CN113486960A - Unmanned aerial vehicle tracking method and device based on long-time memory neural network, storage medium and computer equipment - Google Patents
Unmanned aerial vehicle tracking method and device based on long-time memory neural network, storage medium and computer equipment Download PDFInfo
- Publication number
- CN113486960A CN113486960A CN202110784650.XA CN202110784650A CN113486960A CN 113486960 A CN113486960 A CN 113486960A CN 202110784650 A CN202110784650 A CN 202110784650A CN 113486960 A CN113486960 A CN 113486960A
- Authority
- CN
- China
- Prior art keywords
- neural network
- aerial vehicle
- unmanned aerial
- time
- training
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention provides an unmanned aerial vehicle tracking method based on a long-time memory neural network, which comprises the steps of obtaining an original data set of an unmanned aerial vehicle and classifying the data set; processing the original data set to obtain a training set; establishing a long-term memory tracking neural network of the unmanned aerial vehicle track; fully training the constructed tracking neural network and outputting the trained tracking neural network; collecting measurement data of an unmanned aerial vehicle to be tracked; inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation; the method is based on a large amount of real state data and measurement data of the target unmanned aerial vehicle, so that a stable and good prediction effect is achieved under the motion mode of the unmanned aerial vehicle, the target unmanned aerial vehicle tracking network matched with the radar is obtained through training, the learning capability of the trained network predicts parameters, and the tracking network obtained by the method has the advantages of high accuracy, wide application range, good practical effect and the like, and long-time high-precision target tracking is achieved.
Description
Technical Field
The invention relates to the technical field of fruit classification in machine vision, in particular to an unmanned aerial vehicle tracking method, an unmanned aerial vehicle tracking system, a storage medium and computer equipment based on a long-time and short-time memory neural network.
Background
In recent years, target tracking is a core key technology of radar data processing, a corresponding relation between each frame of measured data of radar and a target track can be established through track initiation, a comprehensive wave gate and track data interconnection, and a motion track and motion parameters of a target are updated in real time, so that the target is tracked. The unmanned aerial vehicle tracking algorithm belongs to a target tracking loop, comprises a state prediction part, a Kalman filtering part and the like, and is used as a core processing module for target tracking, and is very important. The existing target tracking algorithm is hidden behind a track starting module, and the target state is estimated in real time based on starting flight paths and real-time measurement. Common basic target tracking algorithms include a target tracking algorithm based on a constant velocity model, a target tracking algorithm based on a constant acceleration model, a target tracking algorithm based on a cooperative turning model, a target tracking algorithm based on an interactive multi-model and the like.
The disclosure number is CN110610512A, which provides an unmanned aerial vehicle target tracking method based on BP neural network fusion Kalman filtering algorithm, and by fusing BP neural network and Kalman filtering algorithm, the central position coordinates of a target when the target is shielded by an obstacle in the unmanned aerial vehicle target tracking process are accurately predicted, so that the stability of unmanned aerial vehicle target tracking is improved. Meanwhile, the existing prediction method has the problem that parameters cannot be determined, manual repeated modification and debugging are needed, and the optimized track prediction effect after debugging is difficult to achieve. In conclusion, the existing target tracking method has the problems of simple model, low complexity, poor universality, lack of learning ability and the like, and is difficult to realize high-precision target tracking for a long time.
Disclosure of Invention
The invention aims to provide an unmanned aerial vehicle tracking method, an unmanned aerial vehicle tracking system, a storage medium and computer equipment based on a long-time memory neural network, and aims to solve the problems in the background art.
In order to achieve the purpose, the invention provides the following technical scheme:
an unmanned aerial vehicle tracking method based on a long-time and short-time memory neural network comprises the following steps:
s1, acquiring an original data set of the unmanned aerial vehicle and classifying;
s2, processing the original data set to obtain a training set;
s3, constructing a neural network for memorizing and tracking the length of the unmanned aerial vehicle track;
s4, fully training the constructed tracking neural network and outputting the trained tracking neural network;
s5, collecting the measurement data of the unmanned aerial vehicle to be tracked;
and S6, inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation.
