal attenuation intensity and target distance relationship model of the frequency group corresponding to the frequency of the echo to obtain a predicted value of the signal attenuation intensity, input the predicted value of the signal attenuation intensity and the signal attenuation intensity into the judgment formula of the frequency group, and judge whether the target is a bird based on the second bird judgment label output by the judgment formula. If the judgment result of the judgment formula is yes, output a bird label indicating that the target is a bird; if the judgment result of the judgment formula is no, output a drone label indicating that the target is a drone.

[0009] Further, the S3 includes the following steps: Obtain the electromagnetic waves emitted by the radar when detecting various types of birds at various distances from the radar and the corresponding received echoes; Based on the emission time of the electromagnetic wave and the reception time of the echo corresponding to the electromagnetic wave, calculate the distance between the target and the radar when the electromagnetic wave contacts the target to generate an echo, and obtain the target distance; Extract the phase change characteristics of the echo, and associate the phase change characteristics with the corresponding second bird judgment label. Specifically, associate the phase change characteristics when the target is a bird with the corresponding bird label, associate the phase change characteristics when the target is not a bird with the drone label. Further, associate the phase change characteristics when the target is a bird with the corresponding bird category label; Based on the echo, the phase change characteristics and the second bird judgment label, train a second machine learning model to obtain a second bird prediction model. The second bird prediction model is used to, based on the input echo, extract the phase change characteristics of the echo and output the second bird label associated with the phase change characteristics.

[0010] Further, the S3 includes the following steps: Simultaneously emit electromagnetic waves of multiple frequencies towards the target to obtain multiple corresponding echoes; Input the multiple echoes into the second bird prediction model to obtain multiple output second bird judgment labels; Judge whether all the second bird judgment labels indicate that the target is a bird; If so, the target is a bird; if not, the target is a drone.

[0011] Further, in the S4, the time difference method is used to locate the target that emits the interference electromagnetic wave to obtain the interference source position, including the following steps: Synchronize the clock times of multiple radars, and control the clock time error to be less than the set clock error threshold, where the set clock error threshold can be set according to the positioning accuracy. The clock error threshold for a positioning accuracy of 1 meter is 3 nanoseconds, and the clock error threshold for a positioning accuracy of centimeter level is 0.33 nanoseconds; Calculate the time difference of the interfering electromagnetic wave between every two radars; Express the time difference in the form of a distance difference, and obtain the distance difference formula of the interfering source position from every two radars, , where is the distance difference, c is the speed of light, is the time difference, where: , where (x, y, z) are the three-dimensional coordinates of the interfering source position, are the three-dimensional coordinates of one radar, are the three-dimensional coordinates of another radar, and the three-dimensional coordinates of the radars are known.

[0012] Integrate all the distance difference formulas to obtain a distance difference equation system; Solve based on the distance difference equation system to obtain the position of the interfering source.

[0013] A bird and UAV flight track target classification system based on radar, including a radar module, a signal processing module, a first judgment module, a second judgment module, and a third judgment module; There are multiple of the radar modules, which are used to transmit and receive electromagnetic waves. The received electromagnetic waves include echo waves and interfering electromagnetic waves; The signal processing module is used to process the received electromagnetic waves, separate the echo waves and the interfering electromagnetic waves, and calculate the target position corresponding to the echo waves and the interfering source position corresponding to the interfering electromagnetic waves; The first judgment module is used to train a first machine learning model based on historical target trajectory information and a first bird judgment label to obtain a first bird prediction model, and is also used to input the target trajectory information into the first bird prediction model to obtain an output predicted first bird judgment label, and judge whether the target is a bird based on the first bird judgment label; The second judgment module is used to train a second machine learning model based on the echo waves, echo wave features and a second bird judgment label to obtain a second bird prediction model, and is also used to input the echo waves into the second bird prediction model to obtain an output predicted second bird judgment label, and judge whether the target is a bird based on the second bird judgment label; The third judgment module is used to verify whether the judgment results of the first bird prediction model and the second bird prediction model are correct by judging whether there is an overlapping position between the interfering source position and the target position of the target determined to be a bird when the first bird prediction model and the second bird prediction model judge that the target is a bird.

