CN115577305B - Unmanned aerial vehicle signal intelligent recognition method and device - Google Patents
Unmanned aerial vehicle signal intelligent recognition method and device Download PDFInfo
- Publication number
- CN115577305B CN115577305B CN202211350359.2A CN202211350359A CN115577305B CN 115577305 B CN115577305 B CN 115577305B CN 202211350359 A CN202211350359 A CN 202211350359A CN 115577305 B CN115577305 B CN 115577305B
- Authority
- CN
- China
- Prior art keywords
- signal
- model
- unmanned aerial
- aerial vehicle
- representing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R29/00—Arrangements for measuring or indicating electric quantities not covered by groups G01R19/00 - G01R27/00
- G01R29/08—Measuring electromagnetic field characteristics
- G01R29/0864—Measuring electromagnetic field characteristics characterised by constructional or functional features
- G01R29/0892—Details related to signal analysis or treatment; presenting results, e.g. displays; measuring specific signal features other than field strength, e.g. polarisation, field modes, phase, envelope, maximum value
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses an intelligent unmanned aerial vehicle signal recognition method and device, wherein the method comprises the following steps: receiving a space electromagnetic wave signal by using radio reconnaissance equipment, and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal; the zero intermediate frequency signal is sent to signal processing equipment, and the zero intermediate frequency signal is processed by the signal processing equipment to obtain a training signal sample database; constructing an intelligent unmanned aerial vehicle signal recognition initial model, and training the intelligent unmanned aerial vehicle signal recognition initial model by utilizing the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model; and receiving the space electromagnetic wave signal to be identified by using the radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by using the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result. The method can solve the problem of signal detection under the conditions of a large amount of background signals and real-time detection and accuracy in the prior art.
Description
Technical Field
The invention relates to the technical field of electronic information, in particular to an unmanned aerial vehicle signal intelligent recognition method and device.
Background
Along with the continuous updating of science and technology, unmanned aerial vehicles show a rapid development situation in the military and civil fields. Unmanned aerial vehicle technique has also inevitably brought the problem of management and control and reaction when bringing convenience for a great deal of fields, needs urgent solution. The anti-unmanned technology has thus grown and developed vigorously. The precondition for realizing the various technologies of the anti-unmanned aerial vehicle is that the target unmanned aerial vehicle signal is detected rapidly and accurately.
Modern social communication equipment is increased, and electromagnetic environments are complex and changeable along with various noise, interference, fading and multipath effects. In the face of the high dynamic and complex electromagnetic environment, the existing signal detection method has the defects of poor robustness, insufficient generalization capability and the like, and is difficult to cope with the complex electromagnetic environment, so that some difficult problems are needed to be solved.
The difficult problems faced by the intelligent recognition technology of the unmanned aerial vehicle signal under the complex electromagnetic environment are as follows: (1) A large amount of background signals exist in the same frequency band, and various frequency signals mutually interfere; (2) Because the electromagnetic environment is complex, the real-time detection and the accuracy can not achieve better effects at the same time.
Disclosure of Invention
The invention aims to solve the technical problem that in order to meet the requirements of detection and identification of a small unmanned aerial vehicle in a complex electromagnetic environment, the invention provides an intelligent unmanned aerial vehicle signal identification method and device, and detection and identification of the small unmanned aerial vehicle signal and a background signal in the same frequency band are realized through modes of radio detection, signal time-frequency conversion, filtering, signal labeling, target detection based on deep learning, knowledge distillation and the like.
In order to solve the technical problems, a first aspect of the embodiment of the invention discloses an intelligent unmanned aerial vehicle signal recognition method, which comprises the following steps:
s1, receiving a space electromagnetic wave signal by using radio reconnaissance equipment, and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal;
s2, sending the zero intermediate frequency signal to signal processing equipment, and processing the zero intermediate frequency signal by using the signal processing equipment to obtain a training signal sample database, wherein the method comprises the following steps of:
s21, storing and preprocessing the zero intermediate frequency signal by using the signal processing equipment to obtain time-frequency conversion data information;
the time-frequency transformation is a smooth WVD distribution:
where G (t, ω) is the time-frequency distribution of a window function, W x (t, ω) is the WVD distribution of the time domain signal x (), u, ζ is the convolution parameter;
for smooth WVD distribution SW x (t, omega) performing bilateral filtering to obtain time-frequency conversion data information after noise reduction and edge preservation
S22, marking the time-frequency conversion data information to obtain a training signal sample database;
s3, constructing an intelligent unmanned aerial vehicle signal recognition initial model, training the intelligent unmanned aerial vehicle signal recognition initial model by using the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model, wherein the intelligent unmanned aerial vehicle signal recognition model comprises the following components:
s31, constructing an intelligent unmanned aerial vehicle signal recognition initial model;
the unmanned aerial vehicle signal intelligent recognition initial model is a YOLO model, and a loss function of the YOLO model is as follows:
Loss_yolo＝λL CIoU (b,b gt )+λL obj (p o ,p iou )+λL cls (c p ,c gt )
L CIoU (b,b gt )＝αLoss soft (b T ,b S )+(1-α)Loss hard (b T ,b gt )
L obj (p o ,p iou )＝αLoss soft (p T ,p S )+(1-α)Loss hard (p T ,p gt )
L cls (c p ,c gt )＝αLoss soft (c T ,c S )+(1-α)Loss hard (c T ,c gt )
where Loss_yolo represents the overall Loss of the YOLO model, λ represents the weight of the corresponding term, L CIoU Represents the regression loss of the boundary box, b represents the predicted value of the boundary box, b gt Representing the true value of the bounding box, L obj Representing target confidence loss, p o Representing the confidence score, p, of a target in a prediction box iou IoU value, L representing prediction frame and target frame corresponding thereto cls Representing class loss, c p Representing class score of prediction frame, c gt Representing the true value of a class, alpha represents the weight, b s Boundary box predicted value representing student model, b T Boundary box predicted value, p, representing teacher model S Representing the target confidence score, p, in a student model prediction frame T Representing the target confidence score, p, in the teacher model prediction frame gt Representing the true value of the target, c S Class score representing student model prediction box, c T A class score representing a teacher model prediction frame;
