Track matching method combining target motion feature joint discrimination

Abstract

The invention discloses a trajectory matching method combined with target motion feature joint discrimination, and relates to the field of geographic information processing. The method comprises the following steps: randomly generating a trajectory starting point in three sub-regions, generating a simulation trajectory set according to the trajectory starting point and simulation trajectory parameters, obtaining a positive and negative sample set through noise introduction and cross matching, and re-sampling the positive and negative sample set to obtain a simulation trajectory training matching sample pair set; and calculating distance features and direction features of the simulation trajectory training matching sample pair, inputting the distance features and the direction features into an SVM classifier, constructing a trajectory matching classification model, and inputting two to-be-matched trajectories into the trajectory matching classification model to obtain a corresponding matching result. According to the method, the target motion element information in the trajectory can be effectively utilized to construct the matching model, and target trajectory matching is realized; the trajectory matching result is more accurate and reliable, and matching of the most similar path target trajectory is realized, so that powerful technical support is provided for optimization and management of ocean transportation.

Description

Technical Field

[0001] The invention relates to the field of geographic information processing, and in particular to a trajectory matching method combining joint discrimination with target motion features. Background Art

[0002] As an important link connecting all parts of the world, ocean transportation has an increasingly significant impact on economic activities. As the main tool for ocean transportation, the management and optimization of ship operation trajectories is the key to ensuring logistics efficiency and safety. Ship target trajectory matching technology not only plays an important role in shipping logistics, port management, maritime safety and other fields, but also has a far-reaching impact on marine resource development, environmental monitoring and other aspects. However, with the increase in the number of ships and the complexity of routes, traditional trajectory matching methods face huge challenges. Therefore, the development of an efficient ship target trajectory matching method is of great socio-economic significance for improving the overall efficiency and safety of ocean transportation.

[0003] Existing ship target trajectory matching algorithms mainly rely on simple geometric relationships or time series analysis, often ignoring the complexity and diversity of ship motion characteristics. These algorithms often exhibit problems such as low accuracy and weak robustness when processing large-scale, highly dynamic trajectory data. Specifically, existing algorithms often fail to accurately capture key motion characteristics such as the ship's speed, acceleration, and steering during the matching process, resulting in limited reliability and practicality of the matching results. Therefore, there is an urgent need for methods that use target motion element features to match target trajectories. Summary of the invention

[0004] In view of this, the present invention proposes a trajectory matching method that combines target motion features for joint identification. The method comprehensively considers the motion elements of the target's motion speed, acceleration, and direction, thereby improving the reliability of target trajectory matching.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A trajectory matching method combining target motion features with joint discrimination comprises the following steps:

[0007] Step 1: For the trajectory data of different densities in the simulation area R, three sub-areas of different scales are set, denoted as R = {r1, r2, r3};

[0008] Step 2: randomly generate m trajectory starting points in each sub-area of ​​the simulation area R, and construct a trajectory starting point set i=1,2,3;m≥7;

[0009] Step 3: According to the trajectory starting point set Combine trajectory simulation parameters to generate a set of simulation trajectories

[0010] Step 4: For the simulation trajectory set For each simulation trajectory in , two scales of random noise are introduced respectively, so as to obtain two simulation trajectory matching samples of different scales, which are recorded as a simulation trajectory matching sample pair, and finally generate a simulation trajectory matching sample pair set

[0011] Step 5: Match the simulation trajectory to the sample pair set All simulation trajectory matching sample pairs in are marked as positive sample pairs and given a label of 1 to obtain the positive sample pair set

[0012] Step 6: Match the simulation trajectory to the sample pair set Cross-match any two different simulation trajectory matching sample pairs in the simulation trajectory matching sample pair set to obtain multiple sets of matching sample pairs that do not belong to the simulation trajectory matching sample pair set. , 3m simulation trajectory matching sample pairs are randomly selected as negative sample pairs, and the label is given as 0 to obtain the negative sample pair set.

[0013] Step 7: Positive sample pair set And negative sample pair set Resample and construct a set of matching sample pairs for simulation trajectory training

[0014] Step 8: Calculate the simulation trajectory training matching sample pair set Each simulation trajectory in the training matches the distance features of the sample pairs, and obtains the distance feature set

[0015] Step 9: Calculate the simulation trajectory training matching sample pair set Each simulation trajectory in the training matches the directional features of the sample pairs, and obtains the directional feature set

[0016] Step 10: Match the simulation trajectory training sample pair set The distance feature set and direction feature set Input into the SVM classifier to build the trajectory matching classification model M SVM , the trajectory matching classification model M SVM The input is the distance feature and direction feature between two trajectories, and the output is a label 0 or 1;

[0017] Step 11: For two trajectories F to be matched re and S re, calculate the distance feature and direction feature between the two, and then input the distance feature and direction feature into the trajectory matching classification model M SVM In the above example, we get the trajectory F re and S re The corresponding label, if the label is 0, it means the trajectory F re and S re No match, if the label is 1, it means that the trajectory F re and S re Match and complete the trajectory matching by combining the joint identification of target motion features.

