Deep learning-based arbitrary irregular trajectory matching method and device

Abstract

The disclosure provides an arbitrary irregular trajectory matching method and device based on deep learning. The method comprises: obtaining a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library comprises a plurality of candidate trajectories; performing fixed sequence length sampling on the reference trajectory to obtain a first trajectory sequence, and performing fixed sequence length sampling on the candidate trajectory to obtain a corresponding second trajectory sequence; inputting the first trajectory sequence and the second trajectory sequence into a pre-trained trajectory matching model to obtain an alignment matrix, wherein the alignment matrix is used to indicate the point-by-point correspondence relationship between the first trajectory sequence and the second trajectory sequence; and obtaining a trajectory matching result according to the alignment matrix. Thus, the fast and accurate matching of arbitrary irregular trajectories can be realized, so that the trajectory matching process can be adapted to individual application scenarios, and the cost loss in the trajectory matching process can be effectively reduced while improving the trajectory matching efficiency.

Description

Technical Field

[0001] This disclosure relates to the field of trajectory matching technology, and specifically to a method and apparatus for matching arbitrary irregular trajectories based on deep learning. Background Technology

[0002] In robotic operations such as welding, spraying, grinding, gluing, laser cladding, 3D printing, medical puncture, agricultural machinery pruning, and drone inspection, operators often can only provide a rough 3D trajectory for "hands-on teaching," such as using handheld laser pointers, AR gestures, voice + eye tracking, one-click CAD assignment, or real-time SLAM capture. This trajectory suffers from four major problems: incomplete length, large spatial errors, local missing parts, and uncertain direction, making it unsuitable for direct robot execution.

[0003] In related technologies, the matching effect for arbitrary irregular trajectories is not good. Summary of the Invention

[0004] This disclosure aims to at least partially address one of the technical problems in the related art.

[0005] Therefore, the purpose of this disclosure is to propose a method, apparatus, computer device, and storage medium for matching arbitrary irregular trajectories based on deep learning. This enables fast and accurate matching of arbitrary irregular trajectories, ensuring that the trajectory matching process can be adapted to personalized application scenarios, and effectively reducing cost losses in the trajectory matching process while improving trajectory matching efficiency.

[0006] To achieve the above objectives, the first aspect of this disclosure proposes a deep learning-based method for matching arbitrary irregular trajectories, comprising: Obtain a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories; The reference trajectory is sampled with a fixed sequence length to obtain a first trajectory sequence, and the candidate trajectory is sampled with a fixed sequence length to obtain a corresponding second trajectory sequence; The first trajectory sequence and the second trajectory sequence are input into a pre-trained trajectory matching model to obtain an alignment matrix, wherein the alignment matrix is ​​used to indicate the point-by-point correspondence between the first trajectory sequence and the second trajectory sequence; Based on the alignment matrix, the trajectory matching result is obtained.

[0007] To achieve the above objectives, the second aspect of this disclosure provides a deep learning-based arbitrary irregular trajectory matching device, comprising: The first acquisition module is used to acquire a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories; The sampling module is used to sample the reference trajectory with a fixed sequence length to obtain a first trajectory sequence, and to sample the candidate trajectory with a fixed sequence length to obtain a corresponding second trajectory sequence; An input module is used to input the first trajectory sequence and the second trajectory sequence into a pre-trained trajectory matching model to obtain an alignment matrix, wherein the alignment matrix is ​​used to indicate the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence; The second acquisition module is used to obtain the trajectory matching result based on the alignment matrix.

[0008] The computer device proposed in the third aspect of this disclosure includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the deep learning-based arbitrary irregular trajectory matching method proposed in the first aspect of this disclosure.

[0009] The fourth aspect of this disclosure proposes a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the deep learning-based arbitrary irregular trajectory matching method proposed in the first aspect of this disclosure.

[0010] A fifth aspect of this disclosure provides a computer program product in which, when instructions are executed by a processor, an arbitrary irregular trajectory matching method based on deep learning as proposed in a first aspect of this disclosure is performed.