s32, compressing the intelligent unmanned aerial vehicle signal recognition initial model by using a characteristic distillation method based on an FSP matrix to obtain an optimized intelligent unmanned aerial vehicle signal recognition model;
the characteristic distillation method based on the FSP matrix comprises the following steps:
wherein F is 1 Is a shallow feature map, F 2 Is a deep feature map, h is the length of the feature map, w is the width of the feature map, i, j is F 1 、F 2 Corresponding channel indexes, x and W are current inputs and parameters; approximating between FSP matrices of a teacher model and a student model using L2 LossDistance L of (2) FSP (W t ,W s ) L2 Loss is defined as:
in the method, in the process of the invention,
FSP matrix for teacher network, +.>
FSP matrix lambda for student network i Represents weight, N is the number of data points, W t For teacher network parameters, W s For student network parameters, x is the current input, i is channel index, n is channel number;
the teacher model is a YOLOv5l model; the student model is a YOLOv5s model;
s33, training the intelligent recognition model of the optimized unmanned aerial vehicle signal by using the training signal sample database to obtain the intelligent recognition model of the unmanned aerial vehicle signal;
s4, receiving a space electromagnetic wave signal to be identified by utilizing radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by utilizing the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result.
In a first aspect of the embodiment of the present invention, the receiving, by a radio reconnaissance device, a spatial electromagnetic wave signal, and processing the spatial electromagnetic wave signal to obtain a zero intermediate frequency signal, where the zero intermediate frequency signal includes:
s11, the radio reconnaissance equipment receives a detection instruction sent by the signal processing equipment;
s12, the radio reconnaissance equipment receives a space electromagnetic wave signal according to the detection instruction;
s13, the radio reconnaissance equipment processes the received space electromagnetic wave signals to obtain zero intermediate frequency signals.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the time-frequency spectrogram SW x (t, omega) bilateral filtering to obtain a time-frequency spectrogram after noise reduction and edge preservation
Comprising the following steps:
where W () represents the original time-frequency spectrum,
representing a time-frequency spectrogram after bilateral filtering processing, m representing an output pixel point, n representing an input pixel point, s representing a set rectangular frame area for traversing an image, W being a two-dimensional template, and>
and
is two Gaussian functions, representing a spatial domain kernel and a pixel domain kernel, M m The specific expression of (2) is:
wherein a and b represent the abscissa and ordinate of the input pixel, i and j represent the coordinates of the center of the box, σ s 、σ r The standard deviation of the gaussian function is shown.
The second aspect of the invention discloses an intelligent unmanned aerial vehicle signal recognition device, which comprises:
the signal receiving module is used for receiving the space electromagnetic wave signal by utilizing the radio reconnaissance equipment and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal;
the training database generation module is configured to send the zero intermediate frequency signal to a signal processing device, process the zero intermediate frequency signal by using the signal processing device, and obtain a training signal sample database, and includes:
s21, storing and preprocessing the zero intermediate frequency signal by using the signal processing equipment to obtain time-frequency conversion data information;
the time-frequency transformation is a smooth WVD distribution:
where G (t, ω) is the time-frequency distribution of a window function, W x (t, ω) is the WVD distribution of the time domain signal x (), u, ζ is the convolution parameter;
for smooth WVD distribution SW x (t, omega) performing bilateral filtering to obtain time-frequency conversion data information after noise reduction and edge preservation
S22, marking the time-frequency conversion data information to obtain a training signal sample database;
the training module is used for constructing an initial model of intelligent unmanned aerial vehicle signal recognition, training the initial model of intelligent unmanned aerial vehicle signal recognition by using the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model, and comprises the following steps:
s31, constructing an intelligent unmanned aerial vehicle signal recognition initial model;
the unmanned aerial vehicle signal intelligent recognition initial model is a YOLO model, and a loss function of the YOLO model is as follows:
Loss_yolo＝λL CIoU (b,b gt )+λL obj (p o ,p iou )+λL cls (c p ,c gt )
L CIoU (b,b gt )＝αLoss soft (b T ,b S )+(1-α)Loss hard (b T ,b gt )
L obj (p o ,p iou )＝αLoss soft (p T ,p S )+(1-α)Loss hard (p T ,p gt )
L cls (c p ,c gt )＝αLoss soft (c T ,c S )+(1-α)Loss hard (c T ,c gt )
where Loss_yolo represents the overall Loss of the YOLO model, λ represents the weight of the corresponding term, L CIoU Represents the regression loss of the boundary box, b represents the predicted value of the boundary box, b gt Representing the true value of the bounding box, L obj Representing target confidence loss, p o Representing the confidence score, p, of a target in a prediction box iou IoU value, L representing prediction frame and target frame corresponding thereto cls Representing class loss, c p Representing class score of prediction frame, c gt Representing the true value of a class, alpha represents the weight, b s Boundary box predicted value representing student model, b T Boundary box predicted value, p, representing teacher model S Representing the target confidence score, p, in a student model prediction frame T Representing the target confidence score, p, in the teacher model prediction frame gt Representing the true value of the target, c S Class score representing student model prediction box, c T A class score representing a teacher model prediction frame;
s32, compressing the intelligent unmanned aerial vehicle signal recognition initial model by using a characteristic distillation method based on an FSP matrix to obtain an optimized intelligent unmanned aerial vehicle signal recognition model;
the characteristic distillation method based on the FSP matrix comprises the following steps:
wherein F is 1 Is a shallow feature map, F 2 Is a deep feature map, h is the length of the feature map, w is the width of the feature map, i, j is F 1 、F 2 Corresponding channel indexes, x and W are current inputs and parameters; approximating distance L between FSP matrices of teacher model and student model using L2 Loss FSP (W t ,W s ) L2 Loss is defined as:
in the method, in the process of the invention,
FSP matrix for teacher network, +.>
FSP matrix lambda for student network i Represents weight, N is the number of data points, W t For teacher network parameters, W s For student network parameters, x is the current input, i is channel index, n is channel number;
the teacher model is a YOLOv5l model; the student model is a YOLOv5s model;
s33, training the intelligent recognition model of the optimized unmanned aerial vehicle signal by using the training signal sample database to obtain the intelligent recognition model of the unmanned aerial vehicle signal;
the identification module is used for receiving the space electromagnetic wave signal to be identified by utilizing the radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by utilizing the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result.