[0018] Furthermore, the simulation trajectory set in step 3 It can be expressed as:

[0019]

[0020] in, Represents a set of simulation trajectories The jth simulation trajectory in: includes three trajectory parameters and sampling time intervals, and They represent the time intervals corresponding to the three trajectories respectively. For different simulation trajectories, the sum of the time intervals corresponding to the three trajectories is the same; and They represent the initial speed of the corresponding time interval, and Respectively represent the initial direction of the corresponding time interval, and They represent the unit time speed variables of the corresponding time intervals, and They represent the unit time direction variables of the corresponding time intervals respectively; Δt represents the sampling time interval of the trajectory points. For different simulation trajectories, the sampling time interval of the trajectory points is the same.

[0021] Furthermore, the simulation trajectory matching sample pair set in step 4 Specifically:

[0022]

[0023] in, Matching sample pairs for simulation trajectories The jth simulation trajectory matching sample pair in: includes two simulation trajectory matching samples of different scales and They are all based on simulation trajectories Adding position noise corresponds to generating: and Respectively represent the simulation trajectory The noise means of different scales generated, and Respectively represent the simulation trajectory The resulting standard deviations for different scales.

[0024] Furthermore, the positive sample pair set in step 5 Specifically:

[0025]

[0026] Furthermore, the negative sample pair set in step 6 Specifically:

[0027]

[0028] in, is the set of negative sample pairs The kth negative sample pair: including and Matching samples of two simulation trajectories.

[0029] Further, step 7 includes the following steps:

[0030] Step 701: Set the positive sample pairs And negative sample pair set Merge, and in the merged new set, each element belongs to or

[0031] Step 702: Each element obtains the corresponding simulation trajectory training matching sample pair according to the trajectory information contained in itself.

[0032]

[0033] in, and Respectively The trajectory information of the two simulation trajectories matching samples in the corresponding element; express The label of the corresponding element, if The corresponding element belongs to but like The corresponding element belongs to but and Respectively The longitude, latitude and instantaneous heading of the first trajectory point of the corresponding simulation trajectory matching sample, and Respectively The longitude, latitude and instantaneous heading of the first trajectory point of the corresponding simulation trajectory matching sample, n represents and The number of trajectory points corresponding to the simulation trajectory matching sample, the value of n is determined by the simulation trajectory set The sum of the time intervals corresponding to the three trajectories in and the sampling time interval Δt are determined;

[0034] Step 703, finally obtain the simulation trajectory training matching sample pair set

[0035]

[0036] Furthermore, the distance feature set of the matching sample pairs in the simulation trajectory training in step 8 Specifically:

[0037]

[0038] in, Represents the distance feature set The lth simulation trajectory in is used to train the distance features of the matching sample pairs, hist() indicates solving the histogram statistical frequency, and log() indicates solving the logarithm.

[0039] Furthermore, the directional feature set of the simulated trajectory training matching sample pairs in step 9 Specifically:

[0040]

[0041] in, Represents the direction feature set The lth simulation trajectory in is used to train the directional features of the matching sample pairs.

[0042] Furthermore, the trajectory F to be matched in step 11 re and S re Specifically:

[0043] F re ={x F (1),y F (1),w F (1)),(x F (2),y F (2),w F (2)),...,

[0044] (x F (n),y F (n),w F (n))}

[0045] Sre ={x S (1),y S (1),w S (1)),(x S (2),y S (2),w S (2)),...,

[0046] (x S (n),y S (n),w S (n))}

[0047] Among them, x F (1) y F (1) and w F (1) represent the trajectory F re The longitude, latitude and instantaneous heading of the first track point, x S (1) y S (1) and w S (1) represent the simulation trajectory S re The longitude, latitude and instantaneous heading of the first track point, simulation track F re and S re The number of trajectory points is n.

[0048] Due to the adoption of the above technical solution, the beneficial effects of the present invention compared with the prior art are:

[0049] 1. The present invention proposes a trajectory matching method that combines target motion features with joint discrimination, which can effectively utilize the target motion element information in the trajectory, construct a matching model, and achieve target trajectory matching.

