Trajectory recognition method, device, equipment and storage medium

Abstract

The application discloses a trajectory recognition method, device and equipment and a storage medium, and comprises the following steps: constructing a shape model and a position model based on historical trajectory data; the shape model is a classification model of trajectory shapes in the historical trajectory data, and the position model is a classification model of trajectory points in the historical trajectory data; calling the shape model and the position model to process a to-be-recognized trajectory respectively, obtaining a first probability distribution of the to-be-recognized trajectory and a second probability distribution of trajectory points in the to-be-recognized trajectory; calculating an overall probability distribution of the to-be-recognized trajectory according to the first probability distribution and the second probability distribution; and determining a trajectory type of the to-be-recognized trajectory based on the overall probability distribution. The application models trajectory shapes and position characteristics respectively, and then recognizes the trajectory, solves problems such as high-dimensional data sample sparsity, weak anti-fabrication capability and poor interpretability when a single model is used, and can accurately and quickly provide a user with a recognition result and basis.

Description

Technical Field

[0001] This invention relates to the field of trajectory recognition technology, and in particular to a trajectory recognition method, apparatus, device, and storage medium. Background Technology

[0002] In recent years, various sensor observation methods have developed rapidly and become increasingly mature, accumulating a vast amount of data, including rapidly increasing data on the trajectories of aircraft, ships, and submarines. Although the raw trajectory data contains considerable noise due to various limitations, a wealth of valuable information can still be extracted through the proper use of algorithms. One significant area of ​​research is how to utilize historical trajectory information patterns to identify the type of currently observed trajectories. This allows machine learning models to automatically determine the most probable identification result when faced with massive amounts of historical data, providing the identification criteria and detailed probability distribution. The algorithm model, based on rigorous statistical formulas, effectively eliminates human subjective bias, providing fast and high-value auxiliary information for manual verification.

[0003] Existing trajectory recognition methods are mainly based on deep learning and statistical methods. Deep learning-based methods, including LSTM and CNN, suffer from problems such as data sparsity and inability to effectively learn feature details due to the extremely wide and uneven distribution of trajectories. Furthermore, deep learning generally has poor interpretability and cannot provide detailed prediction basis. Statistical methods use information such as the bounding polygon region of the trajectory set or heatmaps for description, but they suffer from poor performance when different types of trajectory regions overlap and cannot utilize the shape information of the trajectory itself.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a trajectory recognition method, apparatus, device, and storage medium, aiming to solve the technical problem of low efficiency in current trajectory recognition.

[0006] To achieve the above objectives, the present invention provides a trajectory recognition method, the method comprising the following steps:

[0007] Based on historical trajectory data, a shape model and a location model are constructed; wherein, the shape model is a classification model of trajectory shapes in historical trajectory data, and the location model is a classification model of trajectory points in historical trajectory data;

[0008] The shape model and position model are called to process the trajectory to be identified, respectively, to obtain the first probability distribution of the trajectory to be identified and the second probability distribution of the trajectory points in the trajectory to be identified;

[0009] Calculate the overall probability distribution of the trajectory to be identified based on the first probability distribution and the second probability distribution;

[0010] Based on the overall probability distribution, the trajectory type of the trajectory to be identified is determined.

[0011] Optionally, the historical trajectory data includes several trajectories and corresponding category labels for the trajectories. Each trajectory consists of several trajectory points, and each trajectory point includes the longitude, latitude, and timestamp of the trajectory point.

[0012] Optionally, constructing the shape model specifically includes:

[0013] The LSTM autoencoder is trained using trajectory points from historical trajectory data, and feature data of the trajectory shape output from the encoding layer of the LSTM autoencoder are extracted.

[0014] A binary classification model is constructed for each category label, and the feature data is input into the binary classification model to train the binary classification model and obtain the shape model corresponding to each category label.

[0015] Optionally, constructing the location model specifically includes:

[0016] Based on the observation timestamp of the trajectory corresponding to each category label, extract the trajectory points corresponding to the observation timestamp to construct a sample set;

[0017] Using the sample set, a Gaussian mixture model is trained to obtain the location model corresponding to each category label.

