CN114663710A - Track recognition method, device, equipment and storage medium - Google Patents

Track recognition method, device, equipment and storage medium Download PDF

Info

Publication number
CN114663710A
CN114663710A CN202210443119.0A CN202210443119A CN114663710A CN 114663710 A CN114663710 A CN 114663710A CN 202210443119 A CN202210443119 A CN 202210443119A CN 114663710 A CN114663710 A CN 114663710A
Authority
CN
China
Prior art keywords
track
model
probability distribution
shape
trajectory
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
Application number
CN202210443119.0A
Other languages
Chinese (zh)
Inventor
战永盛
王鹏珍
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fifth Research Institute Of Telecommunications Technology Co ltd
Original Assignee
Fifth Research Institute Of Telecommunications Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Fifth Research Institute Of Telecommunications Technology Co ltd filed Critical Fifth Research Institute Of Telecommunications Technology Co ltd
Priority to CN202210443119.0A priority Critical patent/CN114663710A/en
Publication of CN114663710A publication Critical patent/CN114663710A/en
Pending legal-status Critical Current

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
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a track identification method, a track identification device, track identification equipment and a storage medium, wherein the method comprises the steps of constructing a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data; calling a shape model and a position model to respectively process a track to be recognized, and obtaining a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized; calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution; and determining the track type of the track to be identified based on the overall probability distribution. The method and the device identify the track by respectively modeling the track shape and the position characteristics, solve the problems of sparse high-dimensional data samples, weak manufacturing resistance, poor interpretability and the like when a single model is used, and can accurately and quickly provide identification results and bases for users.