Preferably, the step S1 of acquiring the raw data set of the drone specifically includes the following steps:
s101, collecting measurement data and target real state data of the unmanned aerial vehicle through a radar;
s102, aggregating the collected data;
s103, comparing the condensed points with clutter values of one unit of the radar terminal, and if the amplitude of each point is more than three times of the clutter values of the unit, determining that the target can be batched;
s104, determining whether the unit clutter values at the subsequent time are updated or not according to the unit clutter values and the freezing factors;
s105, screening out a trace point associated with the flight path from all the measurements by using the angle information, wherein the trace point comprises angle limit and time limit;
and S106, performing dimension conversion on the trace point data processed in the steps to obtain a processed original data set.
Preferably, the classifying the original data set in step S1 includes classifying according to scene and time period.
Preferably, the step S2 specifically includes the following steps:
s201, scaling the data in equal proportion according to a Min-Max standard mode to finish standard pretreatment;
s202, storing the measurement data in the original data set into a one-dimensional matrix according to the measurement information corresponding to each time point, then storing the measurement information of the same time point subjected to multiple Monte Carlo simulations into a two-dimensional array, and finally storing the measurements corresponding to different time points into a three-dimensional array to obtain a training basic data set;
s203, storing the target real state information corresponding to each time point into a one-dimensional matrix, then storing the target real state information of the same time point subjected to multiple Monte Carlo simulations into a two-dimensional array, and finally storing the real states corresponding to different time points into a three-dimensional array to obtain a training label data set;
and S204, integrating the obtained training basic data and training label data under different scenes to obtain a three-dimensional matrix respectively comprising a complete training basic data set and a training label data set.
Preferably, the step S3 specifically includes the following steps:
s301, setting the number of layers of the neural network, the number of neurons in each layer and an adopted excitation function, and setting the number of layers of the neural network as an input layer, a hidden layer and an output layer so as to build a basic long-time and short-time neural network model;
s302, the number of nodes of an input layer is set as a measurement dimension, the number of nodes of an implicit layer is set as a target state dimension, and the number of nodes of an output layer is set as a target state dimension.
Preferably, the step S4 specifically includes the following steps:
s401, disordering the data of the training data set and the label data set to ensure that the corresponding relation is unchanged;
s402, setting training parameters of the network, setting a loss function as a mean square error function, setting an optimizer as adaptive moment estimation, setting batch size as 1 and setting batch as 100;
s403, inputting training data and training labels into a network for training, and calculating a loss value and accuracy;
s404, adjusting and modifying the Epoch and the optimizer model parameters of the tracking neural network, and repeatedly training;
and S405, finishing training and outputting the memory neural network when the loss value and the accuracy rate reach convergence.
Preferably, in step S5, the measurement data of the unmanned aerial vehicle navigation is collected by a radar, and the measurement data information includes a time sequence of the unmanned aerial vehicle and coordinate information corresponding to the time sequence.
In order to achieve the above object, the present invention further provides an unmanned aerial vehicle tracking system based on a long-time and short-time memory neural network, wherein the system includes:
the data acquisition and classification module is used for acquiring an original data set of the unmanned aerial vehicle and classifying the original data set;
the data set processing module is used for processing the original data set to obtain a training set;
the neural network construction module is used for constructing a long-time memory tracking neural network of the unmanned aerial vehicle track;
the network training module is used for fully training the constructed tracking neural network and outputting the trained tracking neural network;
the measurement data acquisition module is used for acquiring the measurement data of the unmanned aerial vehicle to be tracked; and;
and the target state estimation module is used for inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation.
In order to achieve the above object, the present invention further provides a long-term memory neural network-based storage medium for tracking a drone, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the drone tracking method can be implemented.
In order to achieve the above object, the present invention further provides a long-term memory neural network-based unmanned aerial vehicle tracking computing device, wherein the computing device includes: a memory, a processor, and a drone tracking algorithm program stored on the memory and executable on the processor, the drone tracking program configured to implement the steps of the drone tracking method described above.