[0011] This disclosure provides a deep learning-based method, apparatus, computer device, and storage medium for matching arbitrary irregular trajectories. The method involves acquiring a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories. A first trajectory sequence is obtained by sampling the reference trajectory with a fixed sequence length, and a corresponding second trajectory sequence is obtained by sampling the candidate trajectories with a fixed sequence length. The first and second trajectory sequences are input into a pre-trained trajectory matching model to obtain an alignment matrix, which indicates the point-to-point correspondence between the first and second trajectory sequences. The trajectory matching result is obtained based on the alignment matrix. This enables fast and accurate matching of arbitrary irregular trajectories, ensuring that the trajectory matching process can be adapted to personalized application scenarios, and effectively reducing cost losses during the trajectory matching process while improving efficiency.

[0012] Additional aspects and advantages of this disclosure will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this disclosure. Attached Figure Description

[0013] The above and / or additional aspects and advantages of this disclosure will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, in which: Figure 1 This is a flowchart illustrating an embodiment of the deep learning-based arbitrary irregular trajectory matching method proposed in this disclosure. Figure 2 This is a flowchart illustrating an arbitrary irregular trajectory matching method based on deep learning proposed in another embodiment of this disclosure; Figure 3 This is a flowchart of an arbitrary irregular trajectory matching algorithm based on deep learning proposed in this disclosure; Figure 4 This is a schematic diagram of the matching model structure proposed in this disclosure; Figure 5 This is a schematic diagram of the trajectory matching process proposed in this disclosure; Figure 6 This is a schematic diagram of the structure of an arbitrary irregular trajectory matching device based on deep learning proposed in an embodiment of this disclosure; Figure 7 A block diagram of an exemplary computer device suitable for implementing embodiments of the present disclosure is shown. Detailed Implementation

[0014] Embodiments of this disclosure are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are used only to explain this disclosure, and should not be construed as limiting this disclosure. Rather, embodiments of this disclosure include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0015] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0016] Figure 1 This is a flowchart illustrating an embodiment of the arbitrary irregular trajectory matching method based on deep learning proposed in this disclosure.

[0017] It should be noted that the execution subject of the deep learning-based arbitrary irregular trajectory matching method in this embodiment is a deep learning-based arbitrary irregular trajectory matching device. This device can be implemented by software and / or hardware. The device can be configured in a computer device, which may include, but is not limited to, a terminal, a server, etc., such as a mobile phone, a PDA, etc.

[0018] like Figure 1 As shown, this deep learning-based method for matching arbitrary irregular trajectories includes: S101: Obtain the reference trajectory and candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories.

[0019] The reference trajectory can refer to a roughly indicated trajectory Q in a real-world application scenario, such as a handheld laser pointer, AR gestures, eye-tracking trajectory, CAD output, SLAM teaching trajectory sampling, or a sequence of trajectory points recorded by an operator dragging the robot's end effector.

[0020] The candidate trajectory library C can be generated offline in advance or obtained from the semantic segmentation results of the preceding module, such as: Welding: Weld centerline; Spraying: Equidistant spray gun trajectory; Grinding: Sanding belt contact line; Glue application: glue width center line; Medical: Safe access for puncture; Drones: Inspecting flight path corridors; Agricultural machinery: The trajectory of a weed cutter.

[0021] Among them, candidate trajectories, Optionally, in some embodiments, it is only necessary to ensure that the trajectory data is ordered when collecting the trajectory data, such as indicating that each point in the trajectory is sorted by time.

[0022] In this embodiment of the disclosure, when a reference trajectory and a candidate trajectory library are obtained, multiple candidate trajectories in the candidate trajectory library can provide reliable data support for the matching process of the reference trajectory.

[0023] S102: Sample the reference trajectory with a fixed sequence length to obtain the first trajectory sequence, and sample the candidate trajectory with a fixed sequence length to obtain the corresponding second trajectory sequence.

[0024] Fixed-length sampling can refer to the process of systematically and in batches extracting shorter subsequences of identical length (number of points) from a very long time series.

[0025] The first trajectory sequence and the second trajectory sequence can refer to the trajectory sequences obtained by sampling the reference trajectory and the candidate trajectory with a fixed sequence length, respectively.