Description

Track recognition method, device, equipment and storage medium
Technical Field
The present invention relates to the field of trajectory recognition technologies, and in particular, to a trajectory recognition method, apparatus, device, and storage medium.
Background
In recent years, various sensor observation means are rapidly developed and mature, and a large amount of data are gradually accumulated, wherein the track type data of airplanes, ships, submarines and the like are also rapidly increased. Although more noise exists in the original data of the track due to various condition limitations, a large amount of valuable information can be still mined by reasonably utilizing the algorithm. The method is a significant research on how to utilize the historical track information rule to identify the type of the currently observed track, so that when massive historical data are faced, the most possible identification result can be automatically solved by a machine learning model, and the identification basis and detailed probability distribution are shown. The algorithm model is based on a strict statistical formula, can effectively eliminate human subjective bias and provides quick and high-value auxiliary information for manual evidence judgment.
The existing track identification method is mainly based on deep learning and statistical methods. The methods based on deep learning include LSTM, CNN and the like, and have the problems of sparse data and incapability of effectively learning characteristic details due to extremely wide and unbalanced track distribution areas, and the deep learning interpretability is poor, so that a detailed prediction basis cannot be given. The statistical method is used for describing information such as a polygonal area or thermodynamic diagram externally connected with a track set, and the problems that the effect is poor when different track areas are overlapped, the shape information of the track can not be utilized and the like exist.
The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.
Disclosure of Invention
The invention mainly aims to provide a track identification method, a track identification device, track identification equipment and a storage medium, and aims to solve the technical problem that the existing track identification efficiency is not high.
In order to achieve the above object, the present invention provides a trajectory recognition method, including the steps of:
constructing a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data;
calling a shape model and a position model to respectively process a track to be recognized, and obtaining a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized;
calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution;
and determining the track type of the track to be identified based on the overall probability distribution.
Optionally, the historical track data includes a plurality of tracks and category labels corresponding to the tracks, the tracks are composed of a plurality of track points, and the track points include longitudes, latitudes and timestamps of the track points.
Optionally, constructing the shape model specifically includes:
training an LSTM self-encoder by using a track point in historical track data, and extracting characteristic data of a track shape output by an encoding layer of the LSTM self-encoder;
and constructing a binary model for each class label, inputting the characteristic data into the binary model, and training the binary model to obtain a shape model corresponding to each class label.
Optionally, constructing the location model specifically includes:
extracting corresponding track points in the observation time stamps according to the observation time stamps of the tracks corresponding to the category labels, and constructing a sample set;
and training a Gaussian mixture model by using the sample set to obtain a position model corresponding to each class label.
Optionally, before the step of calling the shape model and the position model to process the trajectory to be recognized, the method further includes determining prior probability distribution P of all category labels according to the number of trajectory points corresponding to each category labelPre。
Optionally, the obtaining, according to the first probability distribution and the second probability distribution, an expression of the overall probability distribution of the trajectory to be identified is specifically:
PFinal＝w1*PShape+w2*PLoc，s.t.w1+w2＝1
wherein, PLoc＝PLoc-co*PPre，PShape＝PShape-co*PPre，PShaPe-coA first distribution of probability, P, corresponding to the shape modelShape-coA second probability distribution, w, corresponding to the position model1、w2Is a weight value.
Optionally, the method further includes preprocessing the historical trajectory data and/or the trajectory to be recognized; wherein the preprocessing comprises performing uniform interpolation on the trajectory.
Further, in order to achieve the above object, the present invention also provides a trajectory recognition device including:
the building module is used for building a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data;
the device comprises a calling module, a shape recognition module and a position recognition module, wherein the calling module is used for calling a shape model and a position model to respectively process a track to be recognized so as to obtain a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized;
the calculation module is used for calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution;
and the determining module is used for determining the track type of the track to be identified based on the overall probability distribution.
In addition, in order to achieve the above object, the present invention also provides a trajectory recognition device, including: the track recognition system comprises a memory, a processor and a track recognition program which is stored on the memory and can run on the processor, wherein the track recognition program realizes the steps of the track recognition method when being executed by the processor.
In order to achieve the above object, the present invention also provides a storage medium having a trajectory recognition program stored thereon, the trajectory recognition program implementing the steps of the trajectory recognition method described above when executed by a processor.