Compared with the prior art, the invention has the beneficial effects that:
the target tracking algorithm based on the long-time memory neural network is based on a large amount of real state data and measured data of the target unmanned aerial vehicle, so that a stable and good prediction effect is achieved under the motion mode of the unmanned aerial vehicle, the target unmanned aerial vehicle tracking network matched with the radar is obtained through training, parameters are predicted through the learning capacity of the trained network, the tracking network obtained through the method has the advantages of high accuracy, wide application range, good practical effect and the like, long-time high-precision target tracking is achieved, and the target tracking algorithm is suitable for being popularized and used in practical application.
Drawings
Fig. 1 is a flow chart of a method of the unmanned aerial vehicle tracking method of the present invention;
fig. 2 is a schematic structural diagram of a system of the unmanned aerial vehicle tracking method of the present invention;
FIG. 3 is a schematic structural diagram of a long-term and short-term memory tracking neural network framework constructed by the present invention;
FIG. 4 is a detailed flow chart of the tracking method of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example (b):
referring to fig. 1 to 4, the present invention provides a technical solution:
an unmanned aerial vehicle tracking method based on a long-time and short-time memory neural network comprises the following steps:
and S1, acquiring the original data set of the unmanned aerial vehicle and classifying.
The method comprises the following steps of acquiring an original data set, collecting measured data and target real state data by utilizing radars in different scenes to form the original data set, wherein the step of acquiring the original data set of the unmanned aerial vehicle specifically comprises the following steps:
s101, collecting measurement data and target real state data of the unmanned aerial vehicle through a radar;
specifically, the collected radars should be radars with the same model and arrangement, so that the collection error caused by the radars is avoided; the collected data comprises measured point track data and tracking track data under different time periods, different areas and different motion models, abnormal data caused by radar faults or misoperation are eliminated, and collected normal point track data and tracking track data are obtained.
S102, aggregating the collected data;
specifically, in this embodiment, r02b is selected as the radar model for collecting data, and the limiting conditions during data aggregation are that the angle is greater than 8 degrees and the time is greater than 0.5 s; the data aggregation is to carry out envelope detection and multi-pulse non-coherent accumulation on original data acquired by a radar to obtain data after pulse accumulation, carry out constant false alarm detection and binary sliding window detection, treat the detected data as binary image data to obtain binary image data and a connected domain thereof, filter the first and second connected domains to obtain a residual connected domain containing a target trace and target information, and further realize aggregation of the target trace.
S103, comparing the condensed points with clutter values of one unit of the radar terminal, and if the amplitude of each point is more than three times of the clutter values of the unit, determining that the target can be batched;
the clutter value can be obtained by actually detecting the echo amplitude, the batched targets refer to the reserved targets, and otherwise, the target points are discarded.
S104, determining whether the unit clutter values at the subsequent time are updated or not according to the unit clutter values and the freezing factors;
and if the freezing factor is small, the point is considered as misjudged and discarded.
S105, screening out a trace point associated with the flight path from all the measurements by using the angle information, wherein the trace point comprises angle limit and time limit;
specifically, in the acquisition process, points acquired by the radar in one turn include different coordinate information, each point has corresponding angle information and corresponding time information, which subsequent point of the point determined as the target is determined according to the angle and time difference of the different points, and the different points are associated.
S106, carrying out dimension conversion on the trace point data processed in the steps to obtain a processed original data set;
the dimension conversion is to convert the unit of each data, i.e. the dimension, into a uniform dimension, so as to facilitate the uniformity of the data during subsequent processing.
The classification of the original data set comprises classification according to scenes and time periods, wherein the data can be classified according to different scenes and different time periods for the measured data and the target real state data of the original data set. In this embodiment, according to the classification of the collection time, the finally obtained data time period categories include three time periods of the morning, the middle and the evening of one day, and the scene categories include an open area and a crowded area.
And S2, processing the original data set to obtain a training set.