In a second aspect of the embodiment of the present invention, the receiving, by a radio reconnaissance device, a spatial electromagnetic wave signal, and processing the spatial electromagnetic wave signal to obtain a zero intermediate frequency signal, where the zero intermediate frequency signal includes:
s11, the radio reconnaissance equipment receives a detection instruction sent by the signal processing equipment;
s12, the radio reconnaissance equipment receives a space electromagnetic wave signal according to the detection instruction;
s13, the radio reconnaissance equipment processes the received space electromagnetic wave signals to obtain zero intermediate frequency signals.
In a second aspect of the present invention, the processing the zero intermediate frequency signal by the signal processing device to obtain a training signal sample database includes:
s21, storing and preprocessing the zero intermediate frequency signal by using the signal processing equipment;
the preprocessing comprises the steps of performing time-frequency conversion on the zero intermediate frequency signal to obtain time-frequency conversion data information;
s22, labeling the time-frequency conversion data information to obtain a training signal sample database.
In a second aspect of the embodiment of the present invention, the time-frequency spectrogram SW x (t, omega) bilateral filtering to obtain a time-frequency spectrogram after noise reduction and edge preservation
Comprising the following steps:
where W () represents the original time-frequency spectrum,
representing a time-frequency spectrogram after bilateral filtering processing, m representing an output pixel point, n representing an input pixel point, s representing a set rectangular frame area for traversing an image, W being a two-dimensional template, and>
and
is two Gaussian functions, representing a spatial domain kernel and a pixel domain kernel, M m The specific expression of (2) is:
wherein a and b represent the abscissa and ordinate of the input pixel, i and j represent the coordinates of the center of the box, σ s 、σ r The standard deviation of the gaussian function is shown.
In a second aspect of the embodiment of the present invention, the receiving, by using a radio reconnaissance device, a spatial electromagnetic wave signal to be identified, and processing, by using the intelligent recognition model for an unmanned aerial vehicle signal, the spatial electromagnetic wave signal to be identified to obtain an intelligent recognition result for the unmanned aerial vehicle signal, includes:
s41, receiving a space electromagnetic wave signal to be identified by using a radio reconnaissance device;
s42, processing the space electromagnetic wave signal to be identified to obtain a time-frequency spectrogram to be identified, and marking the time-frequency spectrogram to be identified to obtain a marked time-frequency spectrogram;
s43, processing the labeled time-frequency spectrogram by using the unmanned aerial vehicle signal intelligent recognition model to obtain an unmanned aerial vehicle signal intelligent recognition result.
The third aspect of the invention discloses another unmanned aerial vehicle signal intelligent recognition device, which comprises:
a memory storing executable program code;
a processor coupled to the memory;
the processor invokes the executable program codes stored in the memory to execute part or all of the steps in the unmanned aerial vehicle signal intelligent recognition method disclosed in the first aspect of the embodiment of the invention.
The fourth aspect of the present invention discloses a computer storage medium, where the computer storage medium stores computer instructions, where the computer instructions are used to execute part or all of the steps in the method for intelligent recognition of a signal of an unmanned aerial vehicle disclosed in the first aspect of the present invention when the computer instructions are called.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
(1) The invention provides an intelligent recognition method for unmanned aerial vehicle signals under complex electromagnetic conditions, which solves the problem of signal detection under the conditions of a large amount of background signals and real-time detection and accuracy;
(2) The deep learning knowledge distillation idea is adopted, the knowledge learned by a larger teacher model is transferred to a smaller student model, and the purpose of reducing the model operation amount, improving the model detection speed, reducing the overfitting and finally achieving the purpose of compressing the model to realize equipment end configuration is achieved under the condition that the precision is kept basically unchanged.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of an intelligent unmanned aerial vehicle signal recognition method disclosed by the embodiment of the invention;
FIG. 2 is a schematic diagram of a knowledge distillation flow based on deep learning object detection in accordance with an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an intelligent recognition device for unmanned aerial vehicle signals according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of another intelligent unmanned aerial vehicle signal recognition device according to an embodiment of the present invention.