The embodiment of the invention provides a track identification method, a track identification device, track identification equipment and a storage medium, wherein the method comprises the steps of constructing a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data; calling a shape model and a position model to respectively process a track to be recognized, and obtaining a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized; calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution; and determining the track type of the track to be identified based on the overall probability distribution. According to the invention, the track is identified by respectively modeling the track shape and the position characteristics, so that the problems of sparse high-dimensional data samples, weak anti-manufacturing capability, poor interpretability and the like when a single model is used are solved, and the identification result and basis can be accurately and quickly provided for a user.
Drawings
FIG. 1 is a schematic structural diagram of a trajectory recognition device according to the present invention;
FIG. 2 is a schematic flow chart illustrating a trajectory recognition method according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of a track recognition embodiment of the track recognition method of the present invention;
FIG. 4 is a diagram of a shape model architecture of the present invention;
FIG. 5 is a comparison of the time factor considered and the time factor not considered for the location modeling of the present invention;
fig. 6 is a block diagram of the track recognition device according to the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
The existing track identification method is mainly based on deep learning and statistical methods. The methods based on deep learning include LSTM, CNN and the like, and have the problems of sparse data and incapability of effectively learning characteristic details due to extremely wide and unbalanced track distribution areas, and the deep learning interpretability is poor, so that a detailed prediction basis cannot be given. The statistical method is used for describing information such as a polygonal area or thermodynamic diagram externally connected with a track set, and the problems that the effect is poor when different track areas are overlapped, the shape information of the track can not be utilized and the like exist.
To solve this problem, various embodiments of the trajectory recognition method of the present invention are proposed. The track identification method provided by the invention identifies the track by respectively modeling the track shape and the position characteristics, solves the problems of sparse high-dimensional data samples, weak manufacturing resistance, poor interpretability and the like when a single model is used, and can accurately and quickly provide identification results and bases for users.
Referring to fig. 1, fig. 1 is a schematic structural diagram of a trajectory recognition device according to an embodiment of the present invention.
The device may be a User Equipment (UE) such as a Mobile phone, a smart phone, a laptop, a digital broadcast receiver, a Personal Digital Assistant (PDA), a tablet computer (PAD), etc., a handheld device, a vehicle-mounted device, a wearable device, a computing device or other processing device connected to a wireless modem, a Mobile Station (MS), etc., for trajectory recognition. The device may be referred to as a user terminal, portable terminal, desktop terminal, etc.
Generally, the apparatus comprises: at least one processor 301, a memory 302, and a trajectory recognition program stored on the memory and executable on the processor, the trajectory recognition program being configured to implement the steps of the trajectory recognition method as previously described.
The processor 301 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and so on. The processor 301 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). The processor 301 may also include a main processor and a coprocessor, where the main processor is a processor for processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 301 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. The processor 301 may further include an AI (Artificial Intelligence) processor for processing relevant trajectory recognition operations so that the trajectory recognition model may train learning autonomously, improving efficiency and accuracy.
Memory 302 may include one or more computer-readable storage media, which may be non-transitory. Memory 302 may also include high speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 302 is used to store at least one instruction for execution by processor 301 to implement the trajectory recognition methods provided by method embodiments herein.
In some embodiments, the terminal may further include: a communication interface 303 and at least one peripheral device. The processor 301, the memory 302 and the communication interface 303 may be connected by a bus or signal lines. Various peripheral devices may be connected to communication interface 303 via a bus, signal line, or circuit board. Specifically, the peripheral device includes: at least one of radio frequency circuitry 304, a display screen 305, and a power source 306.
The communication interface 303 may be used to connect at least one peripheral device related to I/O (Input/Output) to the processor 301 and the memory 302. The communication interface 303 is used for receiving the movement tracks of the plurality of mobile terminals uploaded by the user and other data through the peripheral device. In some embodiments, processor 301, memory 302, and communication interface 303 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 301, the memory 302 and the communication interface 303 may be implemented on a single chip or circuit board, which is not limited in this embodiment.
The Radio Frequency circuit 304 is used for receiving and transmitting RF (Radio Frequency) signals, also called electromagnetic signals. The radio frequency circuit 304 communicates with a communication network and other communication devices through electromagnetic signals, so as to obtain the movement tracks and other data of a plurality of mobile terminals. The rf circuit 304 converts an electrical signal into an electromagnetic signal to transmit, or converts a received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit 304 comprises: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and so forth. The radio frequency circuitry 304 may communicate with other terminals via at least one wireless communication protocol. The wireless communication protocols include, but are not limited to: metropolitan area networks, various generation mobile communication networks (2G, 3G, 4G, and 5G), Wireless local area networks, and/or WiFi (Wireless Fidelity) networks. In some embodiments, the rf circuit 304 may further include NFC (Near Field Communication) related circuits, which are not limited in this application.