The method comprises the following steps of acquiring a training data set and a label data set, wherein the training data set comprises a basic data set and a label data set, a target measurement value corresponding to time is used as training data, a corresponding target real state value is used as the label data set, further, point trace data and track point trace data of an original data set are processed, real-time measurement forms a three-dimensional vector according to the sequence of batch, time sequence and measurement values and is used as a network input value, a target initial state value is used as a network hidden layer initial value according to the same sequence, and a target real-time state is used as a network output value, so that the training data set and the label data set are acquired, and the method specifically comprises the following steps:
s201, scaling the data in equal proportion according to a Min-Max standard mode to finish standard pretreatment;
the data normalization method can adopt methods such as max-min normalization, standard deviation normalization and regularization, and in this embodiment, the max-min normalization method is preferably used for processing, and the calculation formula is as follows:
wherein N isR+CRepresents the total number of samples of the training basis data and the training label data set, and x represents one of distance, azimuth angle, and elevation angle.
S202, storing the measurement data in the original data set into a one-dimensional matrix according to the measurement information corresponding to each time point, then storing the measurement information of the same time point subjected to multiple Monte Carlo simulations into a two-dimensional array, and finally storing the measurements corresponding to different time points into a three-dimensional array to obtain a training basic data set;
s203, storing the target real state information corresponding to each time point into a one-dimensional matrix, then storing the target real state information of the same time point subjected to multiple Monte Carlo simulations into a two-dimensional array, and finally storing the real states corresponding to different time points into a three-dimensional array to obtain a training label data set;
in this embodiment, the radar is a three-coordinate radar, so the ith trace sequence can be expressed as a vector:
[xi(1),xi(2),xi(3),...,xi(Ni)]，
wherein xi(n)＝[Ri(n),αi(n),θi(n)]，xi(N) information representing the nth trace in the trace sequence, including distance, azimuth, and pitch data, NiThe length of the point track sequence is represented, and a point track data set corresponding to one track point is composed of a plurality of point tracks.
The trace point data set form is expressed as a vector:
[R1(n),α1(n),θ1(n),R2(n),α2(n),θ2(n),...,Ri(Ni),αi(Ni),θi(Ni)]
and arranging the data sets corresponding to all the track points according to rows to form a training basic data set. Similarly, the track data set can be obtained in the form of:
[z1,z2,...,zn]
wherein z isn＝[Rn,αn,θn]。
S204, integrating the obtained training basic data and training label data under different scenes to obtain a three-dimensional matrix respectively comprising a complete training basic data set and a training label data set;
wherein, the training basic data and the training label data are in one-to-one correspondence.
And S3, constructing a neural network for memorizing and tracking the length of the unmanned aerial vehicle track.
Specifically, the constructed tracking neural network further includes a transmission form of the network, the number of network layers, the number of neurons in each layer, and an adopted excitation function, and specifically includes the following steps:
s301, setting the number of layers of the neural network, the number of neurons in each layer and the adopted excitation function, and setting the number of layers of the neural network as an input layer, a hidden layer and an output layer so as to build a basic long-time and short-time neural network model.
Specifically, the long-term and short-term memory tracking neural network constructed in this embodiment mainly has four use forms, namely one-to-one, many-to-one, one-to-many, and many-to-many, and the data structure required to be input in the unmanned aerial vehicle flight path prediction is multidimensional data, and the output is point trajectory data and also multidimensional data, so that the network form of the corresponding error back propagation neural network is a many-to-many form, and the output nodes of the many-to-many network are set as a plurality of nodes. The transmission form is set to be a many-to-many form, namely, a plurality of inputs correspond to a plurality of outputs.
The construction process can build the neural network by setting the number of layers of the neural network, the number of neurons in each layer, and the excitation function employed in the program.
S302, the number of nodes of an input layer is set as a measurement dimension, the number of nodes of an implicit layer is set as a target state dimension, and the number of nodes of an output layer is set as a target state dimension.
The neural network constructed in the embodiment is a three-layer structure, and the first layer is an input layer, namely a data receiving layer, and is used for receiving real-time data; the second layer is a hidden layer and is used for combining the hidden layer state at the previous moment and the input data at the current moment for processing and transmitting the estimated state of the target at the previous moment; the third layer is an output layer, each data in the process from the previous layer to the next layer is multiplied by a corresponding weight to calculate a result, the weight is given and adjusted by a model, and the output layer corresponds to the current estimation state of the output target according to the processed tracking problem.