[0026] In this embodiment of the present disclosure, when sampling the reference trajectory with a fixed sequence length to obtain the first trajectory sequence and sampling the candidate trajectory with a fixed sequence length to obtain the corresponding second trajectory sequence, the reference trajectory and the candidate trajectory can be input into a pre-trained machine learning model to obtain the corresponding first trajectory sequence and the second trajectory sequence. Alternatively, the first trajectory sequence can be obtained by sampling the reference trajectory with a fixed sequence length based on a combination of numerical and graphical methods, and the second trajectory sequence can be obtained by sampling the candidate trajectory with a fixed sequence length. There are no limitations on this.

[0027] Optionally, in some embodiments, when sampling a reference trajectory with a fixed sequence length to obtain a first trajectory sequence, and sampling a candidate trajectory with a fixed sequence length to obtain a corresponding second trajectory sequence, the process may involve: determining multiple first trajectory points in the reference trajectory and multiple second trajectory points in the candidate trajectory based on a preset fixed sequence length; determining a first parameter corresponding to a first trajectory point and a second parameter corresponding to a second trajectory point, wherein the first parameter includes the three-dimensional coordinates, local discrete curvature, torsion, and angle value of the first trajectory point, and the second parameter includes the three-dimensional coordinates, local discrete curvature, torsion, and angle value of the second trajectory point; constructing a first trajectory sequence based on the first parameter of the first trajectory point; and constructing a second trajectory sequence based on the second parameter of the second trajectory point. Thus, equal-length sampling of the reference trajectory and candidate trajectory can be achieved based on a preset fixed sequence length, thereby ensuring the adaptability of the first trajectory sequence and the second trajectory sequence during the matching process, and ensuring the information indication effect of the obtained first trajectory sequence and second trajectory sequence based on the first and second parameters.

[0028] The first trajectory point can refer to a trajectory point in a fixed-length sequence obtained by sampling a reference trajectory with a fixed sequence length based on a preset fixed sequence length. The second trajectory point can refer to a trajectory point in a fixed-length sequence obtained by sampling a candidate trajectory with a fixed sequence length based on a preset fixed sequence length.

[0029] The first parameter and the second parameter can refer to the relevant parameters of the corresponding trajectory points.

[0030] In other words, in this embodiment of the disclosure, the collected query trajectory and candidate trajectory point set can be sampled with a fixed sequence length. The sampling scheme can handle cases where the number of points exceeds or is insufficient for the preset sequence length by combining random downsampling and linear interpolation upsampling.

[0031]

[0032] Q and C These represent the sampled indicator trajectory and a specific trajectory from the candidate trajectory set, respectively. Represents the dimension of each input point; These are the three-dimensional coordinates of each input point; Represents local discrete curvature, describing the degree of curvature of the trajectory at that point, and is approximated by the difference of three consecutive points; Represents torsion, describing the degree of twisting of the trajectory at that point, and is approximated by the four-point center difference; The angle at which the trajectory bends at that point is calculated using the included angle between three adjacent points.

[0033]

[0034]

[0035] L is a fixed sequence length and a model hyperparameter, which can be determined based on the specific input trajectory features, such as L=1024. The sampled trajectory point set is normalized according to the maximum 3D coordinate of all trajectory points and used as the input for the subsequent trajectory matching model.

[0036] S103: Input the first trajectory sequence and the second trajectory sequence into the pre-trained trajectory matching model to obtain the alignment matrix, wherein the alignment matrix is ​​used to indicate the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence.

[0037] The trajectory matching model can refer to a pre-trained model used to match the first trajectory sequence and the second trajectory sequence.

[0038] The alignment matrix can be used to indicate the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence.

[0039] Optionally, in some embodiments, the trajectory matching model includes: a trajectory segment encoder, a segment matching module, and a point-by-point alignment confidence head; wherein, the alignment matrix is ​​obtained based on the following method: the trajectory segment encoder divides the first trajectory sequence into multiple first trajectory segments, and determines the first length weight and first segment-level feature of each first trajectory segment; the trajectory segment encoder divides the second trajectory sequence into multiple second trajectory segments, and determines the second length weight and second segment-level feature of each second trajectory segment; the segment matching module processes the first segment-level feature and the second segment-level feature to obtain a similarity matrix, wherein the similarity matrix is ​​used to indicate the similarity distribution between each first segment-level feature and different second segment-level features; the point-by-point alignment confidence head processes the first length weight, the first segment-level feature, the second length weight, the second segment-level feature, and the similarity matrix to obtain the alignment matrix. Therefore, the indication effect of the obtained alignment matrix on the point-by-point correspondence between the first trajectory sequence and the second trajectory sequence can be effectively improved.