The display screen 305 is used to display a UI (User Interface). The UI may include graphics, text, icons, video, and any combination thereof. When the display screen 305 is a touch display screen, the display screen 305 also has the ability to capture touch signals on or over the surface of the display screen 305. The touch signal may be input to the processor 301 as a control signal for processing. At this point, the display screen 305 may also be used to provide virtual buttons and/or a virtual keyboard, also referred to as soft buttons and/or a soft keyboard. In some embodiments, the display screen 305 may be one, the front panel of the electronic device; in other embodiments, the display screens 305 may be at least two, respectively disposed on different surfaces of the electronic device or in a folded design; in still other embodiments, the display screen 305 may be a flexible display screen disposed on a curved surface or a folded surface of the electronic device. Even further, the display screen 305 may be arranged in a non-rectangular irregular figure, i.e. a shaped screen. The Display screen 305 may be made of LCD (liquid crystal Display), OLED (Organic Light-Emitting Diode), and the like.
The power supply 306 is used to power various components in the electronic device. The power source 306 may be alternating current, direct current, disposable or rechargeable. When the power source 306 includes a rechargeable battery, the rechargeable battery may support wired or wireless charging. The rechargeable battery may also be used to support fast charge technology.
Those skilled in the art will appreciate that the configuration shown in fig. 1 does not constitute a limitation of the trajectory recognition device, and may include more or less components than those shown, or some components in combination, or a different arrangement of components.
An embodiment of the present invention provides a track identification method, and referring to fig. 2, fig. 2 is a schematic flow chart of an embodiment of the track identification method of the present invention.
In this embodiment, the track identification method includes the following steps:
step S100, constructing a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data.
Specifically, in practical application, the historical track data comprises a plurality of tracks and category labels corresponding to the tracks, each track is composed of a plurality of track points, and each track point comprises longitude, latitude and a timestamp of the track point.
When the shape model is constructed, the LSTM self-encoder is trained by using track points in historical track data, and characteristic data of track shapes output by a coding layer of the LSTM self-encoder are extracted; and constructing a binary model for each class label, inputting the characteristic data into the binary model, and training the binary model to obtain a shape model corresponding to each class label.
In the embodiment, the problem of no negative examples of the track is solved through the self-encoder and the random feature data when the shape is modeled. If the track is directly input into the LSTM-like neural network two classifier, the track is not long enough, the shape space is huge, and the shape of other target tracks can be similar to the target, so that the track covering the whole negative example is difficult to generate. The track is converted into the characteristic data with fixed length through the encoder, and positive and negative samples are generated from the characteristic data, so that the problem is solved.
When a position model is constructed, extracting corresponding track points in an observation time stamp according to the observation time stamp of the track corresponding to each category label, and constructing a sample set; and training a Gaussian mixture model by using the sample set to obtain a position model corresponding to each class label.
In this embodiment, in modeling the position, time dimension information is considered, for example, the historical occurrence times of the target in the two areas AB are the same, but the area a is 5 years ago, and the area B is in the near future, which is obviously more likely to occur in the area B next time, that is, the probability density of the gaussian mixture model in the area B should be greater than that in the area a, rather than equal to that in the area a. Different weights are introduced to the samples through the timestamp information, and the improved model is more consistent with empirical facts.
It should be noted that there are n categories in the history data, which correspond to n tags, and each tag corresponds to one or more tracks, and because the track point densities in each track are different, and the track point densities in each area in the same track may also be different, it is necessary to perform uniform distance interpolation on all track data.
It will be readily appreciated that the use of historical data to model the shape of the trajectory includes the steps of:
A. and performing data enhancement on all track data in a specific mode: and adding a small-range Gaussian noise disturbance to each track point position, and randomly selecting a track sequence subset to form a new data set NS. And (3) shifting all track starting points in the NS to (0,0) and keeping the numerical difference of other points in the track relative to the starting points unchanged.
B. And establishing an LSTM self-encoder, inputting and outputting longitude and latitude sequences of track points, and obtaining characteristic data of the track shape by using an encoding layer of the LSTM self-encoder after training.
C. And respectively establishing a binary model for each label, inputting data as characteristic data of the track, and outputting the probability of the label. When the binary model is trained, the positive example of the data set is the characteristic data corresponding to the track under the label in the data set NS, and the reverse example of the data set is random characteristic data;
D. the shape models used in the estimation are LSTM self-encoder encoding part serial binary models, and since there are n types, there are n binary models, that is, n shape models.
Modeling the location of the trajectory using historical data, comprising the steps of:
A. for a track data set under a certain label, the observation time stamp of each track is obtained and applied to corresponding track points, the track points are taken as a sample set, and the corresponding weight of a sample with a smaller time stamp can be reduced.