And S4, fully training the constructed tracking neural network and outputting the trained tracking neural network.
Wherein, step S4 specifically includes the following steps:
s401, data of the training data set and the label data set are disturbed, and the corresponding relation is guaranteed to be unchanged.
S401, training parameters of the network are set, a loss function is set to be a mean square error function, an optimizer is set to be adaptive moment estimation, the batch size is set to be 1, and the batch size is set to be 100.
And S402, inputting the training data and the training labels into the network for training, and calculating the loss value and the accuracy.
Wherein, the loss value is calculated by a loss function, the calculation formula of the loss function in this embodiment is selected as,
the smaller the loss value, the higher the accuracy of the target state estimate.
And S403, adjusting and modifying the Epoch and the optimizer model parameters of the tracking neural network, and repeatedly training.
The Epoch refers to a process of sending all data into a network to complete one forward calculation and one backward propagation, when a complete data set passes through the neural network once and returns once, the process is called one Epoch, that is, all training samples are subjected to one forward propagation and one backward propagation in the neural network, and the optimizer is an algorithm for enabling the loss value of the loss function to be as small as possible through appropriate parameters.
And S404, finishing training and outputting the memory neural network when the loss value and the accuracy rate reach convergence.
And S5, collecting the measurement data of the unmanned aerial vehicle to be tracked.
The unmanned aerial vehicle navigation measurement data can be acquired through a radar, the acquired echo is converted into navigation measurement data to be processed through processing, and the measurement data information comprises the time sequence of the unmanned aerial vehicle and the corresponding coordinate information.
And S6, inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation.
The neural network outputs a result as a target estimation state corresponding to each time sequence, and the target estimation states of one time sequence can be connected into a complete track.
In order to achieve the above object, the present invention further provides an unmanned aerial vehicle tracking system based on a long-time and short-time memory neural network, wherein the system includes:
the data acquisition and classification module is used for acquiring an original data set of the unmanned aerial vehicle and classifying the original data set;
the data set processing module is used for processing the original data set to obtain a training set;
the neural network construction module is used for constructing a long-time memory tracking neural network of the unmanned aerial vehicle track;
the network training module is used for fully training the constructed tracking neural network and outputting the trained tracking neural network;
the measurement data acquisition module is used for acquiring the measurement data of the unmanned aerial vehicle to be tracked; and;
and the target state estimation module is used for inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation.
In order to achieve the above object, the present invention further provides a long-term memory neural network-based storage medium for tracking a drone, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the drone tracking method can be implemented.
The storage medium may include: any entity or device capable of carrying computer program code, recording medium, U.S. disk, removable hard disk, magnetic disk, optical disk, computer memory, Read Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution media, and the like. It should be noted that the computer readable medium may include any suitable increase or decrease as required by legislation and patent practice in a jurisdiction, for example, in some jurisdictions, in accordance with legislation and patent practice, the computer readable medium may not include electrical carrier signals and telecommunications signals
In order to achieve the above object, the present invention further provides a long-term memory neural network-based unmanned aerial vehicle tracking computing device, wherein the computing device includes: a memory, a processor, and a drone tracking algorithm program stored on the memory and executable on the processor, the drone tracking program configured to implement the steps of the drone tracking method described above.
The electronic terminal can be a desktop computer, an industrial computer, a numerical control device, an industrial robot, a server and other computing devices. It will be appreciated by those skilled in the art that the electronic terminal includes a processor and a memory, and the description of the memory for storing instructions is merely an example of the electronic terminal and is not a limitation of the electronic terminal, and may include more or less components, or combine certain components, or different components, for example, the electronic terminal may further include input and output devices, network access devices, buses, etc.
And (3) working results are as follows: the conventional unmanned aerial vehicle tracking method comprises tracking based on a current statistical model algorithm, tracking based on an interactive multi-model algorithm and the like, and in order to verify that the unmanned aerial vehicle track tracking method has higher accuracy and wider applicability in the aspect of tracking the unmanned aerial vehicle track compared with the conventional method, the following experiments are carried out.