Detailed Description
In order to make the present invention better understood by those skilled in the art, the following description will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The terms first, second and the like in the description and in the claims and in the above-described figures are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, apparatus, article, or device that comprises a list of steps or elements is not limited to the list of steps or elements but may, in the alternative, include other steps or elements not expressly listed or inherent to such process, method, article, or device.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
The invention discloses an intelligent unmanned aerial vehicle signal recognition method and device, which can utilize radio reconnaissance equipment to receive a space electromagnetic wave signal and process the space electromagnetic wave signal to obtain a zero intermediate frequency signal; the zero intermediate frequency signal is sent to signal processing equipment, and the zero intermediate frequency signal is processed by the signal processing equipment to obtain a training signal sample database; constructing an intelligent unmanned aerial vehicle signal recognition initial model, and training the intelligent unmanned aerial vehicle signal recognition initial model by utilizing the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model; and receiving the space electromagnetic wave signal to be identified by using the radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by using the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result. The method can solve the problem of signal detection under the conditions of a large amount of background signals and real-time detection and accuracy in the prior art.
Example 1
Referring to fig. 1, fig. 1 is a schematic flow chart of an intelligent unmanned aerial vehicle signal recognition method according to an embodiment of the invention. The unmanned aerial vehicle signal intelligent recognition method described in fig. 1 can be applied to unmanned aerial vehicle recognition systems or other signal processing fields, and the embodiment of the invention is not limited. As shown in fig. 1, the unmanned aerial vehicle signal intelligent recognition method may include the following operations:
s1, receiving a space electromagnetic wave signal by using radio reconnaissance equipment, and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal;
s2, the zero intermediate frequency signal is sent to signal processing equipment, and the zero intermediate frequency signal is processed by the signal processing equipment to obtain a training signal sample database;
s3, constructing an intelligent unmanned aerial vehicle signal recognition initial model, and training the intelligent unmanned aerial vehicle signal recognition initial model by using the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model;
s4, receiving a space electromagnetic wave signal to be identified by utilizing radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by utilizing the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result.
The radio reconnaissance device is used for signal reception, low noise amplification, filtering, A/D conversion, digital down conversion and finally converting the signal into a zero intermediate frequency signal.
Storing and preprocessing the zero intermediate frequency signal by using a signal processing device;
the preprocessing comprises the steps of performing time-frequency conversion on the zero intermediate frequency signal to obtain time-frequency conversion data information;
and labeling the time-frequency conversion data information to obtain a training signal sample database.
Optionally, the labeling includes labeling the coordinate position (start-stop time, start-stop frequency point) and the signal type, and the signal type includes: unmanned aerial vehicle measurement and control signal, unmanned aerial vehicle image signal, wi-Fi signal, bluetooth signal and other unknown signals. Training a signal recognition model using the annotation data. The model is compressed by using a deep learning knowledge distillation technology, and the complexity of the model is reduced under the condition that the detection precision is basically unchanged, so that the real-time application requirement is met.
The time-frequency transformation is WVD transformation:
wherein t represents time, ω represents angular frequency, x () represents time domain signal, τ represents delay, and W (t, ω) is a time-frequency spectrogram;
alternatively, let G (t, ω) be the time-frequency distribution of a window function, let G (t, ω) and W x The convolution of (t, ω) in both the directions t and ω is called smooth WVD, denoted SW x (t, ω), i.e.:
as for G (t, ω) vs W x The effect of the (t, ω) action depends on the shape of G (t, ω), u, ζ being the convolution parameter.
Bilateral filtering is carried out on the time-frequency spectrogram W (t, omega) to obtain a time-frequency spectrogram after noise reduction and edge preservation
Where W () represents the original time-frequency spectrum,
representing a time-frequency spectrogram after bilateral filtering processing, m representing an output pixel point, n representing an input pixel point, s representing a set rectangular frame area for traversing an image, W being a two-dimensional template, and>
and
is two Gaussian functions, representing a spatial domain kernel and a pixel domain kernel, M m The specific expression of (2) is:
Wherein a and b represent the abscissa and ordinate of the input pixel, i and j represent the coordinates of the center of the box, σ s 、σ r The standard deviation of the gaussian function is shown.
Constructing an intelligent unmanned aerial vehicle signal recognition initial model, wherein the intelligent unmanned aerial vehicle signal recognition initial model is a YOLO model based on deep learning target detection; in the YOLO target detection model, a time-frequency spectrogram is divided into A×B grids, and the grid where the time-frequency center position of a target signal is located is responsible for detection and identification of the signal. The information of the signals in each time-frequency spectrum, namely the starting and ending time, the starting and ending frequency point and the signal type, is expressed as A multiplied by B multiplied by [ (4+1) ×N+C ]. Wherein A x B represents the grid number, 4 represents the time-frequency central coordinate and width and height of the signal, 1 represents the confidence coefficient, N represents the number of prediction frames, and C represents the number of various signal types recognized by the system.
The YOLO target detection model loss function is:
Loss_yolo＝λL CIoU (b,b gt )+λL obj (p o ,p iou )+λL cls (c p ,c gt )
L CIoU (b,b gt )＝αLoss soft (b T ,b S )+(1-α)Loss hard (b T ,b gt )
L obj (p o ,p iou )＝αLoss soft (p T ,p S )+(1-α)Loss hard (p T ,p gt )
L cls (c p ,c gt )＝αLoss soft (c T ,c S )+(1-α)Loss hard (c T ,c gt )
where Loss_yolo represents the overall Loss of the YOLO model, λ represents the weight of the corresponding term, L CIoU Represents the regression loss of the boundary box, b represents the predicted value of the boundary box, b gt Representing the true value of the bounding box, L obj Representing target confidence loss, p o Representing the confidence score, p, of a target in a prediction box iou IoU value, L representing prediction frame and target frame corresponding thereto cls Representing class loss, c p Representing class score of prediction frame, c gt Representing the true value of a class, alpha represents the weight, b s Boundary box predicted value representing student model, b T Boundary box predicted value, p, representing teacher model S Representing the target confidence score, p, in a student model prediction frame T Representing the target confidence score, p, in the teacher model prediction frame gt Representing the true value of the target, c S Class score representing student model prediction box, c T A class score representing the teacher model prediction box.