[0040] Here, a trajectory segment refers to a trajectory segment obtained after segmenting the trajectory sequence using a trajectory segment encoder. The total number of trajectory segments obtained from the trajectory sequence can be flexibly adjusted according to the application scenario and is not limited thereto. The first trajectory segment and the second trajectory segment refer to the trajectory segments obtained after processing the first trajectory sequence and the second trajectory sequence using a trajectory segment encoder.

[0041] The length weight can be used to indicate how many trajectory segments the trajectory sequence should be divided into, and how many trajectory points each trajectory segment should contain. The first length weight and the second length weight refer to the length weights corresponding to the first trajectory segment and the second trajectory segment, respectively.

[0042] Segment-level features can be used to describe the feature information of trajectory sub-segments. For example, this segment-level feature can be a comprehensive feature vector. The first segment-level feature and the second segment-level feature refer to the segment-level features corresponding to the first trajectory sub-segment and the second trajectory sub-segment, respectively. Optionally, in some embodiments, when processing the first length weight, the first segment-level feature, the second length weight, the second segment-level feature, and the similarity matrix based on the point-to-point alignment confidence head to obtain the alignment matrix, the following steps may be taken: obtaining the first point-to-point feature based on the first segment-level feature; obtaining the second point-to-point feature based on the second segment-level feature; obtaining the alignment cost matrix of the first point-to-point feature and the second point-to-point feature; obtaining the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence by minimizing the total path cost based on the alignment cost matrix; and obtaining the alignment matrix based on the point-to-point correspondence.

[0043] For example, in this embodiment of the disclosure, the model first dynamically segments the input Query trajectory sequence and Candidate trajectory sequence, and obtains the high-dimensional features and length of each segment; calculates the similarity between each Query segment and Candidate segment to obtain a complete segment-to-segment similarity matrix; and achieves point-to-point alignment through SoftDTW to return a point-by-point alignment matrix between the Query trajectory and the Candidate trajectory.

[0044] a. Segment-token Encoder Introducing learnable segment tokens, the input trajectory point sequence is processed through multi-head self-attention. Divide into segments and output the length and weight of each segment. and corresponding segment-level features The total number of segments for each trajectory does not exceed [number missing]. Each segment can be based on The product of the input sequence length and the corresponding trajectory point index is calculated, with segments being continuous and without gaps. The weights of this module are shared by the Query / Candidate input sequences.

[0045] b. Segment Matching Module (SegmentMatcher) The input is the query segment features extracted by the segment encoder. and Candidata segment features The similarity matrix between segments is calculated through cross-attention. Then, by using row-wise normalization, the similarity distribution of each Query segment to all Candidate segments is obtained, thus providing initialization information for subsequent point-by-point alignment.

[0046] c. Pointwise AlignHeader The input is the output of the first two modules: the trajectory segment encoder. , , , , and the similarity matrix S obtained by the segmented matching module.

[0047] First, based on the length and feature vector of each segment, segment features are divided... Restored to point-by-point features The feature distance of each point is defined as the sum of the product of each feature vector. The alignment cost matrix is ​​calculated using Soft-DTW, and the point-by-point correspondence between the Query sequence and the Candidate sequence is obtained by minimizing the total path cost.

[0048] The model uses an end-to-end training approach, and the overall loss consists of two parts: 1. Alignment distance loss, calculated from the distance within the Soft-DTW cost matrix itself;

[0049] Where B is batch size, It is the sum of the point-to-point aligned feature distances of the i-th sample in the batch. This is the point-by-point alignment prior matrix extended from the similarity matrix S. It is an adjustable hyperparameter weight ( When =0, it degenerates to prior information obtained without using segmented matching.

[0050] 2. Confidence-supervised loss: The confidence level of the prediction is calculated using BCELoss, and the true value is obtained by interpolation based on the segment-by-segment correspondence between the input Query sequence and the Candidate sequence.

[0051] in, Let i be the pointwise aligned ground truth mask for the i-th sample in the batch. Predict the confidence mask for the corresponding model.