The same flight data of the unmanned aerial vehicle are collected and tracked by respectively adopting a traditional method and the method provided by the embodiment, the tracking data is compared with the actual flight data of the unmanned aerial vehicle, and the detection accuracy is observed, wherein in the experiment, the traditional tracking method is based on an interactive multi-model algorithm, and the comparison result is shown in the following table 1:
table 1: result comparison table for coordinate prediction of unmanned aerial vehicle under different methods
According to the table, compared with the tracking result of the traditional method on the flight coordinate of the unmanned aerial vehicle, the prediction result obtained by adopting the tracking method provided by the embodiment is closer to the real value.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (10)
1. An unmanned aerial vehicle tracking method based on a long-time and short-time memory neural network is characterized by comprising the following steps:
s1, acquiring an original data set of the unmanned aerial vehicle and classifying;
s2, processing the original data set to obtain a training set;
s3, constructing a neural network for memorizing and tracking the length of the unmanned aerial vehicle track;
s4, fully training the constructed tracking neural network and outputting the trained tracking neural network;
s5, collecting the measurement data of the unmanned aerial vehicle to be tracked;
and S6, inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation.
2. The long-time and short-time memory neural network-based unmanned aerial vehicle tracking method according to claim 1, wherein the long-time and short-time memory neural network-based unmanned aerial vehicle tracking method comprises the following steps: the step S1 of acquiring the raw data set of the unmanned aerial vehicle specifically includes the following steps:
s101, collecting measurement data and target real state data of the unmanned aerial vehicle through a radar;
s102, aggregating the collected data;
s103, comparing the condensed points with clutter values of one unit of the radar terminal, and if the amplitude of each point is more than three times of the clutter values of the unit, determining that the target can be batched;
s104, determining whether the unit clutter values at the subsequent time are updated or not according to the unit clutter values and the freezing factors;
s105, screening out a trace point associated with the flight path from all the measurements by using the angle information, wherein the trace point comprises angle limit and time limit;
and S106, performing dimension conversion on the trace point data processed in the steps to obtain a processed original data set.
3. The long-time and short-time memory neural network-based unmanned aerial vehicle tracking method according to claim 2, wherein the long-time and short-time memory neural network-based unmanned aerial vehicle tracking method comprises the following steps: the classifying the raw data set in the step S1 includes classifying according to scene and time period.
4. The long-time and short-time memory neural network-based unmanned aerial vehicle tracking method according to claim 1, wherein the long-time and short-time memory neural network-based unmanned aerial vehicle tracking method comprises the following steps: the step S2 specifically includes the following steps:
s201, scaling the data in equal proportion according to a Min-Max standard mode to finish standard pretreatment;
s202, storing the measurement data in the original data set into a one-dimensional matrix according to the measurement information corresponding to each time point, then storing the measurement information of the same time point subjected to multiple Monte Carlo simulations into a two-dimensional array, and finally storing the measurements corresponding to different time points into a three-dimensional array to obtain a training basic data set;
s203, storing the target real state information corresponding to each time point into a one-dimensional matrix, then storing the target real state information of the same time point subjected to multiple Monte Carlo simulations into a two-dimensional array, and finally storing the real states corresponding to different time points into a three-dimensional array to obtain a training label data set;
and S204, integrating the obtained training basic data and training label data under different scenes to obtain a three-dimensional matrix respectively comprising a complete training basic data set and a training label data set.
5. The long-and-short-term memory neural network-based unmanned aerial vehicle tracking method according to claim, characterized in that: the step S3 specifically includes the following steps:
s301, setting the number of layers of the neural network, the number of neurons in each layer and an adopted excitation function, and setting the number of layers of the neural network as an input layer, a hidden layer and an output layer so as to build a basic long-time and short-time neural network model;
s302, the number of nodes of an input layer is set as a measurement dimension, the number of nodes of an implicit layer is set as a target state dimension, and the number of nodes of an output layer is set as a target state dimension.