B. And carrying out position modeling on the sample set, wherein a two-dimensional Gaussian mixture model based on a Dirichlet process is selected, and the weight data of the samples are applied during training.
C. And processing all the labels by the steps to obtain n position models.
Step S200, calling a shape model and a position model to respectively process a track to be recognized, and obtaining a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized.
It should be noted that after the shape model and the position model are constructed and obtained, the prior probability distribution P of all the category labels needs to be determined according to the number of the trace points corresponding to each category labelPre。
Specifically, the prior probability of the category is calculated through quantity statistics, and the specific method is that the total track point number is tn, the track point number corresponding to a certain label is tx, and the prior probability of the label is tx/tn. Let the probability distribution of all the prior probability components of the labels bePPre。
Meanwhile, when the track to be identified is utilized, uniform distance interpolation can be carried out on all track data.
After the shape model and the position model are obtained, the trajectory to be recognized can be processed by utilizing the shape model and the position model, so that a first probability distribution of the trajectory to be recognized and a second probability distribution of trajectory points in the trajectory to be recognized are obtained.
And step S300, calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution.
In the present embodiment, after obtaining the prior probability, the first probability distribution, and the second probability distribution, the probability distribution of each category to which the trajectory belongs is calculated using the bayesian method. According to the first probability distribution and the second probability distribution, the expression for obtaining the overall probability distribution of the track to be identified is specifically as follows:
PFinal＝w1*PShape+w2*PLoc，s.t.w1+w2＝1
wherein, PLoc＝PLoc-co*PPre，PShape＝PShape-co*PPre，PShape-coA first distribution of probability, P, corresponding to the shape modelShape-coA second probability distribution, w, corresponding to the position model1、w2Is a weight value.
Specifically, the method for calculating the probability distribution of each category to which the track belongs by using the Bayesian method comprises the following steps:
A. carrying out uniform distance interpolation pretreatment on the track to be identified;
B. calculating conditional probability distribution of the track to be recognized according to the n shape models, and setting the conditional probability distribution as PShape-co。
C. All points of the track to be identified are respectively brought into each position model, each model obtains a plurality of probability density values, logarithm sums are respectively obtained, the n logarithm sums are converted into probability distribution, and the probability distribution is set as PLoc-co。
D. Posterior where the trajectory belongs to each model according to Bayes' formulaProbability is proportional to the prior probability multiplied by the conditional probability, let PLoc＝PLoc-co*PPre，PShape＝PShape-co*PPreAnd is combined with PLoc、PShapeRespectively converting the probability distribution into a probability distribution form with the sum of 1, namely the posterior probability that the track belongs to the position and shape model.
E. Weighting w to shape model and position model1、w2The final probability distribution is:
PFinal＝w1*PShape+w2*PLoc，s.t.w1+w2＝1。
and step S400, determining the track type of the track to be identified based on the overall probability distribution.
Specifically, after the overall probability distribution of the trajectory to be recognized is obtained, the trajectory to be recognized can be confirmed according to the trajectory type corresponding to the maximum probability value in the overall probability distribution.
In the embodiment, the track is described by respectively modeling from two angles of position and shape through a track identification method based on an LSTM and Gaussian mixture model, so that the problem that a single model cannot be effectively learned due to overlarge track space range and sparse data is solved. Compared with other existing deep learning models, the scheme provided by the embodiment is strong in interpretability, wherein the position model can intuitively observe an inference process through a two-dimensional probability density map.
In order to more clearly illustrate the present invention, the following description will be given by way of a specific example of trajectory recognition.
As shown in fig. 3, a trajectory recognition method based on LSTM and gaussian mixture model is provided, which includes the following steps:
A. and (4) preprocessing data, and performing uniform distance interpolation on all track data. Specifically, the spacing distance d is specified, and if the spacing between two adjacent points in the track is greater than or equal to (m +0.5) × d and less than (m +1.5) × d, m +1 points are uniformly inserted between the two points, so that the spacing between all adjacent track points is consistent as much as possible on the premise of ensuring that the track shape is unchanged.
B. The shape of the track is modeled by using historical data, as shown in fig. 4, an LSTM self-encoder is trained, input and output are the same track data, and the track positions are moved to the origin of coordinates in a unified manner, so that the track shape is learned, that is, an indefinite-length track is converted into a fixed-length shape characterization vector. The overall shape model connects the LSTM encoder to the neural network two-classifier, corresponding label probabilities are output by a Sigmod function, each label trains one two-classifier, and then there are n shape models, and their LSTM self-encoding parts are shared. Because it is very difficult to generate the negative sample track, the random feature with a fixed length is directly generated as the negative sample for training, thereby solving the problem.