[0052] Overall loss function:

[0053] S104: Obtain the trajectory matching result based on the alignment matrix.

[0054] The trajectory matching result can be used to indicate one or more candidate trajectories that match the reference trajectory.

[0055] In this embodiment, a reference trajectory and a candidate trajectory library are acquired, wherein the candidate trajectory library includes multiple candidate trajectories. A first trajectory sequence is obtained by sampling the reference trajectory with a fixed sequence length, and a corresponding second trajectory sequence is obtained by sampling the candidate trajectories with a fixed sequence length. The first and second trajectory sequences are input into a pre-trained trajectory matching model to obtain an alignment matrix, which indicates the point-to-point correspondence between the first and second trajectory sequences. The trajectory matching result is obtained based on the alignment matrix. Therefore, fast and accurate matching of any irregular trajectory can be achieved, ensuring that the trajectory matching process can be adapted to personalized application scenarios, and effectively reducing cost losses during the trajectory matching process while improving efficiency.

[0056] Figure 2 This is a flowchart illustrating an arbitrary irregular trajectory matching method based on deep learning proposed in another embodiment of this disclosure.

[0057] like Figure 2 As shown, this deep learning-based method for matching arbitrary irregular trajectories includes: S201: Obtain the reference trajectory and candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories.

[0058] S202: Sample the reference trajectory with a fixed sequence length to obtain the first trajectory sequence, and sample the candidate trajectory with a fixed sequence length to obtain the corresponding second trajectory sequence.

[0059] S203: Input the first trajectory sequence and the second trajectory sequence into the pre-trained trajectory matching model to obtain the alignment matrix, wherein the alignment matrix is ​​used to indicate the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence.

[0060] The descriptions of S201-S203 can be found in the above embodiments, and will not be repeated here.

[0061] S204: Based on the alignment matrix, determine the matching confidence of each candidate trajectory with the reference trajectory.

[0062] Among them, the matching confidence score can be used to indicate the degree of matching between the candidate trajectory and the reference trajectory.

[0063] In this embodiment of the disclosure, when the matching confidence of each candidate trajectory with the reference trajectory is determined according to the alignment matrix, reliable reference information can be provided for the subsequent selection of the target trajectory.

[0064] S205: Select candidate trajectories with a matching confidence level greater than a preset threshold as target trajectories.

[0065] The specific value of the preset threshold can be flexibly adjusted according to the application scenario, and there are no restrictions on it.

[0066] The target trajectory refers to the trajectory that matches the entirety or part of the reference trajectory.

[0067] S206: Generate trajectory matching results based on the target trajectory.

[0068] In other words, in this embodiment of the present disclosure, after obtaining the alignment matrix, the matching confidence level between each candidate trajectory and the reference trajectory can be determined based on the alignment matrix; candidate trajectories with a matching confidence level greater than a preset threshold are taken as target trajectories; and trajectory matching results are generated based on the target trajectories. This effectively improves the practicality of the obtained trajectory matching results.

[0069] S207: Determine the positional correspondence between each target trajectory and the reference trajectory in the trajectory matching results.

[0070] Among them, the position correspondence can be used to indicate the correspondence between the target trajectory and the reference trajectory.

[0071] S208: Sort the target trajectories in the trajectory matching results according to the position correspondence.

[0072] In other words, in this embodiment of the present disclosure, after obtaining the trajectory matching result, the positional correspondence between each target trajectory and the reference trajectory in the trajectory matching result can be determined; based on the positional correspondence, the target trajectories in the trajectory matching result are sorted. Therefore, the indicative effect of the obtained trajectory matching result can be effectively improved.

[0073] In this embodiment, the matching confidence level between each candidate trajectory and the reference trajectory is determined based on the alignment matrix; candidate trajectories with a matching confidence level greater than a preset threshold are selected as target trajectories; and trajectory matching results are generated based on the target trajectories. This effectively improves the practicality of the obtained trajectory matching results. The positional correspondence between each target trajectory and the reference trajectory in the trajectory matching results is determined; and the target trajectories in the trajectory matching results are sorted according to the positional correspondence. This effectively improves the indicative effect of the obtained trajectory matching results.