[0018] Optionally, before the steps of calling the shape model and position model to process the trajectory to be identified, the method further includes determining the prior probability distribution P of all category labels based on the number of trajectory points corresponding to each category label. Pre .

[0019] Optionally, the expression for obtaining the overall probability distribution of the trajectory to be identified based on the first probability distribution and the second probability distribution is specifically as follows:

[0020] P Final =w1*P Shape +w2*P Loc stw1+w2=1

[0021] Among them, P Loc =P Loc-co *P Pre P Shape =P Shape-co *P Pre P ShaPe-co Let P be the first probability distribution corresponding to the shape model. Shape-coLet w1 and w2 be the second probability distribution corresponding to the location model, and w1 and w2 be the weight values.

[0022] Optionally, the method further includes preprocessing historical trajectory data and / or the trajectory to be identified; wherein the preprocessing includes uniformly interpolating the trajectory.

[0023] Furthermore, to achieve the above objectives, the present invention also provides a trajectory recognition device, the trajectory recognition device comprising:

[0024] A construction module is used to build a shape model and a location model based on historical trajectory data; wherein, the shape model is a classification model of trajectory shapes in historical trajectory data, and the location model is a classification model of trajectory points in historical trajectory data;

[0025] The calling module is used to call the shape model and the position model to process the trajectory to be identified, and obtain the first probability distribution of the trajectory to be identified and the second probability distribution of the trajectory points in the trajectory to be identified.

[0026] The calculation module is used to calculate the overall probability distribution of the trajectory to be identified based on the first probability distribution and the second probability distribution.

[0027] The determination module is used to determine the trajectory type of the trajectory to be identified based on the overall probability distribution.

[0028] In addition, to achieve the above objectives, the present invention also provides a trajectory recognition device, the device comprising: a memory, a processor, and a trajectory recognition program stored in the memory and executable on the processor, wherein the trajectory recognition program, when executed by the processor, implements the steps of the trajectory recognition method described above.

[0029] In addition, to achieve the above objectives, the present invention also provides a storage medium storing a trajectory recognition program, which, when executed by a processor, implements the steps of the trajectory recognition method described above.

[0030] This invention proposes a trajectory recognition method, apparatus, device, and storage medium. The method includes constructing a shape model and a position model based on historical trajectory data. The shape model is a classification model of trajectory shapes in the historical trajectory data, and the position model is a classification model of trajectory points in the historical trajectory data. The shape model and position model are used to process the trajectory to be recognized, respectively, to obtain a first probability distribution of the trajectory to be recognized and a second probability distribution of the trajectory points in the trajectory to be recognized. Based on the first and second probability distributions, the overall probability distribution of the trajectory to be recognized is calculated. Based on the overall probability distribution, the trajectory type of the trajectory to be recognized is determined. This invention solves the problems of sparse high-dimensional data samples, weak anti-forgery ability, and poor interpretability when using a single model by modeling trajectory shape and position features separately, thereby providing users with accurate and rapid recognition results and evidence. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the trajectory recognition device of the present invention;

[0032] Figure 2 This is a flowchart illustrating an embodiment of the trajectory recognition method of the present invention;

[0033] Figure 3 This is a flowchart illustrating a specific example of trajectory recognition using the trajectory recognition method of the present invention;

[0034] Figure 4 This is a structural diagram of the shape model of the present invention;

[0035] Figure 5 A comparison chart showing the location modeling of this invention with and without considering the time factor;

[0036] Figure 6 This is a structural block diagram of the trajectory recognition device of the present invention.

[0037] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0038] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0039] Existing trajectory recognition methods are mainly based on deep learning and statistical methods. Deep learning-based methods, including LSTM and CNN, suffer from problems such as data sparsity and inability to effectively learn feature details due to the extremely wide and uneven distribution of trajectories. Furthermore, deep learning generally has poor interpretability and cannot provide detailed prediction basis. Statistical methods use information such as the bounding polygon region of the trajectory set or heatmaps for description, but they suffer from poor performance when different types of trajectory regions overlap and cannot utilize the shape information of the trajectory itself.