Compressing the intelligent unmanned aerial vehicle signal recognition initial model by using a deep learning knowledge distillation model, and reducing the complexity of the model to obtain an optimized intelligent unmanned aerial vehicle signal recognition model; the deep learning knowledge distillation model is as follows:
s321, setting a teacher model and a student model;
the teacher model is a YOLOv5l model; the student model is a YOLOv5s model; compared with the YOLOv5s model, the network of YOLOv5l is deeper and wider, the model detection precision is higher, and the operation time is also increased. The purpose of knowledge distillation is to learn knowledge of a larger teacher network with a smaller student network.
S322, setting a temperature coefficient T, and controlling the importance of each soft label by introducing the temperature coefficient T in the distillation process. The specific operation is to divide the output result of the teacher model by the temperature coefficient and then make Softmax calculation to obtain the soft label. The Softmax function was obtained as:
wherein p is i For the i-th class output probability, T is the temperature coefficient, z i The output value is the i-th class, j is the number of classification classes;
the difference between the soft label obtained by the teacher model and the soft prediction obtained by the student model is distillation loss, so that the output result of the student model is ensured to be as consistent as possible with the teacher model, namely the probability distribution of the student model is enabled to be as close as possible to the distribution probability of the teacher network after the temperature coefficient is added. The student model uses KL divergence to supervise the learning process of the teacher network probability distribution.
S323, calculating distillation Loss using Softmax function soft ：
Loss soft ＝T 2 *KLdiv(P S ,P T )
Wherein P is T Probability distribution for teacher model, P S As the probability distribution of the student model, KLdiv () is KL divergence;
the principle of KL divergence is:
from the above formula:
KLdiv(P S ,P T )＝H(P T ,P S )-H(P T )
H(P T ,P S ) Is P T And P S Cross entropy, H (P) T ) Is P T Is a function of the entropy of (a). KL divergence is equal to cross entropy minus entropy. For a given data set, entropy is known, finding the KL divergence is equivalent to finding the cross entropy, and therefore uses the cross entropy as a loss function.
S324, cross entropy of hard label and hard prediction output data of student model is used to ensure that output result of student model is consistent with real label as much as possible. The hard object is the object of normal network training, the purpose of which is to make the student model as correct as possibleAnd (5) classification. Calculating student model Loss through difference between student model and real label hard The method comprises the following steps:
Loss hard ＝H(P S ,y gt )
wherein y is gt Representing the authentic tag, H (P S ,y gt ) Representing P S ,y gt Cross entropy of (2);
s325, weighting the objective function of knowledge distillation by the distillation loss corresponding to the soft target and the student model loss corresponding to the hard target, and setting the objective function of knowledge distillation as follows:
Loss＝αLoss soft +(1-α)Loss hard
in the formula, loss is total Loss, loss soft Loss of distillation hard For student model loss, α is the weight. A schematic diagram of a knowledge distillation flow based on deep learning target detection is shown in fig. 2.
Alternatively, a characteristic distillation method based on an FSP matrix can be used to compress the network. The FSP is used for defining the characteristic relation between layers, channels of the shallow layer characteristic diagram and the deep layer characteristic diagram with the same resolution are in one-to-one correspondence to calculate inner products, and the result is placed at a position corresponding to the FSP matrix. The calculation process is as follows:
wherein F is 1 Is a shallow feature map, F 2 Is a deep feature map, h is the length of the feature map, w is the width of the feature map, i, j is F 1 、F 2 The corresponding channel index, x and W, are the current inputs and parameters. The distance L between FSP matrix of teacher model and student model is then approximated by L2 Loss FSP (W t ,W s ) L2 Loss is defined as follows:
in the method, in the process of the invention,
FSP matrix for teacher network, +.>
FSP matrix lambda for student network i Represents the weight, N is the number of data points, W t For teacher network parameters, W s For the student network parameters, x is the current input, i is the channel index, and n is the channel number.
The identification stage, the radio reconnaissance equipment is utilized to receive the space electromagnetic wave signal to be identified; processing the space electromagnetic wave signal to be identified, including time-frequency conversion and bilateral filtering, to obtain a time-frequency spectrogram to be identified, and marking the time-frequency spectrogram to be identified to obtain a marked time-frequency spectrogram; and processing the marked time-frequency spectrogram by using the unmanned aerial vehicle signal intelligent recognition model, and giving out starting and stopping time, starting and stopping frequency points and type prediction to obtain an unmanned aerial vehicle signal intelligent recognition result.
Example two
Referring to fig. 3, fig. 3 is a schematic structural diagram of an intelligent recognition device for unmanned aerial vehicle signals according to an embodiment of the present invention. The device described in fig. 3 can be applied to unmanned aerial vehicle signal recognition and other target recognition in the electronic information field, and the embodiment of the invention is not limited. As shown in fig. 3, the apparatus may include:
s301, a signal receiving module is used for receiving a space electromagnetic wave signal by using radio reconnaissance equipment and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal;
s302, a training database generation module is used for sending the zero intermediate frequency signal to signal processing equipment, and the signal processing equipment is used for processing the zero intermediate frequency signal to obtain a training signal sample database;
s303, a training module is used for constructing an intelligent unmanned aerial vehicle signal recognition initial model, and training the intelligent unmanned aerial vehicle signal recognition initial model by using the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model;
s304, an identification module is used for receiving the space electromagnetic wave signal to be identified by utilizing the radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by utilizing the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result.