[0074] In other words, in this embodiment of the present disclosure, after trajectory matching, the alignment matrix returned by the model can be used to filter candidates based on confidence level and threshold, and the original candidate trajectories with confidence levels greater than the threshold are added to the matching results. The trajectories in the matching results are then sorted according to the positional correspondence between each trajectory segment in the matching results and the original trajectory.

[0075] In summary, as described in the above embodiments, Figure 3 As shown, Figure 3 This is a flowchart of an arbitrary irregular trajectory matching algorithm based on deep learning proposed in this disclosure.

[0076] Among them, the trajectory matching model can be as follows: Figure 4 As shown, Figure 4 This is a schematic diagram of the matching model structure proposed in this disclosure. The specific trajectory matching process can be as follows: Figure 5 As shown, Figure 5 This is a schematic diagram of the trajectory matching process proposed in this disclosure, in which the red solid line represents the indicator trajectory and the green dashed line represents the matching result.

[0077] Figure 6 This is a schematic diagram of the structure of an arbitrary irregular trajectory matching device based on deep learning proposed in an embodiment of this disclosure.

[0078] like Figure 6 As shown, the deep learning-based arbitrary irregular trajectory matching device 60 includes: The first acquisition module 601 is used to acquire a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories; The sampling module 602 is used to sample the reference trajectory with a fixed sequence length to obtain a first trajectory sequence, and to sample the candidate trajectory with a fixed sequence length to obtain a corresponding second trajectory sequence; The input module 603 is used to input the first trajectory sequence and the second trajectory sequence into a pre-trained trajectory matching model to obtain an alignment matrix, wherein the alignment matrix is ​​used to indicate the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence; The second acquisition module 604 is used to obtain the trajectory matching result based on the alignment matrix.

[0079] It should be noted that the foregoing explanation of the deep learning-based arbitrary irregular trajectory matching method also applies to the deep learning-based arbitrary irregular trajectory matching device of this embodiment, and will not be repeated here.

[0080] In this embodiment, a reference trajectory and a candidate trajectory library are acquired, wherein the candidate trajectory library includes multiple candidate trajectories. A first trajectory sequence is obtained by sampling the reference trajectory with a fixed sequence length, and a corresponding second trajectory sequence is obtained by sampling the candidate trajectories with a fixed sequence length. The first and second trajectory sequences are input into a pre-trained trajectory matching model to obtain an alignment matrix, which indicates the point-to-point correspondence between the first and second trajectory sequences. The trajectory matching result is obtained based on the alignment matrix. Therefore, fast and accurate matching of any irregular trajectory can be achieved, ensuring that the trajectory matching process can be adapted to personalized application scenarios, and effectively reducing cost losses during the trajectory matching process while improving efficiency.

[0081] Figure 7 A block diagram of an exemplary computer device suitable for implementing embodiments of the present disclosure is shown. Figure 7 The computer device 12 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0082] like Figure 7 As shown, the computer device 12 is represented in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0083] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0084] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0085] Memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 7 Not shown; usually referred to as a "hard drive".

[0086] although Figure 7 As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disc drive for reading and writing to a removable non-volatile optical disc (e.g., a compact disc read-only memory (CD-ROM), a digital video disc read-only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this disclosure.

[0087] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of this disclosure.

[0088] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable human interaction with the computer device 12, and / or with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0089] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the deep learning-based arbitrary irregular trajectory matching method mentioned in the foregoing embodiments.

[0090] To implement the above embodiments, this disclosure also proposes a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements the deep learning-based arbitrary irregular trajectory matching method proposed in the foregoing embodiments of this disclosure.

[0091] To implement the above embodiments, this disclosure also proposes a computer program product that, when executed by an instruction processor, performs the deep learning-based arbitrary irregular trajectory matching method proposed in the foregoing embodiments of this disclosure.

[0092] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this disclosure all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0093] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.

[0094] This disclosure is intended to provide implementation schemes for users to selectively prevent the use or access to their personal information data. Specifically, this disclosure is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information is de-identified to protect user privacy.

[0095] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0096] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this disclosure, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0097] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of this disclosure includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of this disclosure pertain.