6. The long-time and short-time memory neural network-based unmanned aerial vehicle tracking method according to claim 1, wherein the long-time and short-time memory neural network-based unmanned aerial vehicle tracking method comprises the following steps: the step S4 specifically includes the following steps:
s401, disordering the data of the training data set and the label data set to ensure that the corresponding relation is unchanged;
s402, setting training parameters of the network, setting a loss function as a mean square error function, setting an optimizer as adaptive moment estimation, setting batch size as 1 and setting batch as 100;
s403, inputting training data and training labels into a network for training, and calculating a loss value and accuracy;
s404, adjusting and modifying the Epoch and the optimizer model parameters of the tracking neural network, and repeatedly training;
and S405, finishing training and outputting the memory neural network when the loss value and the accuracy rate reach convergence.
7. The long-time and short-time memory neural network-based unmanned aerial vehicle tracking method according to claim 1, wherein the long-time and short-time memory neural network-based unmanned aerial vehicle tracking method comprises the following steps: in the step S5, the measurement data of the unmanned aerial vehicle navigation is collected by a radar, and the measurement data information includes the time sequence of the unmanned aerial vehicle and the coordinate information corresponding to the time sequence.
8. The utility model provides an unmanned aerial vehicle tracker based on long-time memory neural network which characterized in that: the system comprises:
the data acquisition and classification module is used for acquiring an original data set of the unmanned aerial vehicle and classifying the original data set;
the data set processing module is used for processing the original data set to obtain a training set;
the neural network construction module is used for constructing a long-time memory tracking neural network of the unmanned aerial vehicle track;
the network training module is used for fully training the constructed tracking neural network and outputting the trained tracking neural network;
the measurement data acquisition module is used for acquiring the measurement data of the unmanned aerial vehicle to be tracked; and;
and the target state estimation module is used for inputting the measured data into the trained tracking neural network to obtain the target real-time state estimation.
9. The utility model provides an unmanned aerial vehicle trails storage medium based on long-time memory neural network which characterized in that: the storage medium having stored thereon a computer program enabling, when executed by a processor, the steps of the drone tracking method of any one of claims 1 to 7.
10. The utility model provides an unmanned aerial vehicle tracks computing equipment based on long-time memory neural network which characterized in that: the computing device includes: memory, a processor, and a drone tracking algorithm program stored on the memory and executable on the processor, the drone tracking program configured to implement the steps of the drone tracking method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110784650.XA CN113486960A (en) | 2021-07-12 | 2021-07-12 | Unmanned aerial vehicle tracking method and device based on long-time memory neural network, storage medium and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110784650.XA CN113486960A (en) | 2021-07-12 | 2021-07-12 | Unmanned aerial vehicle tracking method and device based on long-time memory neural network, storage medium and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113486960A true CN113486960A (en) | 2021-10-08 |
Family
ID=77938640
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110784650.XA Pending CN113486960A (en) | 2021-07-12 | 2021-07-12 | Unmanned aerial vehicle tracking method and device based on long-time memory neural network, storage medium and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113486960A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445456A (en) * | 2021-12-23 | 2022-05-06 | 西北工业大学 | Data-driven intelligent maneuvering target tracking method and device based on partial model |
| CN117037006A (en) * | 2023-10-09 | 2023-11-10 | 山东中宇航空科技发展有限公司 | Unmanned aerial vehicle tracking method with high endurance capacity |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961310A (en) * | 2018-06-13 | 2018-12-07 | 北京清瑞维航技术发展有限公司 | Device, system, the medium, calculating device and method that automaticidentifying& tracking is realized |
| CN109740742A (en) * | 2019-01-14 | 2019-05-10 | 哈尔滨工程大学 | A kind of method for tracking target based on LSTM neural network |