C. And (3) performing position modeling, wherein a Gaussian mixture model based on a Dirichlet process is selected, and compared with a common Gaussian mixture model, the algorithm has the characteristic of automatically determining parameters according to the characteristics of a data set, so that the problem of over-fitting or under-fitting of different data sets caused by manually specifying the parameters is solved. In addition, weighting factors are added in the model training process, namely when the model calculates the mean value and the variance, the sample value of each sample is changed into the sample value weighting, and the number of samples is changed into the sum of the weighting. As shown in fig. 5, when trajectory point data is subjected to gaussian distribution fitting, the longer the data timestamp is, the lower the corresponding weight is, and the probability density after fitting is further influenced, so that the probability density is closer to the real situation. When the position model is applied for calculation, each point in the trajectory is brought into the model to obtain a series of probability density values, and the values are multiplied to obtain the probability of generating the trajectory for the model. The value is small so the continuous multiplication is changed to a logarithmic sum.
D. Calculating probability distribution of each category to which the track belongs by using a Bayes method, namely, taking the occurrence probability of the label into consideration on the basis of a model result, wherein the posterior probability of the track belonging to each model is proportional to the prior probability multiplied by the conditional probability according to a Bayes formula, and setting PLoc＝PLoc-co*PPre，PShape＝PShape-co*PPreAnd is combined with PLoc、PShapeSeparately converted into probability distribution form, i.e. scaled by equal ratio to make the number thereinThe sum is 1. The posterior probability that the trajectory belongs to the position and shape model is obtained. And (3) giving weight to the shape model and the position model, wherein the final probability distribution is as follows:
PFinal＝w1*PShape+w2*PLoc，s.t.w1+w2＝1
wherein w1、w2Can be specified by experts, and can also be learned by models through optimization algorithms.
In the embodiment, the shape and the position of the track are respectively modeled by using historical data, the prior probability of the category is calculated by quantity statistics, and finally the probability distribution of each category to which the track belongs is calculated by using a Bayesian method and is displayed to a user by an image inference process. The invention realizes the track identification method based on the LSTM and Gaussian mixture model, solves the problems of sparse high-dimensional data samples, weak manufacturing resistance, poor interpretability and the like when a single model is used by respectively modeling the track shape and the position characteristics, and can accurately and quickly provide identification results and bases for users.
Referring to fig. 6, fig. 6 is a block diagram of a track recognition device according to an embodiment of the present invention.
As shown in fig. 6, the trajectory recognition apparatus according to the embodiment of the present invention includes:
a construction module 10 for constructing a shape model and a position model based on the historical trajectory data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data;
the calling module 20 is configured to call the shape model and the position model to respectively process the trajectory to be recognized, so as to obtain a first probability distribution of the trajectory to be recognized and a second probability distribution of trajectory points in the trajectory to be recognized;
a calculating module 30, configured to calculate an overall probability distribution of the trajectory to be identified according to the first probability distribution and the second probability distribution;
and the determining module 40 is used for determining the track type of the track to be identified based on the overall probability distribution.
In the embodiment, a track recognition device is provided, which respectively models the track shape and the position characteristics to recognize the track, so that the problems of sparse high-dimensional data samples, weak anti-manufacturing capability, poor interpretability and the like when a single model is used are solved, and a recognition result and a basis can be accurately and quickly provided for a user.
Other embodiments or specific implementation manners of the track recognition device of the present invention may refer to the above method embodiments, and are not described herein again.
Furthermore, an embodiment of the present invention further provides a storage medium, where a trajectory recognition program is stored, and the trajectory recognition program, when executed by a processor, implements the steps of the trajectory recognition method as described above. Therefore, a detailed description thereof will be omitted. In addition, the beneficial effects of the same method are not described in detail. For technical details not disclosed in embodiments of the computer-readable storage medium referred to in the present application, reference is made to the description of embodiments of the method of the present application. It is determined that, by way of example, the program instructions may be deployed to be executed on one computing device or on multiple computing devices at one site or distributed across multiple sites and interconnected by a communication network.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
It should be noted that the above-described embodiments of the apparatus are merely schematic, where the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. In addition, in the drawings of the embodiment of the apparatus provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, and may be specifically implemented as one or more communication buses or signal lines. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that the present invention may be implemented by software plus necessary general hardware, and may also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components and the like. Generally, functions performed by computer programs can be easily implemented by corresponding hardware, and specific hardware structures for implementing the same functions may be various, such as analog circuits, digital circuits, or dedicated circuits. However, the implementation of a software program is a more preferable embodiment for the present invention. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a readable storage medium, such as a floppy disk, a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk of a computer, and includes instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods according to the embodiments of the present invention.