[0014] 1. Compared with the prior art, a method and system for classifying flight path targets of birds and unmanned aerial vehicles (UAVs) based on radar provided by the present invention can preliminarily screen out targets of UAVs and suspected birds through step S2, and then, through step S3, based on the analysis of echo signal intensity, phase change characteristics, etc., determine again whether the target of the suspected bird is a bird, improving the judgment accuracy for cross-type UAVs, flapping-wing UAVs and birds.

[0015] 2. Compared with the prior art, a method and system for classifying flight path targets of birds and UAVs based on radar provided by the present invention can further verify whether the target is a UAV by verifying whether the target position determined to be a bird emits electromagnetic waves through step S4, further ensuring the accuracy of the judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0017] Figure 1 It is a flowchart of the method steps provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the system structure block diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the following will further introduce the present invention in detail with reference to the drawings.

[0019] In the description of the present invention, it should be understood that the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more unless otherwise specifically defined. In addition, the terms "installation", "connection" and "connection" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection or an integral connection; it may be a mechanical connection or an electrical connection; it may be a direct connection or an indirect connection through an intermediate medium, and it may be the internal connection of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0020] Example embodiments will be described more fully hereinafter with reference to the accompanying drawings, but the example embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0021] In the case of no conflict, the various embodiments of the present disclosure and the features in the embodiments may be combined with each other.

[0022] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0023] The terms used herein are only for the purpose of describing particular embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are also intended to include the plural forms unless the context clearly indicates otherwise.

[0024] Please refer to Figure 1 , a method for classifying the flight tracks of birds and drones based on radar, comprising the following steps: S1. Detect and locate the target position based on radar, and obtain the target trajectory information based on the target position. The target trajectory information is the target position at each moment of the target. Among them, the target speed and target acceleration at each moment of the target can be calculated based on the target position at each moment of the target. The target position can be represented by (x, y, z), and the target speed and target acceleration can be decomposed into three directions of the x-axis, y-axis, and z-axis, and the x-axis, y-axis, and z-axis are perpendicular to each other in pairs; S2. Obtain historical target trajectory information, set the first bird judgment label to be associated with the corresponding historical target trajectory information, and train a first machine learning model based on the first bird judgment label and the historical target trajectory information to obtain a first bird prediction model. The first bird prediction model is used to output a predicted first bird judgment label according to the input target trajectory information. The first bird judgment label includes a drone label and a suspected bird label. The present invention does not limit the specific machine learning model. For example, support vector machine (SVM), random forest, K-nearest neighbor algorithm (K-NN), convolutional neural network (CNN), recurrent neural network (RNN), and long short-term memory network (LSTM) can be selected, etc. During the training process, the data set (i.e., the first bird judgment label and the historical target trajectory information) is divided into a training set and a validation set. Using the historical target trajectory information as the model input, extracting the features of the historical target trajectory information as the output of the model hidden layer, and the first bird judgment label as the model output to obtain the parameters of the model. The performance of the model is evaluated through techniques such as cross-validation. The cross-entropy loss function can be used to measure the difference between the predicted category and the true label. Metrics such as accuracy, recall, and F1-score are used to evaluate the model performance. In the case of class imbalance, the F1-score is ensured first and then the accuracy is ensured if possible.

[0025] In one embodiment, S2 includes the following steps: (1) Obtain historical trajectory information and set the first bird judgment label, where the first bird judgment label includes a drone label and a suspected bird label; (2) Screen out the historical trajectory information of the determined drones from the historical trajectory information and associate it with the drone label; (3) Associate the historical trajectory information after screening out the historical trajectory information of the determined drones with the suspected bird label; (4) Select a first machine learning model, use the historical trajectory information as the input of the first machine learning model, extract the features of the historical trajectory information, associate the first bird judgment label associated with the historical trajectory information with the features of the historical trajectory information, and use the first bird judgment label as the output of the first machine learning model to train the first machine learning model to obtain a first bird prediction model. Among them, the features of the historical trajectory information of the determined drones should be significantly different from the features of the bird trajectory information to reduce the possibility of misidentification, so that the first bird prediction model will only identify the target whose trajectory information conforms to the features of the historical trajectory information of the determined drones as a drone, improving the accuracy of drone identification, and the targets identified as drones can be separated first to reduce the workload of subsequent identification and classification work.