[0040] To address this problem, various embodiments of the trajectory recognition method of the present invention are proposed. The trajectory recognition method provided by the present invention identifies the trajectory by modeling the trajectory shape and position features separately. This solves the problems of sparse high-dimensional data samples, weak resistance to manipulation, and poor interpretability when using a single model, and can accurately and quickly provide users with recognition results and evidence.

[0041] Reference Figure 1 , Figure 1 This is a schematic diagram of the trajectory recognition device involved in the embodiment of the present invention.

[0042] The device can be a user equipment (UE) such as a mobile phone, smartphone, laptop, digital broadcast receiver, personal digital assistant (PDA), tablet computer (PAD), handheld device, in-vehicle device, wearable device, computing device or other processing device connected to a wireless modem, mobile station (MS), etc., used for trajectory recognition. The device may be referred to as a user terminal, portable terminal, desktop terminal, etc.

[0043] Typically, the device includes: at least one processor 301, a memory 302, and a trajectory recognition program stored in the memory and executable on the processor, the trajectory recognition program being configured to implement the steps of the trajectory recognition method as described above.

[0044] Processor 301 may include one or more processing cores, such as a quad-core processor or an octa-core processor. Processor 301 may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). Processor 301 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 301 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. Processor 301 may also include an AI (Artificial Intelligence) processor, which handles trajectory recognition operations, enabling the trajectory recognition model to train and learn autonomously, improving efficiency and accuracy.

[0045] The memory 302 may include one or more computer-readable storage media, which may be non-transitory. The memory 302 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 302 are used to store at least one instruction, which is executed by the processor 301 to implement the trajectory recognition method provided in the method embodiments of this application.

[0046] In some embodiments, the terminal may also optionally include a communication interface 303 and at least one peripheral device. The processor 301, memory 302, and communication interface 303 can be connected via a bus or signal line. Each peripheral device can be connected to the communication interface 303 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 304, a display screen 305, and a power supply 306.

[0047] The communication interface 303 can be used to connect at least one I / O (Input / Output) related peripheral device to the processor 301 and the memory 302. The communication interface 303 is used to receive movement trajectories and other data uploaded by the user from multiple mobile terminals via the peripheral device. In some embodiments, the processor 301, memory 302, and communication interface 303 are integrated on the same chip or circuit board; in other embodiments, any one or two of the processor 301, memory 302, and communication interface 303 can be implemented on separate chips or circuit boards, and this embodiment is not limited to this.

[0048] The radio frequency (RF) circuit 304 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The RF circuit 304 communicates with communication networks and other communication devices via electromagnetic signals, thereby acquiring the movement trajectories and other data of multiple mobile terminals. The RF circuit 304 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals back into electrical signals. Optionally, the RF circuit 304 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, etc. The RF circuit 304 can communicate with other terminals through at least one wireless communication protocol. This wireless communication protocol includes, but is not limited to: metropolitan area networks (MANs), various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks (WLANs), and / or WiFi (Wireless Fidelity) networks. In some embodiments, the RF circuit 304 may also include circuitry related to NFC (Near Field Communication), which is not limited in this application.

[0049] Display screen 305 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 305 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 301 for processing. In this case, display screen 305 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, display screen 305 can be a single screen, the front panel of an electronic device; in other embodiments, display screen 305 can be at least two screens, respectively disposed on different surfaces of the electronic device or in a folded design; in still other embodiments, display screen 305 can be a flexible display screen, disposed on a curved or folded surface of the electronic device. Furthermore, display screen 305 can also be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. Display screen 305 can be made of materials such as LCD (Liquid Crystal Display) or OLED (Organic Light-Emitting Diode).

[0050] Power supply 306 is used to supply power to various components in an electronic device. Power supply 306 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When power supply 306 includes a rechargeable battery, the rechargeable battery can support wired or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0051] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the trajectory recognition device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0052] This invention provides a trajectory recognition method, referring to... Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the trajectory recognition method of the present invention.

[0053] In this embodiment, the trajectory recognition method includes the following steps:

[0054] Step S100: Based on historical trajectory data, construct a shape model and a location model; wherein, the shape model is a classification model of trajectory shapes in historical trajectory data, and the location model is a classification model of trajectory points in historical trajectory data.