Claims (10)

1. A trajectory recognition method, comprising:
constructing a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data;
calling a shape model and a position model to respectively process a track to be recognized, and obtaining a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized;
calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution;
and determining the track type of the track to be identified based on the overall probability distribution.
2. The trajectory recognition method of claim 1, wherein the historical trajectory data includes a plurality of trajectories and category labels corresponding to the trajectories, the trajectories are formed by a plurality of trajectory points, and the trajectory points include longitudes, latitudes, and timestamps of the trajectory points.
3. The trajectory recognition method of claim 2, wherein constructing the shape model specifically comprises:
training the LSTM self-encoder by using a track point in the historical track data, and extracting the characteristic data of the track shape output by the encoding layer of the LSTM self-encoder;
and constructing a binary model for each class label, inputting the characteristic data into the binary model, and training the binary model to obtain a shape model corresponding to each class label.
4. The trajectory recognition method of claim 2, wherein constructing the location model specifically comprises:
extracting corresponding track points in the observation time stamps according to the observation time stamps of the tracks corresponding to the category labels, and constructing a sample set;
and training a Gaussian mixture model by using the sample set to obtain a position model corresponding to each class label.
5. The trajectory recognition method of claim 2, wherein, before the step of processing the trajectory to be recognized by calling the shape model and the position model respectively,the method also comprises the step of determining the prior probability distribution P of all the category labels according to the number of the track points corresponding to each category labelPre。
6. The trajectory recognition method according to claim 5, wherein the expression for obtaining the overall probability distribution of the trajectory to be recognized according to the first probability distribution and the second probability distribution is specifically:
PFinal＝w1*PShape+w2*PLoc，s.t.w1+w2＝1
wherein, PLoc＝PLoc-co*PPre，PShape＝PShape-co*PPre，PShape-coA first distribution of probability, P, corresponding to the shape modelShape-coA second probability distribution, w, corresponding to the position model1、w2Is a weight value.
7. The trajectory recognition method according to any one of claims 1 to 6, characterized in that it further comprises preprocessing the historical trajectory data and/or the trajectory to be recognized; wherein the preprocessing comprises performing uniform interpolation on the trajectory.
8. A trajectory recognition device, characterized in that the trajectory recognition device comprises:
the building module is used for building a shape model and a position model based on historical track data; the shape model is a classification model of a track shape in historical track data, and the position model is a classification model of track points in the historical track data;
the device comprises a calling module, a shape recognition module and a position recognition module, wherein the calling module is used for calling a shape model and a position model to respectively process a track to be recognized so as to obtain a first probability distribution of the track to be recognized and a second probability distribution of track points in the track to be recognized;
the calculation module is used for calculating the overall probability distribution of the track to be identified according to the first probability distribution and the second probability distribution;
and the determining module is used for determining the track type of the track to be identified based on the overall probability distribution.
9. A trajectory recognition device characterized by comprising: memory, a processor and a trajectory recognition program stored on the memory and executable on the processor, the trajectory recognition program, when executed by the processor, implementing the steps of the trajectory recognition method according to any one of claims 1 to 7.
10. A storage medium, characterized in that the storage medium has stored thereon a trajectory recognition program which, when executed by a processor, implements the steps of the trajectory recognition method according to any one of claims 1 to 7.
CN202210443119.0A 2022-04-25 2022-04-25 Track recognition method, device, equipment and storage medium Pending CN114663710A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210443119.0A CN114663710A (en) 2022-04-25 2022-04-25 Track recognition method, device, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210443119.0A CN114663710A (en) 2022-04-25 2022-04-25 Track recognition method, device, equipment and storage medium