[0026] S3. Collect the echoes of electromagnetic waves of various frequencies emitted by the radar from various types of flying birds, extract the echo features, set a variety of second flying bird judgment tags, associate the echo features with the corresponding second flying bird judgment tags, and train a second machine learning model based on the echoes, echo features, and second flying bird judgment tags to obtain a second flying bird prediction model. The second flying bird prediction model is used to extract the echo features according to the input echo and output the second flying bird judgment tag corresponding to the extracted echo features. The second flying bird judgment tag includes a drone tag and a flying bird tag. Further, the flying bird tag may include a variety of flying bird category tags, which can indicate which specific type of flying bird the target belongs to; In one embodiment, S3 includes the following steps: (1) Obtain the electromagnetic waves emitted by the radar when detecting various types of flying birds at various distances from the radar and the corresponding received echoes; (2) Based on the emission time of the electromagnetic wave and the reception time of the echo corresponding to the electromagnetic wave, calculate the distance between the target and the radar when the electromagnetic wave contacts the target to generate an echo, and obtain the target distance; (3) Obtain the signal intensities of the electromagnetic wave and the corresponding echo, and calculate the signal intensity by which the echo decays relative to the corresponding electromagnetic wave to obtain the signal attenuation intensity; (4) Classify the signal attenuation intensity and the target distance based on the types of flying birds to obtain multiple flying bird groups, and then classify the signal attenuation intensity and the target distance based on the frequencies of the electromagnetic wave and the corresponding echo in each flying bird group to obtain multiple frequency groups; (5) Perform a regression analysis on the relationship between the signal attenuation intensity and the corresponding target distance in each frequency group to obtain a relationship model between the signal attenuation intensity and the target distance, and obtain the error of the relationship model between the signal attenuation intensity and the target distance to obtain a first error threshold. The relationship model between the signal attenuation intensity and the target distance is used to output a predicted value of the signal attenuation intensity based on the input target distance; (6) Set a judgment formula, which is used to judge whether the absolute value of the difference between the predicted value of the signal attenuation intensity corresponding to the same target distance in each frequency group and the signal attenuation intensity is less than the first error threshold, and associate the judgment result with the corresponding second flying bird judgment tag. Specifically, associate the result of "yes" with the flying bird tag, and associate the result of "no" with the drone tag; (7) Integrate all the signal attenuation intensity - target distance relationship models and the judgment formula to obtain the second bird prediction model. The second bird prediction model is used to, based on the input echo, obtain the frequency of the echo, retrieve the corresponding electromagnetic wave, calculate the target distance and the signal attenuation intensity, input the calculated target distance into the signal attenuation intensity - target distance relationship model of the frequency group corresponding to the frequency of the echo to obtain the predicted value of the signal attenuation intensity, input the predicted value of the signal attenuation intensity and the signal attenuation intensity into the judgment formula of the input frequency group, and judge whether the target is a bird based on the second bird judgment label output by the judgment formula. If the judgment result of the judgment formula is yes, output the bird label indicating that the target is a bird; if the judgment result of the judgment formula is no, output the drone label indicating that the target is a drone.

[0027] Further, in this embodiment, S3 may further include the following steps: (1) Simultaneously transmit electromagnetic waves of multiple frequencies to the target to obtain multiple corresponding echoes; (2) Input the multiple echoes into the second bird prediction model to obtain multiple output second bird judgment labels; (3) Judge whether all the second bird judgment labels indicate that the target is a bird; (4) If so, the target is a bird; if not, the target is a drone. Through this embodiment, it can prevent drones from having a certain absorption effect on electromagnetic waves of certain frequencies emitted by the radar through technical means such as plating and coating, resulting in an error in prediction due to the similarity in intensity between the echoes of these frequencies at corresponding distances and the echoes reflected by birds.