[0050] 2. The present invention can provide more accurate and reliable trajectory matching results and achieve the matching of the most similar path target trajectory, thereby providing strong technical support for the optimization and management of marine transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The figure is an overall flow chart of a trajectory matching method combined with target motion feature joint discrimination in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0053] A trajectory matching method that combines target motion features with joint discrimination, such as Figure 1 As shown, the following steps are included;

[0054] Step 1: For the trajectory data of different densities in the simulation area R, three sub-areas of different scales are set, denoted as R = {r1, r2, r3};

[0055] Specifically, sub-region r i It can be expressed as:

[0056]

[0057] In the formula, Respectively represent the region r i The longitude and latitude of the starting point, Respectively represent the region r i The longitude and latitude variables of .

[0058] Starting point sub-area r i is a rectangular area, so the coordinates of the other diagonal endpoint of the area are as follows:

[0059]

[0060] Step 2: randomly generate m trajectory starting points in each sub-area of ​​the simulation area R, and construct a trajectory starting point set i=1,2,3;m≥7;

[0061] Specifically, The starting point set of the trajectory The jth starting point in the subregion r i Inside.

[0062] Step 3: According to the trajectory starting point set Combine trajectory simulation parameters to generate a set of simulation trajectories

[0063] Step 4: For the simulation trajectory set For each simulation trajectory in , two scales of random noise are introduced respectively, so as to obtain two simulation trajectory matching samples of different scales, which are recorded as a simulation trajectory matching sample pair, and finally generate a simulation trajectory matching sample pair set

[0064] Step 5: Match the simulation trajectory to the sample pair set All simulation trajectory matching sample pairs in are marked as positive sample pairs and given a label of 1 to obtain the positive sample pair set

[0065] Step 6: Match the simulation trajectory to the sample pair set Cross-match any two different simulation trajectory matching sample pairs in the simulation trajectory matching sample pair set to obtain multiple sets of matching sample pairs that do not belong to the simulation trajectory matching sample pair set. , 3m simulation trajectory matching sample pairs are randomly selected as negative sample pairs, and the label is given as 0 to obtain the negative sample pair set.

[0066] Step 7: Positive sample pair set And negative sample pair set Resample and construct a set of matching sample pairs for simulation trajectory training

[0067] Step 8: Calculate the simulation trajectory training matching sample pair set Each simulation trajectory in the training matches the distance features of the sample pairs, and obtains the distance feature set

[0068] Step 9: Calculate the simulation trajectory training matching sample pair set Each simulation trajectory in the training matches the directional features of the sample pairs, and obtains the directional feature set

[0069] Step 10: Match the simulation trajectory training sample pair set The distance feature set and direction feature set Input into the SVM classifier to build the trajectory matching classification model M SVM , the trajectory matching classification model M SVM The input is the distance feature and direction feature between two trajectories, and the output is a label 0 or 1;

[0070] Step 11: For two trajectories F to be matched re and S re , calculate the distance feature and direction feature between the two, and then input the distance feature and direction feature into the trajectory matching classification model M SVM In the above example, we get the trajectory F re and S re The corresponding label, if the label is 0, it means the trajectory F re and S re No match, if the label is 1, it means that the trajectory F re and S re Match and complete the trajectory matching by combining the joint identification of target motion features.

[0071] Furthermore, the simulation trajectory set in step 3 It can be expressed as:

[0072]

[0073] in, Represents a set of simulation trajectories The jth simulation trajectory in: includes three trajectory parameters and sampling time intervals, and They represent the time intervals corresponding to the three trajectories respectively. For different simulation trajectories, the sum of the time intervals corresponding to the three trajectories is the same; and They represent the initial speed of the corresponding time interval, and Respectively represent the initial direction of the corresponding time interval, and They represent the unit time speed variables of the corresponding time intervals, and They represent the unit time direction variables of the corresponding time intervals respectively; Δt represents the sampling time interval of the trajectory points. For different simulation trajectories, the sampling time interval of the trajectory points is the same.

[0074] Furthermore, the simulation trajectory matching sample pair set in step 4 Specifically:

[0075]

[0076] in, Matching sample pairs for simulation trajectories The jth simulation trajectory matching sample pair in: includes two simulation trajectory matching samples of different scales and They are all based on simulation trajectories Adding position noise corresponds to generating: and Respectively represent the simulation trajectory The noise means of different scales generated, and Respectively represent the simulation trajectory The resulting standard deviations for different scales.

[0077] Furthermore, the positive sample pair set in step 5 Specifically:

[0078]

[0079] Furthermore, the negative sample pair set in step 6 Specifically:

[0080]

[0081] in, is the set of negative sample pairs The kth negative sample pair: including and Matching samples of two simulation trajectories.