Publications (1)

Publication Number Publication Date
CN114663710A true CN114663710A (en) 2022-06-24

Family

ID=82036370

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210443119.0A Pending CN114663710A (en) 2022-04-25 2022-04-25 Track recognition method, device, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN114663710A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117372969A (en) * 2023-12-08 2024-01-09 暗物智能科技(广州)有限公司 Monitoring scene-oriented abnormal event detection method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117372969A (en) * 2023-12-08 2024-01-09 暗物智能科技(广州)有限公司 Monitoring scene-oriented abnormal event detection method
CN117372969B (en) * 2023-12-08 2024-05-10 暗物智能科技(广州)有限公司 Monitoring scene-oriented abnormal event detection method

Similar Documents

Publication Publication Date Title
CN110364144B (en) Speech recognition model training method and device
US11307864B2 (en) Data processing apparatus and method
CN112100841B (en) Method and device for predicting attitude of shield tunneling machine, terminal equipment and storage medium
CN110147533B (en) Encoding method, apparatus, device and storage medium
CN114881379B (en) Method, device and equipment for estimating power generation capacity of wind power plant and storage medium
CN110147532B (en) Encoding method, apparatus, device and storage medium
CN112990390A (en) Training method of image recognition model, and image recognition method and device
CN114663710A (en) Track recognition method, device, equipment and storage medium
CN116543076A (en) Image processing method, device, electronic equipment and storage medium
CN115758893A (en) Method, device, equipment and storage medium for predicting aerodynamic coefficient of overhead ice-coated wire
CN116580211B (en) Key point detection method, device, computer equipment and storage medium
CN112862021A (en) Content labeling method and related device
CN115035313B (en) Black-neck crane identification method, device, equipment and storage medium
CN114862000A (en) Trajectory prediction method, apparatus, device and storage medium
CN114610911B (en) Multi-modal knowledge intrinsic representation learning method, device, equipment and storage medium
CN114969360A (en) Text information processing method, device, equipment and storage medium
CN114252100B (en) Online calibration method, device, equipment and storage medium for sensor
CN113282925B (en) Malicious file detection method, malicious file detection device, terminal equipment and storage medium
CN114972861A (en) Countermeasure sample generation method, device, equipment and storage medium
CN117852624B (en) Training method, prediction method, device and equipment of time sequence signal prediction model
CN113886688B (en) Method, device, terminal equipment and storage medium for predicting association relation of objects
CN117852624A (en) Training method, prediction method, device and equipment of time sequence signal prediction model
CN115701866B (en) E-commerce platform risk identification model training method and device
CN117115596B (en) Training method, device, equipment and medium of object action classification model
CN114511824A (en) Method, device and equipment for detecting quality of images shot by illegal vehicles and storage medium

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