[0055] Specifically, in practical applications, historical trajectory data includes several trajectories and corresponding category labels. The trajectories are composed of several trajectory points, and each trajectory point includes its longitude, latitude, and timestamp.

[0056] In constructing the shape model, the LSTM autoencoder is trained using trajectory points from historical trajectory data, and feature data of the trajectory shape output by the encoding layer of the LSTM autoencoder is extracted. A binary classification model is constructed for each category label, and the feature data is input into the binary classification model to train the binary classification model and obtain the shape model corresponding to each category label.

[0057] In this embodiment, the problem of no negative examples in the trajectory is solved by using an autoencoder and random feature data when modeling the shape. If the trajectory is directly input into an LSTM neural network binary classifier, it is difficult to generate comprehensive negative example trajectories because the trajectory length is variable, the shape space is huge, and the shapes of other target trajectories may be similar to the target. However, by converting the trajectory into fixed-length feature data through an encoder, and generating positive and negative samples from the feature data, this problem is solved.

[0058] When constructing the location model, the trajectory points corresponding to the observation timestamps of the trajectories corresponding to each category label are extracted to construct a sample set; using the sample set, a Gaussian mixture model is trained to obtain the location model corresponding to each category label.

[0059] In this embodiment, time-related information is considered when modeling the location. For example, the target has appeared the same number of times in regions A and B, but all occurrences in region A are from 5 years ago, while those in region B are more recent. Obviously, it is more likely to appear in region B next time. That is, the probability density of the Gaussian mixture model in region B should be higher, rather than equal to that in region A. By introducing different weights to the samples using timestamp information, the improved model is more consistent with empirical facts.

[0060] It should be noted that there are n categories in the historical data, corresponding to n labels. The trajectory data corresponding to each label consists of one or more trajectories. Since the density of trajectory points in each trajectory is different, and the density of trajectory points in different areas of the same trajectory may also be different, it is necessary to perform uniform distance interpolation on all trajectory data.

[0061] To put it simply, modeling the shape of a trajectory using historical data involves the following steps:

[0062] A. Perform data augmentation on all trajectory data. Specifically, add a small-range Gaussian noise perturbation to the location of each trajectory point, and randomly select a subset of the trajectory sequences to form a new dataset NS. Offset the starting point of all trajectories in NS to (0,0) while keeping the numerical differences between the coordinates of other points in the trajectory and the starting point unchanged.

[0063] B. Establish an LSTM autoencoder with latitude and longitude sequences of trajectory points as both input and output. After training, the coding layer of the LSTM autoencoder can be used to obtain feature data of the trajectory shape.

[0064] C. Build a binary classification model for each label. The input data is the feature data of the trajectory, and the output is the probability of being labeled. When training the binary classification model, the positive examples in the dataset are the feature data corresponding to the trajectory under that label in the dataset NS, and the negative examples are random feature data.

[0065] D. The shape model used during inference is the cascaded binary classification model of the LSTM autoencoder encoding part. Since there are n categories, there are n binary classification models, that is, there are a total of n shape models.

[0066] Modeling the location of a trajectory using historical data includes the following steps:

[0067] A. For a trajectory dataset under a certain label, obtain the observation timestamp of each trajectory and apply it to the corresponding trajectory points. Take these trajectory points as the sample set. The weight of samples with smaller timestamps will be reduced.

[0068] B. Model the location of the sample set. Here, a two-dimensional Gaussian mixture model based on the Dirichlet process is selected, and the weight data of the samples is applied during training.

[0069] C. After processing all labels using the above steps, n location models are obtained.

[0070] Step S200: The shape model and position model are called to process the trajectory to be identified, respectively, to obtain the first probability distribution of the trajectory to be identified and the second probability distribution of the trajectory points in the trajectory to be identified.

[0071] It should be noted that after constructing and obtaining the shape model and position model, it is also necessary to determine the prior probability distribution P of all category labels based on the number of trajectory points corresponding to each category label. Pre .