[0028] In another embodiment, S3 includes the following steps: (1) Obtain the electromagnetic waves emitted by the radar when detecting various types of birds at various distances from the radar and the corresponding received echoes; (2) Based on the emission time of the electromagnetic wave and the reception time of the echo corresponding to the electromagnetic wave, calculate the distance between the target and the radar when the electromagnetic wave contacts the target to generate an echo, and obtain the target distance; (3) Extract the phase change characteristics of the echo, and associate the phase change characteristics with the corresponding second bird judgment labels. Specifically, associate the phase change characteristics when the target is a bird with the corresponding bird label, associate the phase change characteristics when the target is not a bird with the drone label. Further, associate the phase change characteristics when the target is a bird with the corresponding bird category label; (4) Based on the echo, the phase change characteristics, and the second bird judgment labels, train the second machine learning model to obtain the second bird prediction model. The second bird prediction model is used to, based on the input echo, extract the phase change characteristics of the echo and output the second bird label associated with the phase change characteristics.

[0029] S4. Extract the non-echo electromagnetic wave from the electromagnetic waves received by the radar to obtain the interfering electromagnetic wave. Based on the time when each radar receives the interfering electromagnetic wave, use the time difference method to locate the target emitting the interfering electromagnetic wave and obtain the position of the interference source; In S4, the time difference method is used to locate the target emitting the interfering electromagnetic wave and obtain the position of the interference source, including the following steps: (1) Synchronize the clock times of multiple radars, and control the clock time error to be less than the set clock error threshold. The set clock error threshold can be set according to the positioning accuracy. The clock error threshold for 1-meter-level accuracy is 3 nanoseconds, and the clock error threshold for centimeter-level accuracy is 0.33 nanoseconds; (2) Calculate the time difference between the interfering electromagnetic wave arriving at every two radars; (3) Represent the time difference in the form of a distance difference to obtain the distance difference formula between every two radars and the position of the interference source, , where is the distance difference, c is the speed of light, is the time difference, where: , where (x, y, z) are the three-dimensional coordinates of the interference source position, is the three-dimensional coordinate of one radar, is the three-dimensional coordinate of another radar, and the three-dimensional coordinates of the radar are known.

[0030] (4) Integrate all the distance difference formulas to obtain a distance difference equation system; (5) Solve based on the distance difference equation system to obtain the position of the interference source.

[0031] S5. Analyze the electromagnetic waves received by the radar, extract the echo of the electromagnetic wave emitted by the radar, obtain the target trajectory information of the target corresponding to the echo based on the echo, input the target trajectory information into the first bird prediction model. If the first bird prediction model outputs a suspected bird label, input the echo into the second bird prediction model. If the second bird prediction model outputs a bird label, determine whether there is a coincidence between the position of the interference source and the target position of the target. If so, determine that the target is a drone; if not, determine that the target is a bird.

[0032] A radar-based classification system for bird and drone flight track targets includes a radar module, a signal processing module, a first judgment module, a second judgment module, and a third judgment module; There are multiple radar modules, which are used to transmit and receive electromagnetic waves. The received electromagnetic waves include echoes and interfering electromagnetic waves; The signal processing module is used to process the received electromagnetic waves, separate the echoes and interfering electromagnetic waves, and calculate the target position corresponding to the echo and the interference source position corresponding to the interfering electromagnetic wave; The first judgment module is used to train a first machine learning model based on historical target trajectory information and a first bird judgment label to obtain a first bird prediction model, and is also used to input the target trajectory information into the first bird prediction model to obtain an output predicted first bird judgment label, and judge whether the target is a bird based on the first bird judgment label; The second judgment module is used to train a second machine learning model based on the echo, echo features and a second bird judgment label to obtain a second bird prediction model, and is also used to input the echo into the second bird prediction model to obtain an output second bird judgment label, and judge whether the target is a bird based on the second bird judgment label; The third judgment module is used to verify whether the judgment results of the first bird prediction model and the second bird prediction model are correct by judging whether there is an interference source position that coincides with the target position of the target determined to be a bird when the first bird prediction model and the second bird prediction model judge that the target is a bird.