Example III
Referring to fig. 4, fig. 4 is a schematic structural diagram of another intelligent unmanned aerial vehicle signal recognition device according to an embodiment of the present invention. The device described in fig. 4 can be applied to unmanned aerial vehicle signal recognition and other target recognition in the electronic information field, and the embodiment of the invention is not limited. As shown in fig. 4, the apparatus may include:
a memory 401 storing executable program codes;
a processor 402 coupled with the memory 401;
the processor 402 invokes executable program code stored in the memory 401 for performing the steps in the unmanned aerial vehicle signal intelligent recognition method described in embodiment one.
Example IV
The embodiment of the invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute the steps in the unmanned aerial vehicle signal intelligent recognition method described in the embodiment one.
The apparatus embodiments described above are merely illustrative, in which the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical, i.e., may be located in one place, or may be distributed over multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above detailed description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course by means of hardware. Based on such understanding, the foregoing technical solutions may be embodied essentially or in part in the form of a software product that may be stored in a computer-readable storage medium including Read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), one-time programmable Read-Only Memory (OTPROM), electrically erasable programmable Read-Only Memory (EEPROM), compact disc Read-Only Memory (Compact Disc Read-Only Memory, CD-ROM) or other optical disc Memory, magnetic disc Memory, tape Memory, or any other medium that can be used for computer-readable carrying or storing data.
Finally, it should be noted that: the embodiment of the invention discloses an intelligent unmanned aerial vehicle signal recognition method and device, which are disclosed by the embodiment of the invention and are only used for illustrating the technical scheme of the invention, but not limiting the technical scheme; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme recorded in the various embodiments can be modified or part of technical features in the technical scheme can be replaced equivalently; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (7)
1. An unmanned aerial vehicle signal intelligent recognition method is characterized by comprising the following steps:
s1, receiving a space electromagnetic wave signal by using radio reconnaissance equipment, and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal;
s2, sending the zero intermediate frequency signal to signal processing equipment, and processing the zero intermediate frequency signal by using the signal processing equipment to obtain a training signal sample database, wherein the method comprises the following steps of:
s21, storing and preprocessing the zero intermediate frequency signal by using the signal processing equipment to obtain time-frequency conversion data information;
the time-frequency transformation is a smooth WVD distribution:
where G (t, ω) is the time-frequency distribution of a window function, W x (t, ω) is the WVD distribution of the time domain signal x (), u, ζ is the convolution parameter;
for smooth WVD distribution SW x (t, omega) performing bilateral filtering to obtain time-frequency conversion data information after noise reduction and edge preservation
S22, marking the time-frequency conversion data information to obtain a training signal sample database;
s3, constructing an intelligent unmanned aerial vehicle signal recognition initial model, training the intelligent unmanned aerial vehicle signal recognition initial model by using the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model, wherein the intelligent unmanned aerial vehicle signal recognition model comprises the following components:
s31, constructing an intelligent unmanned aerial vehicle signal recognition initial model;
the unmanned aerial vehicle signal intelligent recognition initial model is a YOLO model, and a loss function of the YOLO model is as follows:
Loss_yolo＝λL CIoU (b,b gt )+λL obj (p o ,p iou )+λL cls (c p ,c gt )
L CIoU (b,b gt )＝αLoss soft (b T ,b S )+(1-α)Loss hard (b T ,b gt )
L obj (p o ,p iou )＝αLoss soft (p T ,p S )+(1-α)Loss hard (p T ,p gt )
L cls (c p ,c gt )＝αLoss soft (c T ,c S )+(1-α)Loss hard (c T ,c gt )
where Loss_yolo represents the overall Loss of the YOLO model, λ represents the weight of the corresponding term, L CIoU Represents the regression loss of the boundary box, b represents the predicted value of the boundary box, b gt Representing the true value of the bounding box, L obj Representing target confidence loss, p o Representing the confidence score, p, of a target in a prediction box iou IoU value, L representing prediction frame and target frame corresponding thereto cls Representing class loss, c p Representing class score of prediction frame, c gt Representing the true value of a class, alpha represents the weight, b s Boundary box predicted value representing student model, b T Boundary box predicted value, p, representing teacher model S Representing the target confidence score, p, in a student model prediction frame T Representing the target confidence score, p, in the teacher model prediction frame gt Representing the true value of the target, c S Class score representing student model prediction box, c T A class score representing a teacher model prediction frame;
s32, compressing the intelligent unmanned aerial vehicle signal recognition initial model by using a characteristic distillation method based on an FSP matrix to obtain an optimized intelligent unmanned aerial vehicle signal recognition model;
the characteristic distillation method based on the FSP matrix comprises the following steps:
wherein F is 1 Is a shallow feature map, F 2 Is a deep feature map, h is the length of the feature map, w is the width of the feature map, i, j is F 1 、F 2 Corresponding channel indexes, x and W are current inputs and parameters; approximating distance L between FSP matrices of teacher model and student model using L2 Loss FSP (W t ,W s ) L2 Loss is defined as:
in the method, in the process of the invention,
FSP matrix for teacher network, +.>
FSP matrix lambda for student network i Represents weight, N is the number of data points, W t For teacher network parameters, W s For student network parameters, x is the current input, i is channel index, n is channel number;
the teacher model is a YOLOv5l model; the student model is a YOLOv5s model;
s33, training the intelligent recognition model of the optimized unmanned aerial vehicle signal by using the training signal sample database to obtain the intelligent recognition model of the unmanned aerial vehicle signal;
s4, receiving a space electromagnetic wave signal to be identified by utilizing radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by utilizing the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result.