| CN110940971A (en) * | 2019-11-06 | 2020-03-31 | 四川川大智胜软件股份有限公司 | Radar target point trace recording method and device and storage medium |
| CN111898749A (en) * | 2020-05-28 | 2020-11-06 | 南京理工雷鹰电子科技有限公司 | Unmanned aerial vehicle track prediction method based on neural network |
| CN111932583A (en) * | 2020-06-05 | 2020-11-13 | 西安羚控电子科技有限公司 | Space-time information integrated intelligent tracking method based on complex background |
-
2021
- 2021-07-12 CN CN202110784650.XA patent/CN113486960A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961310A (en) * | 2018-06-13 | 2018-12-07 | 北京清瑞维航技术发展有限公司 | Device, system, the medium, calculating device and method that automaticidentifying& tracking is realized |
| CN109740742A (en) * | 2019-01-14 | 2019-05-10 | 哈尔滨工程大学 | A kind of method for tracking target based on LSTM neural network |
| CN110940971A (en) * | 2019-11-06 | 2020-03-31 | 四川川大智胜软件股份有限公司 | Radar target point trace recording method and device and storage medium |
| CN111898749A (en) * | 2020-05-28 | 2020-11-06 | 南京理工雷鹰电子科技有限公司 | Unmanned aerial vehicle track prediction method based on neural network |
| CN111932583A (en) * | 2020-06-05 | 2020-11-13 | 西安羚控电子科技有限公司 | Space-time information integrated intelligent tracking method based on complex background |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445456A (en) * | 2021-12-23 | 2022-05-06 | 西北工业大学 | Data-driven intelligent maneuvering target tracking method and device based on partial model |
| CN117037006A (en) * | 2023-10-09 | 2023-11-10 | 山东中宇航空科技发展有限公司 | Unmanned aerial vehicle tracking method with high endurance capacity |
| CN117037006B (en) * | 2023-10-09 | 2023-12-15 | 山东中宇航空科技发展有限公司 | Unmanned aerial vehicle tracking method with high endurance capacity |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yi et al. | Multi-frame track-before-detect algorithm for maneuvering target tracking | |
| CN108254741B (en) | Target track prediction method based on cyclic neural network | |
| CN110865343B (en) | LMB-based particle filter tracking-before-detection method and system | |
| CN112613532B (en) | Moving target tracking method based on radar and cyclic neural network complement infrared fusion | |
| CN113486960A (en) | Unmanned aerial vehicle tracking method and device based on long-time memory neural network, storage medium and computer equipment | |
| CN111736145B (en) | Multi-maneuvering-target Doppler radar tracking method based on Gaussian mixed probability hypothesis density filtering | |
| CN110376582B (en) | Maneuvering target tracking method of self-adaptive GM-PHD | |
| CN111898749A (en) | Unmanned aerial vehicle track prediction method based on neural network | |
| CN104156984A (en) | PHD (Probability Hypothesis Density) method for multi-target tracking in uneven clutter environment | |
| CN104021519A (en) | Maneuvering multi-target tracking algorithm under dense clutter condition based on GPU architecture | |
| CN105761276A (en) | Iteration RANSAC-based adaptive birth target intensity estimation GM-PHD multi-target tracking algorithm | |
| CN111259332B (en) | Fuzzy data association method and multi-target tracking method in clutter environment | |
| CN115204212A (en) | Multi-target tracking method based on STM-PMBM filtering algorithm | |
| Chen et al. | Improved infrared small target detection and tracking method based on new intelligence particle filter | |
| Jiang et al. | A radar filtering model for aerial surveillance base on kalman filter and neural network | |
| CN112415468A (en) | DOA tracking method based on multi-Bernoulli filtering | |
| Ebert et al. | Deep radar sensor models for accurate and robust object tracking | |
| CN110045363B (en) | Multi-radar track association method based on relative entropy | |
| CN112114309A (en) | JPDA multi-target tracking method based on optimal contour coefficient self-adaptive K-means clustering | |
| Ren et al. | Improved shape-based distance method for correlation analysis of multi-radar data fusion in self-driving vehicle | |
| CN116340876A (en) | Spatial target situation awareness method for local multisource data fusion | |
| CN115508824A (en) | Multi-target big data association fusion tracking method and system | |
| CN115169136A (en) | Rapid UK-GMPHD multi-target tracking method in three-dimensional space | |
| CN114415157A (en) | Underwater target multi-model tracking method based on underwater acoustic sensor network | |
| CN114740467A (en) | Cluster target tracking and number and contour dynamic estimation method based on amplitude trace points |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211008 |
|
| RJ01 | Rejection of invention patent application after publication |