[0082] Further, step 7 includes the following steps:

[0083] Step 701: Set the positive sample pairs And negative sample pair set Merge, and in the merged new set, each element belongs to or

[0084] Step 702: Each element obtains the corresponding simulation trajectory training matching sample pair according to the trajectory information contained in itself.

[0085]

[0086]

[0087] in, and Respectively The trajectory information of the two simulation trajectories matching samples in the corresponding element; express The label of the corresponding element, if The corresponding element belongs to but like The corresponding element belongs to but and Respectively The longitude, latitude and instantaneous heading of the first trajectory point of the corresponding simulation trajectory matching sample, and Respectively The longitude, latitude and instantaneous heading of the first trajectory point of the corresponding simulation trajectory matching sample, n represents and The number of trajectory points corresponding to the simulation trajectory matching sample, the value of n is determined by the simulation trajectory set The sum of the time intervals corresponding to the three trajectories in and the sampling time interval Δt are determined;

[0088] Step 703, finally obtain the simulation trajectory training matching sample pair set

[0089]

[0090] Furthermore, the distance feature set of the matching sample pairs in the simulation trajectory training in step 8 Specifically:

[0091]

[0092] in, Represents the distance feature set The lth simulation trajectory in is used to train the distance features of the matching sample pairs, hist() indicates solving the histogram statistical frequency, and log() indicates solving the logarithm.

[0093] Furthermore, the directional feature set of the simulated trajectory training matching sample pairs in step 9 Specifically:

[0094]

[0095] in, Represents the direction feature set The lth simulation trajectory in is used to train the directional features of the matching sample pairs.

[0096] Furthermore, the trajectory F to be matched in step 11 re and S re Specifically:

[0097] F re ={x F (1),y F (1),w F (1)),(x F (2),y F (2),w F (2)),...,

[0098] (x F (n),y F (n),w F (n))}

[0099] S re ={x S (1),y S (1),w S (1)),(x S (2),y S (2),w S (2)),...,

[0100] (x S (n),y S (n),w S (n))}

[0101] Among them, x F (1) y F (1) and w F (1) represent the trajectory F re The longitude, latitude and instantaneous heading of the first track point, x S (1) yS (1) and w S (1) represent the simulation trajectory S re The longitude, latitude and instantaneous heading of the first track point, simulation track F re and S re The number of trajectory points is n.

[0102] In summary, the present invention can effectively utilize the target motion element information in the trajectory, construct a matching model, and realize the matching of the target trajectory; the trajectory matching result is more accurate and reliable, and the matching of the most similar path target trajectory is realized, thereby providing strong technical support for the optimization and management of marine transportation.

[0103] Those skilled in the art will appreciate that the embodiments described are intended to help readers understand the principles of the present invention, and should be understood that the scope of protection of the present invention is not limited to the embodiments described. For those skilled in the art, the present invention may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A trajectory matching method combining target motion features with joint discrimination, characterized in that: The following steps are involved: Step 1: For the trajectory data of different densities in the simulation area R, three sub-areas of different scales are set, denoted as R = {r1, r2, r3}; Step 2: randomly generate m trajectory starting points in each sub-area of ​​the simulation area R, and construct a trajectory starting point set i=1,2,3;m≥7; Step 3: According to the trajectory starting point set Combine trajectory simulation parameters to generate a set of simulation trajectories Step 4: For the simulation trajectory set For each simulation trajectory in , two scales of random noise are introduced respectively, so as to obtain two simulation trajectory matching samples of different scales, which are recorded as a simulation trajectory matching sample pair, and finally generate a simulation trajectory matching sample pair set Step 5: Match the simulation trajectory to the sample pair set All simulation trajectory matching sample pairs in are marked as positive sample pairs and given a label of 1 to obtain the positive sample pair set Step 6: Match the simulation trajectory to the sample pair set Cross-match any two different simulation trajectory matching sample pairs in the simulation trajectory matching sample pair set to obtain multiple sets of matching sample pairs that do not belong to the simulation trajectory matching sample pair set. , 3m simulation trajectory matching sample pairs are randomly selected as negative sample pairs, and the label is given as 0 to obtain the negative sample pair set. Step 7: Positive sample pair set And negative sample pair set Resample and construct a set of matching sample pairs for simulation trajectory training Step 8: Calculate the simulation trajectory training matching sample pair set Each simulation trajectory in the training matches the distance features of the sample pairs, and obtains the distance feature set Step 9: Calculate the simulation trajectory training matching sample pair set Each simulation trajectory in the training matches the directional features of the sample pairs, and obtains the directional feature set Step 10: Match the simulation trajectory training sample pair set The distance feature set and direction feature set Input into the SVM classifier to build the trajectory matching classification model M SVM , the trajectory matching classification model M SVM The input is the distance feature and direction feature between two trajectories, and the output is a label 0 or 1; Step 11: For two trajectories F to be matched re and S re , calculate the distance feature and direction feature between the two, and then input the distance feature and direction feature into the trajectory matching classification model M SVM In the above example, we get the trajectory F re and S re The corresponding label, if the label is 0, it means the trajectory F re and S re No match, if the label is 1, it means that the trajectory F re and S re Match and complete the trajectory matching by combining the joint identification of target motion features.