[0072] Specifically, the prior probability of a category is calculated through statistical analysis. The method is as follows: let the total number of trajectory points be tn, and the number of trajectory points corresponding to a certain label be tx. Then the prior probability of this label is tx / tn. Let the probability distribution formed by the prior probabilities of all labels be P. Pre .

[0073] At the same time, when using the trajectory to be identified, all trajectory data can also be interpolated at uniform distances.

[0074] After obtaining the shape model and position model, the trajectory to be identified can be processed using the shape model and position model to obtain the first probability distribution of the trajectory to be identified and the second probability distribution of the trajectory points in the trajectory to be identified.

[0075] Step S300: Calculate the overall probability distribution of the trajectory to be identified based on the first probability distribution and the second probability distribution.

[0076] In this 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. Based on the first and second probability distributions, the expression for the overall probability distribution of the trajectory to be identified is as follows:

[0077] P Final =w1*P Shape +w2*P Loc stw1+w2=1

[0078] Among them, P Loc =P Loc-co *P Pre P Shape =P Shape-co *P Pre P Shape-co Let P be the first probability distribution corresponding to the shape model. Shape-co Let w1 and w2 be the second probability distribution corresponding to the location model, and w1 and w2 be the weight values.

[0079] Specifically, the Bayesian method is used to calculate the probability distribution of each category to which the trajectory belongs, including the following steps:

[0080] A. Perform uniform distance interpolation preprocessing on the trajectory to be identified;

[0081] B. The conditional probability distribution of the trajectory to be identified is calculated based on n shape models, denoted as P. Shape-co .

[0082] C. Substitute all points of the trajectory to be identified into the respective location models. Each model yields multiple probability density values. Calculate the logarithmic sum of each value, and convert these n logarithmic sums into a probability distribution, denoted as P. Loc-co .

[0083] D. According to Bayes' theorem, the posterior probability of a trajectory belonging to each model is proportional to the prior probability multiplied by the conditional probability. Let P... Loc =P Loc-co *P Pre P Shape =P Shape-co *P Pre and P Loc P ShapeConvert them into probability distributions that sum to 1, which are the posterior probabilities of the trajectory belonging to the position and shape models.

[0084] E. Assign weights w1 and w2 to the shape model and the position model, and the final probability distribution is:

[0085] P Final =w1*P Shape +w2*P Loc stw1+w2=1.

[0086] Step S400: Based on the overall probability distribution, determine the trajectory type of the trajectory to be identified.

[0087] Specifically, after obtaining the overall probability distribution of the trajectory to be identified, the trajectory to be identified can be confirmed according to the trajectory type corresponding to the maximum probability value in the overall probability distribution.

[0088] In this embodiment, a trajectory recognition method based on LSTM and Gaussian mixture model is used to describe the trajectory from both position and shape perspectives. This avoids the problem that a single model cannot effectively learn due to the large spatial range of the trajectory and the sparse data. Compared with other existing deep learning models, the solution provided in this embodiment has strong interpretability, and the position model can be intuitively observed through a two-dimensional probability density map to understand the inference process.

[0089] To illustrate the invention more clearly, specific examples of trajectory recognition will be used below.

[0090] like Figure 3 As shown, a trajectory recognition method based on LSTM and Gaussian mixture model is provided, including the following steps:

[0091] A. Data preprocessing: Insert points at uniform distances into all trajectory data. Specifically, specify the interval distance d. If the interval between two adjacent points in the trajectory is greater than or equal to (m+0.5)*d and less than (m+1.5)*d, then insert m+1 points evenly between the two points. This ensures that the intervals between all adjacent trajectory points are as consistent as possible while maintaining the shape of the trajectory.

[0092] B. Use historical data to model the shape of the trajectory, such as Figure 4As shown, the LSTM autoencoder is first trained, with the same trajectory data for both input and output. Since the trajectory positions are uniformly moved to the origin, it learns the trajectory shape, converting variable-length trajectories into fixed-length shape representation vectors. The overall shape model connects the LSTM encoder to a neural network binary classifier, with the Sigmoid function outputting the corresponding label probability. A binary classifier is trained for each label, resulting in n shape models whose LSTM autoencoder parts are shared. Because generating negative sample trajectories is very difficult, fixed-length random features are directly generated as negative samples for training, thus solving this problem.