[0033] Only some exemplary embodiments of the present invention have been described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. A method for classifying the flight track targets of birds and unmanned aerial vehicles based on radar, characterized in that: Including the following steps: S1. Locate the target position based on radar detection, and obtain target trajectory information based on the target position. The target trajectory information is the target position at each moment of the target. S2. Obtain historical target trajectory information, set the first bird judgment label associated with the corresponding historical target trajectory information, and train the first machine learning model based on the first bird judgment label and the historical target trajectory information to obtain the first bird prediction model. The first bird prediction model is used to output the predicted first bird judgment label according to the input target trajectory information. The first bird judgment label includes a drone label and a suspected bird label. S3. Collect the echoes of various types of birds on the electromagnetic waves of each frequency emitted by the radar, extract the echo features, set multiple second bird judgment labels, associate the echo features with the corresponding second bird judgment labels, and train the second machine learning model based on the echo, the echo features, and the second bird judgment labels to obtain the second bird prediction model. The second bird prediction model is used to extract the echo features according to the input echo and output the second bird judgment label corresponding to the extracted echo features. The second bird judgment label includes a drone label and a bird label. S4. Extract the electromagnetic waves other than the echo from the electromagnetic waves received by the radar to obtain interference electromagnetic waves, and use the time difference method to locate the target emitting the interference electromagnetic waves based on the time when each radar receives the interference electromagnetic waves to obtain the interference source position. S5. Analyze the electromagnetic waves received by the radar, extract the echo of the electromagnetic waves emitted by the radar, obtain the target trajectory information of the target corresponding to the echo based on the echo, input the target trajectory information into the first bird prediction model. If the first bird prediction model outputs a suspected bird label, input the echo into the second bird prediction model. If the second bird prediction model outputs a bird label, determine whether there is an interference source position that coincides with the target position of the target. If so, determine that the target is a drone; if not, determine that the target is a bird.

2. The method for classifying flight path targets of birds and unmanned aerial vehicles based on radar according to claim 1, wherein: The S2 includes the following steps: Obtain historical trajectory information and set the first bird judgment label. The first bird judgment label includes a drone label and a suspected bird label. Screen out the historical trajectory information of the determined drones from the historical trajectory information and associate it with the drone label. Associate the historical trajectory information after screening out the historical trajectory information of the determined drones with the suspected bird label. Select the first machine learning model, use the historical trajectory information as the input of the first machine learning model, extract the features of the historical trajectory information, associate the first bird judgment label associated with the historical trajectory information with the features of the historical trajectory information, and use the first bird judgment label as the output of the first machine learning model to train the first machine learning model to obtain the first bird prediction model.

3. A method for classifying flight path targets of birds and drones based on radar according to claim 1, characterized in that: The S3 includes the following steps: Obtain the electromagnetic waves emitted by the radar when detecting various types of birds at various distances from the radar and the corresponding received echoes. Based on the emission time of the electromagnetic wave and the reception time of the echo corresponding to the electromagnetic wave, calculate the distance between the target and the radar when the electromagnetic wave contacts the target to generate an echo, and obtain the target distance; Obtain the signal intensities of the electromagnetic wave and the corresponding echo, and calculate the signal intensity by which the echo attenuates relative to the corresponding electromagnetic wave to obtain the signal attenuation intensity; Classify the signal attenuation intensity and the target distance based on the types of birds to obtain multiple bird groups, and then classify the signal attenuation intensity and the target distance based on the frequencies of the electromagnetic wave and the corresponding echo in each bird group to obtain multiple frequency groups; Conduct a regression analysis on the relationship between the signal attenuation intensity and the corresponding target distance for each frequency group to obtain a relationship model between the signal attenuation intensity and the target distance, and obtain the error of the relationship model between the signal attenuation intensity and the target distance to obtain a first error threshold. The relationship model between the signal attenuation intensity and the target distance is used to output a predicted value of the signal attenuation intensity based on the input target distance; Set a judgment formula, which is used to judge whether the absolute value of the difference between the predicted value of the signal attenuation intensity corresponding to the same target distance in each frequency group and the signal attenuation intensity is less than the first error threshold, and associate the judgment result with the corresponding second bird judgment label; Integrate all the relationship models between the signal attenuation intensity and the target distance and the judgment formula to obtain a second bird prediction model. The second bird prediction model is used to, based on the input echo, obtain the frequency of the echo, retrieve the corresponding electromagnetic wave, calculate the target distance and the signal attenuation intensity, input the calculated target distance into the relationship model between the signal attenuation intensity and the target distance of the frequency group corresponding to the frequency of the echo to obtain a predicted value of the signal attenuation intensity, input the predicted value of the signal attenuation intensity and the signal attenuation intensity into the judgment formula of the frequency group, and judge whether the target is a bird based on the second bird judgment label output by the judgment formula.