[0098] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0099] It should be understood that various parts of this disclosure can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0100] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0101] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0102] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A method for matching arbitrary irregular trajectories based on deep learning, characterized in that, include: Obtain a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories; The reference trajectory is sampled with a fixed sequence length to obtain a first trajectory sequence, and the candidate trajectory is sampled with a fixed sequence length to obtain a corresponding second trajectory sequence; The first trajectory sequence and the second trajectory sequence are input into a pre-trained trajectory matching model to obtain an alignment matrix, wherein the alignment matrix is ​​used to indicate the point-by-point correspondence between the first trajectory sequence and the second trajectory sequence; Based on the alignment matrix, the trajectory matching result is obtained.

2. The method as described in claim 1, characterized in that, The step of sampling the reference trajectory with a fixed sequence length to obtain a first trajectory sequence, and sampling the candidate trajectory with a fixed sequence length to obtain a corresponding second trajectory sequence, includes: Based on a preset fixed sequence length, multiple first trajectory points in the reference trajectory and multiple second trajectory points in the candidate trajectory are determined. Determine a first parameter corresponding to the first trajectory point and a second parameter corresponding to the second trajectory point, wherein the first parameter includes the three-dimensional coordinates, local discrete curvature, torsion, and angle value of the first trajectory point, and the second parameter includes the three-dimensional coordinates, local discrete curvature, torsion, and angle value of the second trajectory point; Construct the first trajectory sequence based on the first parameters of the first trajectory points; The second trajectory sequence is constructed based on the second parameters of the second trajectory points.

3. The method as described in claim 1, characterized in that, The trajectory matching model includes: a trajectory segment encoder, a segment matching module, and a point-by-point alignment confidence head; The alignment matrix is ​​obtained in the following manner: Based on the trajectory segment encoder, the first trajectory sequence is divided into multiple first trajectory sub-segments, and a first length weight and a first segment-level feature are determined for each first trajectory sub-segment; The trajectory segment encoder divides the second trajectory sequence into multiple second trajectory segments and determines the second length weight and second segment-level feature for each second trajectory segment. The segmented matching module processes the first segment-level features and the second segment-level features to obtain a similarity matrix, wherein the similarity matrix is ​​used to indicate the similarity distribution between each first segment-level feature and different second segment-level features; The alignment matrix is ​​obtained by processing the first length weight, the first segment-level feature, the second length weight, the second segment-level feature, and the similarity matrix based on the point-to-point alignment confidence head.

4. The method as described in claim 3, characterized in that, The process of processing the first length weight, the first segment-level feature, the second length weight, the second segment-level feature, and the similarity matrix based on the point-to-point alignment confidence head to obtain the alignment matrix includes: Based on the first segment-level features, obtain the first point-by-point features; Based on the second segment-level features, obtain the second point-by-point features; Obtain the alignment cost matrix of the first pointwise feature and the second pointwise feature; Based on the alignment cost matrix, the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence is obtained by minimizing the total path cost, and the alignment matrix is ​​obtained according to the point-to-point correspondence.

5. The method as described in claim 1, characterized in that, The step of obtaining the trajectory matching result based on the alignment matrix includes: Based on the alignment matrix, determine the matching confidence level between each candidate trajectory and the reference trajectory; The candidate trajectory with a matching confidence score greater than a preset threshold is taken as the target trajectory; Based on the target trajectory, the trajectory matching result is generated.

6. The method as described in claim 5, characterized in that, The method further includes: Determine the positional correspondence between each target trajectory and the reference trajectory in the trajectory matching result; Based on the location correspondence, the target trajectory in the trajectory matching result is sorted.

7. A deep learning-based device for matching arbitrary irregular trajectories, characterized in that, include: The first acquisition module is used to acquire a reference trajectory and a candidate trajectory library, wherein the candidate trajectory library includes multiple candidate trajectories; The sampling module is used to sample the reference trajectory with a fixed sequence length to obtain a first trajectory sequence, and to sample the candidate trajectory with a fixed sequence length to obtain a corresponding second trajectory sequence; An input module is used to input the first trajectory sequence and the second trajectory sequence into a pre-trained trajectory matching model to obtain an alignment matrix, wherein the alignment matrix is ​​used to indicate the point-to-point correspondence between the first trajectory sequence and the second trajectory sequence; The second acquisition module is used to obtain the trajectory matching result based on the alignment matrix.

8. A computer device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, in, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1-6.

Images