[0093] C. Location modeling is performed. Here, a Gaussian mixture model based on the Dirichlet process is chosen. Compared to a general Gaussian mixture model, this algorithm automatically determines parameters based on dataset characteristics, avoiding overfitting or underfitting problems caused by manually specifying parameters for different datasets. Furthermore, weighting factors are added during model training; that is, when calculating the mean and variance, the sample value for each sample is changed to sample value * weight, and the number of samples is changed to the sum of the weights. For example... Figure 5 As shown, when fitting a Gaussian distribution to trajectory point data, the older the data timestamp, the lower its corresponding weight, thus affecting the fitted probability density and making it closer to the reality. When applying a location model for calculation, each point in the trajectory is substituted into the model to obtain a series of probability density values. The product of these values ​​is the probability that the model generated this trajectory. Since the value is very small, the multiplication is changed to taking the logarithmic sum.

[0094] D. Use the Bayesian method to calculate the probability distribution of each category to which the trajectory belongs. That is, based on the model results, the probability of the label itself appearing is taken into account. According to Bayes' theorem, the posterior probability of the trajectory belonging to each model is proportional to the prior probability multiplied by the conditional probability. Let P... Loc =P Loc-co *P Pre P Shape =P Shape-co *P Pre and P Loc P Shape The data are then converted to probability distribution form by scaling them proportionally until the sum of the data is 1. This yields the posterior probabilities of the trajectory belonging to the position and shape models. Weights are then assigned to the shape and position models, resulting in the final probability distribution:

[0095] P Final =w1*P Shape +w2*P Loc stw1+w2=1

[0096] w1 and w2 can be specified by experts or learned by the model through optimization algorithms.

[0097] In this embodiment, historical data is used to model the shape and position of the trajectory, and prior probabilities of the categories are calculated through statistical analysis. Finally, the probability distribution of each category to which the trajectory belongs is calculated using a Bayesian method, and the inference process is visualized and presented to the user. This invention implements a trajectory recognition method based on an LSTM and Gaussian mixture model. By modeling the shape and position features of the trajectory separately, it solves the problems of sparse high-dimensional data samples, weak anti-forgery ability, and poor interpretability when using a single model, and can accurately and quickly provide users with recognition results and evidence.

[0098] Reference Figure 6 , Figure 6 This is a structural block diagram of an embodiment of the trajectory recognition device of the present invention.

[0099] like Figure 6 As shown, the trajectory recognition device proposed in this embodiment of the invention includes:

[0100] The construction module 10 is used to construct a shape model and a position model based on historical trajectory data; wherein, the shape model is a classification model of trajectory shapes in historical trajectory data, and the position model is a classification model of trajectory points in historical trajectory data;

[0101] Module 20 is used to call the shape model and the position model to process the trajectory to be identified, respectively, to obtain the first probability distribution of the trajectory to be identified and the second probability distribution of the trajectory points in the trajectory to be identified.

[0102] The calculation module 30 is used to calculate the overall probability distribution of the trajectory to be identified based on the first probability distribution and the second probability distribution;

[0103] The determination module 40 is used to determine the trajectory type of the trajectory to be identified based on the overall probability distribution.

[0104] In this embodiment, a trajectory recognition device is provided. By modeling the trajectory shape and position features separately, the trajectory is recognized. This solves the problems of sparse high-dimensional data samples, weak resistance to manipulation, and poor interpretability when using a single model. It can accurately and quickly provide users with recognition results and evidence.

[0105] Other embodiments or specific implementations of the trajectory recognition device of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0106] Furthermore, embodiments of the present invention also propose a storage medium storing a trajectory recognition program, which, when executed by a processor, implements the steps of the trajectory recognition method described above. Therefore, further details will not be repeated here. Additionally, the beneficial effects of employing the same method will not be repeated. For technical details not disclosed in the computer-readable storage medium embodiments related to this application, please refer to the description of the method embodiments of this application. As an example, program instructions may be deployed to execute on a single computing device, or on multiple computing devices located at one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0107] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0108] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware, and of course, it can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memory, special components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for the present invention, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, portable hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