2. The method for intelligent recognition of unmanned aerial vehicle signals according to claim 1, wherein the steps of receiving a spatial electromagnetic wave signal by a radio reconnaissance device, and processing the spatial electromagnetic wave signal to obtain a zero intermediate frequency signal, comprise:
s11, receiving a detection instruction sent by the signal processing equipment by utilizing the radio reconnaissance equipment;
s12, receiving a space electromagnetic wave signal by using the radio reconnaissance equipment according to the detection instruction;
s13, the radio reconnaissance equipment is utilized to process the received space electromagnetic wave signals, and zero intermediate frequency signals are obtained.
3. The unmanned aerial vehicle signal intelligent recognition method of claim 1, wherein the following is performed
Time-frequency spectrum diagram SW x (t, omega) bilateral filtering to obtain a time-frequency spectrogram after noise reduction and edge preservation
Comprising the following steps:
where W () represents the original time-frequency spectrum,
representing a time-frequency spectrogram after bilateral filtering processing, m representing an output pixel point, n representing an input pixel point, s representing a set rectangular frame area for traversing an image, W being a two-dimensional template, and>
and->
Is two Gaussian functions, representing a spatial domain kernel and a pixel domain kernel, M m The specific expression of (2) is:
wherein a and b represent the abscissa and ordinate of the input pixel, i and j represent the coordinates of the center of the box, σ s 、σ r The standard deviation of the gaussian function is shown.
4. The method for intelligent recognition of a signal of an unmanned aerial vehicle according to claim 1, wherein the receiving the signal of the electromagnetic wave of the space to be recognized by the radio reconnaissance device, and the processing the signal of the electromagnetic wave of the space to be recognized by the intelligent recognition model of the signal of the unmanned aerial vehicle, to obtain the intelligent recognition result of the signal of the unmanned aerial vehicle, comprises:
s41, receiving a space electromagnetic wave signal to be identified by using a radio reconnaissance device;
s42, processing the space electromagnetic wave signal to be identified to obtain a time-frequency spectrogram to be identified,
s43, labeling the time-frequency spectrogram to be identified to obtain a labeled time-frequency spectrogram;
s44, processing the labeled time-frequency spectrogram by using the unmanned aerial vehicle signal intelligent recognition model to obtain an unmanned aerial vehicle signal intelligent recognition result.
5. Unmanned aerial vehicle signal intelligent recognition device, its characterized in that, the device includes:
the signal receiving module is used for receiving the space electromagnetic wave signal by utilizing the radio reconnaissance equipment and processing the space electromagnetic wave signal to obtain a zero intermediate frequency signal;
the training database generation module is configured to send the zero intermediate frequency signal to a signal processing device, process the zero intermediate frequency signal by using the signal processing device, and obtain a training signal sample database, and includes:
s21, storing and preprocessing the zero intermediate frequency signal by using the signal processing equipment to obtain time-frequency conversion data information;
the time-frequency transformation is a smooth WVD distribution:
where G (t, ω) is the time-frequency distribution of a window function, W x (t, ω) is the WVD distribution of the time domain signal x (), u, ζ is the convolution parameter;
for smooth WVD distribution SW x (t, omega) performing bilateral filtering to obtain time-frequency conversion data information after noise reduction and edge preservation
S22, marking the time-frequency conversion data information to obtain a training signal sample database;
the training module is used for constructing an initial model of intelligent unmanned aerial vehicle signal recognition, training the initial model of intelligent unmanned aerial vehicle signal recognition by using the training signal sample database to obtain an intelligent unmanned aerial vehicle signal recognition model, and comprises the following steps:
s31, constructing an intelligent unmanned aerial vehicle signal recognition initial model;
the unmanned aerial vehicle signal intelligent recognition initial model is a YOLO model, and a loss function of the YOLO model is as follows:
Loss_yolo＝λL CIoU (b,b gt )+λL obj (p o ,p iou )+λL cls (c p ,c gt )
L CIoU (b,b gt )＝αLoss soft (b T ,b S )+(1-α)Loss hard (b T ,b gt )
L obj (p o ,p iou )＝αLoss soft (p T ,p S )+(1-α)Loss hard (p T ,p gt )
L cls (c p ,c gt )＝αLoss soft (c T ,c S )+(1-α)Loss hard (c T ,c gt )
where Loss_yolo represents the overall Loss of the YOLO model, λ represents the weight of the corresponding term, L CIoU Represents the regression loss of the boundary box, b represents the predicted value of the boundary box, b gt Representing the true value of the bounding box, L obj Representing target confidence loss, p o Representing the confidence score, p, of a target in a prediction box iou IoU value, L representing prediction frame and target frame corresponding thereto cls Representing class loss, c p Representing class score of prediction frame, c gt Representing the true value of a class, alpha represents the weight, b s Boundary box predicted value representing student model, b T Boundary box predicted value, p, representing teacher model S Representing the target confidence score, p, in a student model prediction frame T Representing the target confidence score, p, in the teacher model prediction frame gt Representing the true value of the target, c S Class score representing student model prediction box, c T A class score representing a teacher model prediction frame;
s32, compressing the intelligent unmanned aerial vehicle signal recognition initial model by using a characteristic distillation method based on an FSP matrix to obtain an optimized intelligent unmanned aerial vehicle signal recognition model;
the characteristic distillation method based on the FSP matrix comprises the following steps:
wherein F is 1 Is a shallow feature map, F 2 Is a deep feature map, h is the length of the feature map, w is the width of the feature map, i, j is F 1 、F 2 Corresponding channel indexes, x and W are current inputs and parameters; approximating distance L between FSP matrices of teacher model and student model using L2 Loss FSP (W t ,W s ) L2 Loss is defined as:
in the method, in the process of the invention,
FSP matrix for teacher network, +.>
FSP matrix lambda for student network i Represents weight, N is the number of data points, W t For teacher network parameters, W s For student network parameters, x is the current input, i is channel index, n is channel number;
the teacher model is a YOLOv5l model; the student model is a YOLOv5s model;
s33, training the intelligent recognition model of the optimized unmanned aerial vehicle signal by using the training signal sample database to obtain the intelligent recognition model of the unmanned aerial vehicle signal;
the identification module is used for receiving the space electromagnetic wave signal to be identified by utilizing the radio reconnaissance equipment, and processing the space electromagnetic wave signal to be identified by utilizing the unmanned aerial vehicle signal intelligent identification model to obtain an unmanned aerial vehicle signal intelligent identification result.