4. A method for classifying flight path targets of birds and drones based on radar according to claim 1, characterized in that: The S3 includes the following steps: Obtain the electromagnetic waves emitted by the radar when detecting various types of birds at various distances from the radar and the corresponding received echoes; Based on the emission time of the electromagnetic wave and the reception time of the echo corresponding to the electromagnetic wave, calculate the distance between the target and the radar when the electromagnetic wave contacts the target to generate an echo, and obtain the target distance; Extract the phase change characteristics of the echo, and associate the phase change characteristics with the corresponding second bird judgment label; Train a second machine learning model based on the echo, the phase change characteristics, and the second bird judgment label to obtain a second bird prediction model. The second bird prediction model is used to, based on the input echo, extract the phase change characteristics of the echo and output a second bird label associated with the phase change characteristics.

5. A method for classifying flight track targets of birds and unmanned aerial vehicles based on radar according to claim 3, characterized in that: The S3 includes the following steps: Simultaneously emit electromagnetic waves of multiple frequencies towards the target to obtain multiple corresponding echoes; Input the multiple echoes into the second bird prediction model to obtain multiple output second bird judgment labels; Judge whether all the second bird judgment labels indicate that the target is a bird; If so, the target is a bird; if not, the target is a drone.

6. A method for classifying flight path targets of birds and drones based on radar according to claim 1, characterized in that: In S4, the time difference method is used to locate the target that emits the interfering electromagnetic wave, and the position of the interference source is obtained, including the following steps: Synchronize the clock times of multiple radars, and control the clock time error to be less than the set clock error threshold; Calculate the time difference between the interfering electromagnetic wave and each pair of radars; Express the time difference in the form of a distance difference to obtain the distance difference formula between each pair of radars and the position of the interference source; Integrate all the distance difference formulas to obtain a distance difference equation system; Solve based on the distance difference equation system to obtain the position of the interference source.

7. A radar-based bird and UAV track target classification system for implementing the radar-based bird and UAV track target classification method according to any one of claims 1-6, characterized in that: It includes a radar module, a signal processing module, a first judgment module, a second judgment module, and a third judgment module; There are multiple radar modules, which are used to transmit and receive electromagnetic waves. The received electromagnetic waves include echoes and interfering electromagnetic waves; The signal processing module is used to process the received electromagnetic waves, separate the echo and the interfering electromagnetic wave, and calculate the target position corresponding to the echo and the interference source position corresponding to the interfering electromagnetic wave; The first judgment module is used to train a first machine learning model based on historical target trajectory information and a first bird judgment label to obtain a first bird prediction model, and is also used to input the target trajectory information into the first bird prediction model to obtain the output predicted first bird judgment label, and judge whether the target is a bird based on the first bird judgment label; The second judgment module is used to train a second machine learning model based on the echo, echo features, and a second bird judgment label to obtain a second bird prediction model, and is also used to input the echo into the second bird prediction model to obtain the output second bird judgment label, and judge whether the target is a bird based on the second bird judgment label; The third judgment module is used to verify whether the judgment results of the first bird prediction model and the second bird prediction model are correct by judging whether there is an overlap between the interference source position and the target position of the target that is a bird when the first bird prediction model and the second bird prediction model judge that the target is a bird.

Images