Claims

1. A trajectory recognition method, characterized in that, include: Based on historical trajectory data, a shape model and a location model are constructed; wherein, the shape model is a classification model of trajectory shapes in historical trajectory data, and the location model is a classification model of trajectory points in historical trajectory data; The historical trajectory data includes several trajectories and corresponding category labels. Each trajectory consists of several trajectory points, and each trajectory point includes the longitude, latitude, and timestamp of the trajectory point. Constructing the location model specifically includes: Obtain the observation timestamp of the trajectory corresponding to each category label, extract the trajectory points corresponding to the observation timestamp, and construct a sample set. The weight of samples with smaller timestamps will be reduced. Position modeling is performed on the sample set, and a two-dimensional Gaussian mixture model based on the Dirichlet process is trained to obtain the position model corresponding to each category label; The method calls a shape model and a position model to process the trajectory to be identified, respectively, to obtain a first probability distribution of the trajectory to be identified and a second probability distribution of the trajectory points in the trajectory to be identified. Before the step of calling the shape model and the position model to process the trajectory to be identified, the method further includes determining the prior probability distribution of all category labels based on the number of trajectory points corresponding to each category label. ; Based on the first and second probability distributions, the overall probability distribution of the trajectory to be identified is calculated, and the specific expression is as follows: in, This represents the first probability distribution corresponding to the shape model. This represents the second probability distribution corresponding to the location model. , These are weight values; Based on the overall probability distribution, the trajectory type of the trajectory to be identified is determined.

2. The trajectory recognition method as described in claim 1, characterized in that, Constructing the shape model specifically includes: The LSTM autoencoder is trained using trajectory points from historical trajectory data, and feature data of the trajectory shape output from the encoding layer of the LSTM autoencoder are extracted. A binary classification model is constructed for each category label, and the feature data is input into the binary classification model to train the binary classification model and obtain the shape model corresponding to each category label.

3. The trajectory recognition method according to any one of claims 1-2, characterized in that, The method further includes preprocessing historical trajectory data and / or the trajectory to be identified; wherein the preprocessing includes uniformly interpolating the trajectory.

4. A trajectory recognition device, characterized in that, The trajectory recognition device includes: A construction module is used to build shape models and location models based on historical trajectory data. The historical trajectory data includes several trajectories and corresponding category labels. Each trajectory consists of several trajectory points, and each trajectory point includes the longitude, latitude, and timestamp of the trajectory point. The shape model is a classification model of trajectory shapes in the historical trajectory data, and the location model is a classification model of trajectory points in the historical trajectory data. Constructing the location model specifically includes: Based on the observation timestamp of the trajectory corresponding to each category label, extract the trajectory points corresponding to the observation timestamp to construct a sample set; Using the sample set, a Gaussian mixture model is trained to obtain the location model corresponding to each category label; The calling module is used to call the shape model and the position model to process the trajectory to be identified, respectively, to obtain the first probability distribution of the trajectory to be identified and the second probability distribution of the trajectory points in the trajectory to be identified; before the step of calling the shape model and the position model to process the trajectory to be identified, the module further includes determining the prior probability distribution of all category labels based on the number of trajectory points corresponding to each category label. ; The calculation module is used to calculate the overall probability distribution of the trajectory to be identified based on a first probability distribution and a second probability distribution; the expression for obtaining the overall probability distribution of the trajectory to be identified based on the first probability distribution and the second probability distribution is specifically as follows: in, This represents the first probability distribution corresponding to the shape model. This represents the second probability distribution corresponding to the location model. , These are weight values; The determination module is used to determine the trajectory type of the trajectory to be identified based on the overall probability distribution.

5. A trajectory recognition device, characterized in that, The trajectory recognition device includes: a memory, a processor, and a trajectory recognition program stored in the memory and executable on the processor, wherein the trajectory recognition program, when executed by the processor, implements the steps of the trajectory recognition method as described in any one of claims 1 to 3.

6. A storage medium, characterized in that, The storage medium stores a trajectory recognition program, which, when executed by a processor, implements the steps of the trajectory recognition method as described in any one of claims 1 to 3.

Images