2. A trajectory matching method combining target motion features with joint discrimination according to claim 1, characterized in that: Simulation trajectory collection in step 3 It can be expressed as: in, Represents a set of simulation trajectories The jth simulation trajectory in: includes three trajectory parameters and sampling time intervals, and They represent the time intervals corresponding to the three trajectories respectively. For different simulation trajectories, the sum of the time intervals corresponding to the three trajectories is the same; and They represent the initial speed of the corresponding time interval, and Respectively represent the initial direction of the corresponding time interval, and They represent the unit time speed variables of the corresponding time intervals, and They represent the unit time direction variables of the corresponding time intervals respectively; Δt represents the sampling time interval of the trajectory points. For different simulation trajectories, the sampling time interval of the trajectory points is the same.

3. The trajectory matching method according to claim 2, characterized in that: Simulation trajectory matching sample pair set in step 4 Specifically: in, Matching sample pairs for simulation trajectories The jth simulation trajectory matching sample pair in: includes two simulation trajectory matching samples of different scales and They are all based on simulation trajectories Adding position noise corresponds to generating: and Respectively represent the simulation trajectory The noise means of different scales generated, and Respectively represent the simulation trajectory The resulting standard deviations for different scales.

4. The trajectory matching method according to claim 3, characterized in that: The positive sample pair set in step 5 Specifically:

5. The trajectory matching method according to claim 4, characterized in that: The negative sample pair set in step 6 Specifically: in, is the set of negative sample pairs The kth negative sample pair: including and Matching samples of two simulation trajectories.

6. A trajectory matching method combining target motion feature joint discrimination according to claim 5, characterized in that: Step 7 includes the following steps: Step 701: collect the positive sample pairs And negative sample pair set Merge, and in the merged new set, each element belongs to or Step 702: Each element obtains the corresponding simulation trajectory training matching sample pair according to the trajectory information contained in itself. in, and Respectively The trajectory information of the two simulation trajectories matching samples in the corresponding element; express The label of the corresponding element, if The corresponding element belongs to but like The corresponding element belongs to but and Respectively The longitude, latitude and instantaneous heading of the first trajectory point of the corresponding simulation trajectory matching sample, and Respectively The longitude, latitude and instantaneous heading of the first trajectory point of the corresponding simulation trajectory matching sample, n represents and The number of trajectory points corresponding to the simulation trajectory matching sample, the value of n is determined by the simulation trajectory set The sum of the time intervals corresponding to the three trajectories in and the sampling time interval Δt are determined; Step 703, finally obtain the simulation trajectory training matching sample pair set 7. The trajectory matching method according to claim 6, characterized in that: The distance feature set of the matching sample pairs in simulation trajectory training in step 8 Specifically: in, Represents the distance feature set The lth simulation trajectory in is used to train the distance features of the matching sample pairs, hist() indicates solving the histogram statistical frequency, and log() indicates solving the logarithm.

8. The trajectory matching method according to claim 7, characterized in that: The directional feature set of the matching sample pairs in simulation trajectory training in step 9 Specifically: in, Represents the direction feature set The lth simulation trajectory in is used to train the directional features of the matching sample pairs.

9. The trajectory matching method according to claim 8, characterized in that: The trajectory F to be matched in step 11 re and S re Specifically: F re ={x F (1),y F (1),w F (1)),(x F (2),y F (2),w F (2)),..., ( x F ( n ), y F ( n ), w F (n))} S re ={x S (1),y S (1),w S (1)),(x S (2),y S (2),w S (2)),..., ( x S ( n ), y S ( n ), w S (n))} Among them, x F (1) y F (1) and w F (1) represent the trajectory F re The longitude, latitude and instantaneous heading of the first track point, x S (1) y S (1) and w S (1) represent the simulation trajectory S re The longitude, latitude and instantaneous heading of the first track point, simulation track F re and S re The number of trajectory points is n.

Images