6. Unmanned aerial vehicle signal intelligent recognition device, its characterized in that, the device includes:
a memory storing executable program code;
a processor coupled to the memory;
the processor invokes the executable program code stored in the memory to perform the unmanned aerial vehicle signal intelligent recognition method of any of claims 1-4.
7. A computer-storable medium storing computer instructions that, when invoked, are used to perform the unmanned aerial vehicle signal intelligent recognition method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211350359.2A CN115577305B (en) | 2022-10-31 | 2022-10-31 | Unmanned aerial vehicle signal intelligent recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211350359.2A CN115577305B (en) | 2022-10-31 | 2022-10-31 | Unmanned aerial vehicle signal intelligent recognition method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115577305A CN115577305A (en) | 2023-01-06 |
| CN115577305B true CN115577305B (en) | 2023-05-30 |
Family
ID=84589416
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211350359.2A Active CN115577305B (en) | 2022-10-31 | 2022-10-31 | Unmanned aerial vehicle signal intelligent recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115577305B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116805291B (en) * | 2023-06-27 | 2024-02-09 | 中国人民解放军军事科学院系统工程研究院 | Unmanned aerial vehicle signal frequency band detection method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114154545A (en) * | 2021-12-07 | 2022-03-08 | 中国人民解放军32802部队 | Intelligent unmanned aerial vehicle measurement and control signal identification method under strong mutual interference condition |
| CN114241282A (en) * | 2021-11-04 | 2022-03-25 | 河南工业大学 | Knowledge distillation-based edge equipment scene identification method and device |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180268292A1 (en) * | 2017-03-17 | 2018-09-20 | Nec Laboratories America, Inc. | Learning efficient object detection models with knowledge distillation |
| CN113343796B (en) * | 2021-05-25 | 2022-04-05 | 哈尔滨工程大学 | Knowledge distillation-based radar signal modulation mode identification method |
-
2022
- 2022-10-31 CN CN202211350359.2A patent/CN115577305B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114241282A (en) * | 2021-11-04 | 2022-03-25 | 河南工业大学 | Knowledge distillation-based edge equipment scene identification method and device |
| CN114154545A (en) * | 2021-12-07 | 2022-03-08 | 中国人民解放军32802部队 | Intelligent unmanned aerial vehicle measurement and control signal identification method under strong mutual interference condition |
Non-Patent Citations (1)
| Title |
|---|
| Early Detection of Forest Fire Based on Unmaned Aerial Vehicle Platform;Xingsha Yang等;《2019 ICSIDP》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115577305A (en) | 2023-01-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108734723B (en) | Relevant filtering target tracking method based on adaptive weight joint learning | |
| CN110222760B (en) | Quick image processing method based on winograd algorithm | |
| CN109902763B (en) | Method and device for generating feature map | |
| CN109948700B (en) | Method and device for generating feature map | |
| CN114154545B (en) | Intelligent unmanned aerial vehicle measurement and control signal identification method under strong mutual interference condition | |
| CN110942472B (en) | Nuclear correlation filtering tracking method based on feature fusion and self-adaptive blocking | |
| CN115577305B (en) | Unmanned aerial vehicle signal intelligent recognition method and device | |
| CN113591978B (en) | Confidence penalty regularization-based self-knowledge distillation image classification method, device and storage medium | |
| US20190332894A1 (en) | Method for Processing Automobile Image Data, Apparatus, and Readable Storage Medium | |
| CN114449452A (en) | Indoor positioning algorithm for heterogeneous equipment | |
| CN113537085A (en) | Ship target detection method based on two-time transfer learning and data augmentation | |
| CN114627447A (en) | Road vehicle tracking method and system based on attention mechanism and multi-target tracking | |
| CN110378932B (en) | Correlation filtering visual tracking method based on spatial regularization correction | |
| CN114972439A (en) | Novel target tracking algorithm for unmanned aerial vehicle | |
| CN114527441A (en) | Radar signal identification method of LSTM network based on multi-head attention mechanism | |
| CN112364916B (en) | Image classification method based on transfer learning, related equipment and storage medium | |
| CN115600101B (en) | Priori knowledge-based unmanned aerial vehicle signal intelligent detection method and apparatus | |
| CN112261609A (en) | Millimeter wave internet of vehicles resource allocation method and device, electronic equipment and storage medium | |
| CN111539390A (en) | Small target image identification method, equipment and system based on Yolov3 | |
| CN116894207A (en) | Intelligent radiation source identification method based on Swin transducer and transfer learning | |
| CN114708307B (en) | Target tracking method, system, storage medium and device based on correlation filter | |
| CN114724245B (en) | Incremental learning human body action recognition method based on CSI | |
| CN113269808B (en) | Video small target tracking method and device | |
| CN113298136B (en) | Twin network tracking method based on alpha divergence | |
| CN112946567B (en) | Moving target fingerprint indoor positioning method based on domain antagonism neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |