Trajectory compression method and apparatus

Abstract

The application discloses a trajectory compression method. In view of the deficiency of the prior art, the solution idea of the application is that, based on the global optimization idea, the feature trajectory sub-chain obtained by simplifying and compressing a line segment is globally optimized, and multiple influence factors such as an offset angle, a perpendicular Euclidean distance, a trajectory segmentation number and the like are comprehensively considered, an evaluation index capable of comprehensively reflecting space and time characteristics of a compression result is established, and therefore the contradiction between compression efficiency and effective information reservation and between the original trajectory and the possibility of being as similar as possible is well balanced. The application also discloses a trajectory compression device. Compared with the prior art, the application can realize offline compression of all trajectories containing more effective information and having higher accuracy, and the overall operation accuracy and efficiency of trajectory compression is improved.

Description

Technical Field

[0001] This invention relates to the field of big data processing technology, and in particular to a trajectory compression method. Background Technology

[0002] With the rapid development of Location Based Services (LBS), Global Navigation Satellite System (GNSS), and the Internet of Things (IoT), and the increasing prevalence of mobile smart terminals, residents have accumulated massive amounts of mobile trajectory data. This provides a huge data foundation and application prospects for in-depth mining of valuable hidden travel information. The exponential growth in the scale of mobile trajectory data has led to the consumption of large amounts of data storage resources, and the performance of analysis based on massive amounts of mobile trajectory data has declined. Therefore, trajectory compression has become one of the current research hotspots.

[0003] Trajectory compression is a prerequisite for travel information mining, helping managers understand urban spatial structure and traffic operation, and providing a reference for residents' travel. Currently, scholars both domestically and internationally have conducted extensive research on trajectory compression, which can be mainly categorized as follows: The first category is line segment simplification compression, which can be further divided into offline compression and online compression. Both compression methods use line segments to approximate the original trajectory, compressing the original trajectory based on Euclidean distance, with the goal of reducing trajectory points as much as possible. (Wu Jiagao et al.) [1] By comprehensively analyzing the spatiotemporal characteristics of trajectory data, offline compression of trajectory points was achieved; Wang Qiandong [2] A fast algorithm based on the longest common subsequence similarity metric is proposed to achieve trajectory compression; Muckell et al. [3] The first method improves the trajectory point priority calculation method by using the SQUISH-E algorithm to compress local trajectory segments and reduce the number of trajectory points. The second method is trajectory compression based on road network structure. By combining this with a stable road network structure, it ensures that the trajectory matches the road segment, making it more meaningful for display. (Jiang Junwen et al.) [4] A trajectory compression technology based on driving characteristics was proposed; Zuo Yimeng et al. [5] An online trajectory compression method based on road network perception was implemented, including a distance-bounded hidden Markov map matching algorithm and an error-bounded efficient trajectory compression algorithm; Song RC et al. [6] The optimal compressed path model is constructed by combining the minimum description length to achieve road network trajectory compression. The third type is semantic compression, which makes it easier for readers to understand information such as trajectory origin and destination, travel route, and geographical location. References [7-8] propose semantic compression, which divides a trajectory into several semantically meaningful events and locations, but takes more compression time; Gao CM et al. [9]A clustering analysis model for semantic regions of interest based on multi-resolution synchronization is constructed, and the trajectory is compressed hierarchically.

[0004] While existing research on trajectory compression has yielded significant results, it also has certain limitations. Firstly, offline processing of line segments often results in the underutilization of crucial attribute information, leading to trajectory feature distortion. Since different attributes have varying degrees of impact on trajectory compression, relying solely on a single feature fails to comprehensively measure the trajectory, thus affecting compression accuracy. Secondly, offline compression depends on manually input parameters, and in practical applications, insufficient experience in setting appropriate parameter thresholds prevents the achievement of expected compression results. Furthermore, the lack of gradual analysis of trajectory point motion feature parameters hinders the development of adaptive dynamic optimization models for key feature parameters both internally and globally. Finally, research on constructing a multi-dimensional utility evaluation objective function for trajectory compression is scarce. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a trajectory compression method that can achieve offline compression of all trajectories containing a large amount of effective information and with high accuracy, thereby improving the overall accuracy and efficiency of trajectory compression.

[0006] The present invention specifically adopts the following technical solutions to solve the above-mentioned technical problems:

[0007] A trajectory compression method involves simplifying and compressing the original trajectory into line segments, and then performing global optimization on the feature trajectory sub-chains obtained from the line segment simplification and compression. The specific optimization model for the global optimization is as follows:

[0008] J = min(tad)

[0009]

[0010] Where tad=ε·d′ space +β·d′ time +μ·α′,d′ space d′ time , are the normalized values ​​of the modulus of the spatial distance matrix and the temporal distance matrix between all feature trajectory points in the current feature trajectory subchain and the original trajectory, respectively; α′ is the normalized value of the modulus of the offset angle matrix of all feature trajectory points in the current feature trajectory subchain; ε, β, and μ are the normalized values ​​of d′. space d′ time , the weight of α′, ε+β+μ=1, α(P i ) represents the feature trajectory point P in the current feature trajectory subchain. i The offset angle, PED(P) i ) represents the feature trajectory point P in the current feature trajectory subchain. iThe vertical Euclidean distance, n, between the original trajectory and the trajectory sub θ represents the number of feature trajectory points in the current feature trajectory subchain, and θ, D, and N represent the offset angle threshold, spatial distance threshold, and trajectory segment number threshold, respectively.

[0011] Preferably, any feature trajectory point P in the current feature trajectory subchain sub Spatial distance d to the original trajectory Trj space The specific definition is as follows:

[0012]

[0013] In the formula, n sub P represents the number of feature trajectory points in the current feature trajectory subchain; ⊥ P represents sub The formula for calculating the vertical coordinates to the original trajectory Trj and their corresponding position coordinates is as follows:

[0014]

[0015] P ⊥ ·l yi =P start ·l xi +m·[(P ⊥ ·l xi )-(P start ·l xi )]

[0016]

[0017]

[0018] In the formula, l xi l yi t i These represent the x-coordinate, y-coordinate, and timestamp of the i-th trajectory point in the original trajectory after its latitude and longitude have been converted into two-dimensional plane coordinates based on the same geographic coordinate system; P start P end ζ represents the first and last trajectory points of the original trajectory, respectively; ζ represents the internal correlation degree between the various trajectory sub-segments contained in the original trajectory Trj, that is, the ratio of the sum of the products of the length of the trajectory sub-segment and the corresponding importance of the trajectory sub-segment to the total number of trajectory points contained in the original trajectory, where n i This represents the total number of trajectory points contained in the original trajectory Trj. Dis(P) represents the importance of a segment of the trajectory. i ,P i+1 P represents the length of a trajectory segment calculated based on the latitude and longitude of adjacent original trajectory points; K ·lUi Point P K The location coordinates and the x or y coordinate l Ui The product of P, where P K Includes: P ⊥ P sub P start P end , l Ui Includes: l xi l yi .

[0019] Preferably, any feature trajectory point P in the current feature trajectory subchain sub Time distance d to the original trajectory Trj time The specific definition is as follows:

[0020]

[0021] Among them, l xi l yi These represent the x and y coordinates of the i-th trajectory point in the original trajectory after its latitude and longitude have been converted into two-dimensional plane coordinates based on the same geographic coordinate system.

[0022]

[0023]

[0024] In the formula, Represents the x-coordinate of the i-th time position; Represents the ordinate of the i-th time position; T = (T1, T2, ..., T... M ) represents the time discrete interval state parameter of any feature trajectory point in the current feature trajectory subchain; M represents the total number of time discrete interval state parameters; Trj′ represents the current feature trajectory subchain; T b (Trj′) represents the value of the state parameter of the b-th time discrete interval in the current feature subchain Trj′ out of the total number M; Δs represents the mean of the state parameters at the b-th time discrete interval of the current feature subchain; T This represents the time discrete interval state step size difference between the preceding and following adjacent feature trajectory points in the current feature trajectory subchain; and These represent the upper and lower time bounds of the original trajectory, respectively. and These represent the upper and lower time bounds of the current feature trajectory subchain, respectively; t i This represents the timestamp of the i-th trajectory point in the original trajectory.

[0025] Preferably, in any trajectory, trajectory point P iThe specific definition of the offset angle α is as follows:

[0026]

[0027] In the formula, i≥2, P i-1 P i+1 P i The trajectory points that are adjacent before and after, t i-1 t i t i+1 They are the trajectory points P i-1 P i P i+1 The timestamp of the record.

[0028] Preferably, an improved DP algorithm is used to simplify and compress the original trajectory. The improved DP algorithm is as follows: First, all trajectory points with an offset angle α greater than a preset offset angle threshold θ are searched from the original trajectory. These searched trajectory points, together with the first and last trajectory points of the original trajectory, constitute a preliminary compressed trajectory. Then, the DP algorithm is used to further compress the preliminary compressed trajectory to obtain the final compressed trajectory. Wherein, any trajectory point P in the original trajectory... i The specific definition of the offset angle α is as follows:

[0029]

[0030] Where i≥2, P i-1 P i+1 P i The trajectory points that are adjacent before and after, t i-1 t i t i+1 They are the trajectory points P i-1 P i P i+1 The timestamp of the record.

[0031] Preferably, the trajectory segment number threshold N is the number of trajectory segments obtained by compressing the original trajectory using a trajectory compression algorithm based on the MDL model.

[0032] Preferably, the global optimization is performed using a particle swarm optimization algorithm.

[0033] Preferably, d′ space d′ time The normalization process is based on the min-max standardization method, and α′ is based on the z-score standard score.

[0034] A trajectory compression apparatus includes a memory and a processor for storing and executing a computer program, wherein the computer program is used to implement the trajectory compression method described in any of the above technical solutions.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] This invention, based on the idea of ​​global optimization, performs global optimization on the feature trajectory sub-chains obtained by simplifying and compressing line segments. Compared with local trajectory segmentation compression, the selection of feature trajectory points is simpler and more accurate. It also comprehensively considers various influencing factors such as offset angle, vertical Euclidean distance, and the number of trajectory segments to establish an evaluation index that can comprehensively reflect the spatial and temporal characteristics of the compression result. This index is dimensionless and can effectively solve the problem of dynamically setting various parameters such as offset angle threshold, spatial distance threshold, and trajectory segment number threshold when compressing trajectories with different spatial location changes, different time zone spans, and different angular offsets. It better balances the contradiction between compression efficiency and the preservation of effective information and the similarity to the original trajectory as much as possible, and has strong practicality. Detailed Implementation

[0037] To address the shortcomings of existing technologies, the present invention addresses the issue by globally optimizing the feature trajectory sub-chains obtained from the simplified compression of line segments. It comprehensively considers various influencing factors such as offset angle, vertical Euclidean distance, and the number of trajectory segments to establish an evaluation index that can fully reflect the spatial and temporal characteristics of the compression result. This approach effectively balances the contradiction between compression efficiency and the preservation of effective information, as well as the goal of achieving the best possible similarity to the original trajectory.

[0038] The trajectory compression method proposed in this invention is as follows:

[0039] The original trajectory is simplified and compressed using line segments, and the resulting feature trajectory sub-chains are then globally optimized. The specific optimization model for this global optimization is as follows:

[0040] J = min(tad)

[0041]

[0042] Where tad=ε·d′ space +β·d′ time +μ·α′,d′ space d′ time , are the normalized values ​​of the modulus of the spatial distance matrix and the temporal distance matrix between all feature trajectory points in the current feature trajectory subchain and the original trajectory, respectively; α′ is the normalized value of the modulus of the offset angle matrix of all feature trajectory points in the current feature trajectory subchain; ε, β, and μ are the normalized values ​​of d′. space d′ time , the weight of α′, ε+β+μ=1, α(P i ) represents the feature trajectory point P in the current feature trajectory subchain.i The offset angle, PED(P) i ) represents the feature trajectory point P in the current feature trajectory subchain. i The vertical Euclidean distance, n, between the original trajectory and the trajectory sub θ represents the number of feature trajectory points in the current feature trajectory subchain, and θ, D, and N represent the offset angle threshold, spatial distance threshold, and trajectory segment number threshold, respectively.

[0043] To facilitate public understanding, the technical solution of the present invention will be described in detail below through a preferred embodiment:

[0044] Step 1: Data Preprocessing

[0045] Due to the instability of mobile sensor operation, packet loss during data transmission, and other uncontrollable external factors, the raw data often contains incomplete trajectory sequences, missing or incorrect attribute information. Therefore, to ensure the validity of mobile trajectory data, it is necessary to preprocess the raw data before trajectory compression. This involves combining advanced big data computing and analysis techniques to remove invalid and redundant trajectory points, achieving preprocessing such as cleaning, denoising, and deletion of the raw data to obtain valid trajectory data. The data preprocessing of this invention can employ various existing data preprocessing methods. The specific preprocessing steps in this embodiment are as follows:

[0046] Step 1: Clean the data based on the abnormal values ​​(i.e. invalid data) recorded by the Flag attribute in the original data table.

[0047] Step 2: Calculate the error value of noise points in the original data using the Kalman filter method, and then perform noise removal processing on the original data to eliminate noise points.

[0048] Step 3: Extract the original data with the same desensitized trajectory number, and compare whether the recorded timestamp, longitude, latitude and other information are consistent. If the above recorded information is consistent, delete the duplicate data and keep only one data; otherwise, continue to search for duplicate data, and finally achieve data deduplication.

[0049] Step 4: Repeat the above preprocessing steps to output all valid trajectory data.

[0050] Step 2: Simplify and compress the preprocessed effective trajectory into line segments:

[0051] Trajectory compression algorithms should retain appropriate trajectory points; dynamic programming (DP) algorithm. [11-12]This is a compression algorithm based on linear fitting to preserve important points. The main idea is to use the line connecting the start and end points of the original trajectory as an approximation of the original trajectory. The Euclidean distance from each point in the original trajectory to this approximation is calculated sequentially. Trajectories with Euclidean distances greater than a given threshold are used as dividing points, splitting the current trajectory into two parts. This process is repeated until the distances from all original trajectory points to their corresponding approximation segments are less than the distance threshold. The line connecting these dividing points represents the compressed trajectory.

[0052] The line segment simplification and compression of this invention can employ various existing schemes, such as the classic dynamic programming (DP) algorithm or various improved DP algorithms; this embodiment uses an improved DP algorithm, as detailed below:

[0053] Step 1: Search for all trajectory points in the original trajectory whose offset angle α is greater than the preset offset angle threshold θ. These searched trajectory points, together with the first and last trajectory points of the original trajectory, form a preliminary compressed trajectory.

[0054] Wherein, any trajectory point P i The specific definition of the offset angle α is as follows:

[0055]

[0056] Where i≥2, P i-1 P i+1 P i The trajectory points that are adjacent before and after, t i-1 t i t i+1 They are the trajectory points P i-1 P i P i+1 The timestamp of the record.

[0057] According to the above definition, the offset angles of all trajectory points in the original trajectory, except for the first and last trajectory points, can be calculated. Trajectory points with offset angles α less than or equal to the offset angle threshold θ are deleted. The remaining trajectory points with larger offset angles α are used as feature trajectory points along with the first and last trajectory points of the original trajectory. These feature trajectory points are then connected by line segments to obtain the preliminary compressed trajectory.

[0058] Step 2: Use the DP algorithm to further compress the initial compressed trajectory to obtain the final compressed trajectory;

[0059] Using the initial compressed trajectory as input to the DP algorithm, and according to a preset PED threshold D, the DP algorithm further compresses the initial compressed trajectory to obtain the final compressed trajectory, also known as the feature trajectory sub-chain. The DP algorithm is a classic algorithm in this field, and will not be elaborated upon here for the sake of brevity.

[0060] Step 3: Dynamically adjust the compression parameters of the line segment simplification and compression algorithm and perform global optimization on the obtained feature trajectory subchain:

[0061] The compression parameters of the improved DP algorithm used in this embodiment are the offset angle threshold θ and the PED threshold D. For the original trajectory, a new feature trajectory sub-chain can be obtained by adjusting the compression parameters once. The purpose of global optimization is to find the globally optimal feature trajectory sub-chain and its corresponding compression parameters from these feature trajectory sub-chains according to a certain evaluation index.

[0062] The optimization model used in this invention is as follows:

[0063] J = min(tad)

[0064]

[0065] Where tad=ε·d′ space +β·d′ time +μ·α′,d′ space d′ time , are the normalized values ​​of the modulus of the spatial distance matrix and the temporal distance matrix between all feature trajectory points in the current feature trajectory subchain and the original trajectory, respectively; α′ is the normalized value of the modulus of the offset angle matrix of all feature trajectory points in the current feature trajectory subchain; ε, β, and μ are the normalized values ​​of d′. space d′ time , the weight of α′, ε+β+μ=1, α(P i ) represents the feature trajectory point P in the current feature trajectory subchain. i The offset angle, PED(P) i ) represents the feature trajectory point P in the current feature trajectory subchain. i The vertical Euclidean distance, n, between the original trajectory and the trajectory sub θ represents the number of feature trajectory points in the current feature trajectory subchain, and θ, D, and N represent the offset angle threshold, spatial distance threshold, and trajectory segment number threshold, respectively.

[0066] The spatial and temporal distances between any feature trajectory point in the current feature trajectory subchain and the original trajectory can be defined using existing spatial and temporal distance definitions. For example, references [13-14] have proposed spatial distance metrics between points and subtrajectories, including vertical distance, horizontal distance, and angular distance; references [15-16] have considered the temporal information in the trajectory sequence and defined the synchronization temporal distance between points and subtrajectories.

[0067] However, existing studies do not consider the intrinsic correlation and similarity between the original trajectory points and trajectory segments in terms of spatial distance, and they define temporal distance in a coarse-grained manner, failing to efficiently mine and analyze the temporal information of the original trajectory by combining long time spans and the degree of temporal discrepancy. Therefore, in this embodiment, the spatial and temporal distances between any feature trajectory point in the current feature trajectory subchain and the original trajectory are defined with new definitions. The spatial distance fully considers the intrinsic correlation and similarity between the original trajectory points and trajectory segments, while the temporal distance considers the impact of long time spans and the degree of temporal discrepancy, as detailed below:

[0068] (1) Any feature trajectory point P in the current feature trajectory subchain sub Spatial distance d to the original trajectory Trj space The specific definition is as follows:

[0069]

[0070] In the formula, n sub P represents the number of feature trajectory points in the current feature trajectory subchain; ⊥ P represents sub The formula for calculating the vertical coordinates to the original trajectory Trj and their corresponding position coordinates is as follows:

[0071]

[0072] P ⊥ ·l yi =P start ·l xi +m·[(P ⊥ ·l xi )-(P start ·l xi )]

[0073]

[0074]

[0075] In the formula, l xi l yi t i These represent the x-coordinate, y-coordinate, and timestamp of the i-th trajectory point in the original trajectory after its latitude and longitude have been converted into two-dimensional plane coordinates based on the same geographic coordinate system; P start P end ζ represents the first and last trajectory points of the original trajectory, respectively; ζ represents the internal correlation degree between the various trajectory sub-segments contained in the original trajectory Trj, that is, the ratio of the sum of the products of the length of the trajectory sub-segment and the corresponding importance of the trajectory sub-segment to the total number of trajectory points contained in the original trajectory, where n iThis represents the total number of trajectory points contained in the original trajectory Trj. Dis(P) represents the importance of a segment of the trajectory. i ,P i+1 P represents the length of a trajectory segment calculated based on the latitude and longitude of adjacent original trajectory points; K ·l Ui Point P K The location coordinates and the x or y coordinate l Ui The product of P, where P K Includes: P ⊥ P sub P start P end , l Ui Includes: l xi l yi , such as P ⊥ ·l xi Represents the vertical coordinate P ⊥ with x-coordinate l xi The product between them.

[0076] (2) Any feature trajectory point P in the current feature trajectory subchain sub Time distance d to the original trajectory Trj time The specific definition is as follows:

[0077]

[0078] Among them, l xi l yi These represent the x and y coordinates of the i-th trajectory point in the original trajectory after its latitude and longitude have been converted into two-dimensional plane coordinates based on the same geographic coordinate system.

[0079]

[0080]

[0081] In the formula, Represents the x-coordinate of the i-th time position; Represents the ordinate of the i-th time position; T = (T1, T2, ..., T... M ) represents the time discrete interval state parameter of any feature trajectory point in the current feature trajectory subchain; M represents the total number of time discrete interval state parameters; Trj′ represents the current feature trajectory subchain; T b (Trj′) represents the value of the state parameter of the b-th time discrete interval in the current feature subchain Trj′ out of the total number M; Δs represents the mean of the state parameters at the b-th time discrete interval of the current feature subchain; TThis represents the time discrete interval state step size difference between the preceding and following adjacent feature trajectory points in the current feature trajectory subchain; and These represent the upper and lower time bounds of the original trajectory, respectively. and These represent the upper and lower time bounds of the current feature trajectory subchain, respectively; t i This represents the timestamp of the i-th trajectory point in the original trajectory.

[0082] Based on the above d space d time By using the definitions of the current feature trajectory subchain and the offset angle α mentioned above, we can obtain the spatial and temporal distances between all feature trajectory points in the current feature trajectory subchain and the original trajectory, as well as the offset angles of all feature trajectory points in the current feature trajectory subchain. This series of spatial distances, temporal distances, and offset angles constitutes the spatial distance matrix d between all feature trajectory points in the current feature trajectory subchain and the original trajectory. space Time distance matrix d time And the offset angle matrix α of all feature trajectory points in the current feature trajectory subchain, the modulus of these three matrices ||d space ||、||d time ||、||α|| can be used as three dimensions of features for evaluating compression effect.

[0083] To make the established optimization model more accurate, this embodiment further refines ||d space ||、||d time The values ​​|| and ||α|| are standardized; specifically, the min-max standardization method is used to standardize ||d. space ||and||d time The result is normalized to map to the range [0,1], and then standardized based on the z-score. The normalized result is ||d||. space ||and||d time || Let them be d′ respectively space d′ time The standardized ||α|| is denoted as α′, and the weights of the three are ε, β, and μ, respectively, ε+β+μ=1. The weight allocation can be adjusted according to the different focuses of the actual attention.

[0084] Specific optimization algorithms can employ existing techniques such as particle swarm optimization, annealing, and ant colony optimization. This embodiment uses particle swarm optimization, where the fitness function of each particle is... The threshold N for the number of trajectory segments is the number of trajectory segments obtained by compressing the original trajectory using a trajectory compression algorithm based on the MDL model (see reference

[10] ).

[0085] To verify the technical effectiveness of this invention, the method was compared with existing classical techniques. One week's worth of open-source taxi GPS tracks from a certain city were selected, with data collection intervals of 5 seconds per track, totaling 1,467,026 tracks. The algorithm program of this invention was written in Python, developed using PyCharm 2022, and ran on an AMD Ryzen 7 5800H x64 processor (3.20GHz), 16GB of RAM, and Windows 10 operating system.

[0086] (1) Sensitivity analysis of algorithm parameters

[0087] The algorithm is evaluated using metrics such as compression ratio σ and mean error MErr. Further research is conducted on the effects of distance threshold D and angle threshold θ on σ and MErr. As multiple compression thresholds are dynamically adjusted, the spatiotemporal attribute information reflected in the trajectory data may change in different ways.

[0088] (2) Algorithm performance analysis

[0089] In the trajectory compression stage, the classic MDL algorithm was selected as the comparison object for the method of this invention, verifying that the method of this invention has certain compression advantages compared with the classic algorithm. The algorithm of this invention and the MDL algorithm were applied to a real taxi GPS trajectory dataset to compare their accuracy and simplicity.

[0090] Step 1: Accuracy Evaluation. Under the premise of preserving trajectory characteristics and effectively reducing data size, the angle threshold θ and distance threshold D are set to [29.571°, 62.691°] and [62.632m, 133.411m] respectively, at which point tad reaches its minimum value. With the same number of trajectory segments, this invention achieves a smaller value than the MDL algorithm, and tad converges to a stable interval more quickly, indicating higher accuracy in trajectory compression.

[0091] Step 2: Simplification Evaluation. The simplification metric is defined as the number of trajectory segment compressions. That is, under the same trajectory data parameters, the number of trajectory segment compressions corresponding to the present invention and the MDL algorithm is shown in Table 1. As shown in Table 1, the trajectory compression variation of the present invention is reduced by an average of 56.056% compared to the MDL algorithm, indicating that the method of the present invention has better simplification.

[0092] Table 1 Comparison results of the two methods

[0093]

[0094] References:

[0095] [1] Wu Jiagao, Qian Keyu, Liu Min, et al. Hybrid trajectory compression algorithm based on comprehensive spatiotemporal characteristics [J]. Computer Applications, 2015, 35(5): 1209-1212.

[0096] [2] Wang Qiandong. A fast algorithm for similarity measurement of classical trajectories [J]. Systems Engineering and Electronics, 2020, 42(10):2189-2196.

[0097] [3]Muckell J,Olsen PW,Hwang JH,et al.Compression of trajectorydata:a comprehensive evaluation and new approach[J].GeoInformatica,2014,18(3):435-460.

[0098] [4] Jiang Junwen, Zhang Kai, Wang Xiaoling, et al. Trajectory compression technology based on driving characteristics [J]. Computer Science and Exploration, 2016, 10(9): 1240-1249.

[0099] [5] Zuo Yimeng, Lin Xuelian, Ma Shuai, et al. Online trajectory compression method based on road network perception [J]. Journal of Software, 2018, 29(3):734-755.

[0100] [6]Song Renchu, Sun Weiwei, Zheng Baihua, et al. PRESS: a novel framework of trajectory compression in road net-works[J]. Proceedings of the VLDBEndowment, 2014, 7(9): 661-672.

[0101] [7]Schmid F,Richter KF,Laube P.Semantic trajectory com-pression[M].Advances in Spatial and Temporal Databases.Berlin,Heidelberg:Springer,2009:411-416.

[0102] [8]Richter K F,Schmid F,Laube P.Semantic trajectory com-pression:representing urban movement in a nutshell[J].Journal of Spatial InformationScience,2012(4):3-30.

[0103] [9]Chongming Gao,Yi Zhao,Ruizhi Wu,et al.Semantic trajectorycompression via multi-resolution synchronization-based clustering[J].Knowledge-Based Systems,2019,174:177-193.

[0104]

[10] LEE J G,HAN J,WHANG K Y.Trajectory clustering:a partition-and-group framework[C].ACM SIGMOD International Conference on Management ofData.ACM,2007,593-604.

[0105]

[11] DOUGLAS DH,PEUCKER TK.Algorithms for the reduction of the numberof points required to represent a digitized line or its caricaturelJ].Cartographica:The International Journal for Geographic Information andGeovisualization,1973,10(2):112-122.

[0106]

[12] HERSHBERGER JE,SNOEYINK J.Speeding up the douglas-peucker line-simplification algorithm[M].International Symposium on Spatial DataHandling.Berlin:Springer,1992:134-143.

[0107]

[13] Yang Shuliang, Bi Shuoben, Nkunzimana A, et al. A spatial clustering method for taxi passenger trajectories[J]. Computer Engineering and Applications, 2018, 54(14):249-255.

[0108]

[14] Guo Wei, Tang Huifeng. Research on aircraft trajectory pattern mining based on trajectory clustering [J]. Computer Applications Research, 2020, 37(2):416-420.

[0109]

[15] Meratnia N,By R AD.Spatiotemporal Compression Techniques for Moving Point Objects[M].Advances in Database Technology-EDBT 2004.SpringerBerlin Heidelberg,2004:765-782.

[0110]

[16] Chen Yu, Jiang Wei, Zhou Ji'en. Trajectory data compression method based on prediction model [J]. Computer Applications, 2018, 38(1):171-175.

Claims

1. A trajectory compression method, comprising simplifying and compressing the original movement trajectory data into line segments, and globally optimizing the feature trajectory sub-chains obtained from the simplified and compressed line segments; characterized in that, The specific optimization model for global optimization is as follows: in, , , These are the normalized values ​​of the spatial distance matrix and the temporal distance matrix between all feature trajectory points in the current feature trajectory subchain and the original trajectory, respectively. This is the standardized value of the modulus of the offset angle matrix of all feature trajectory points in the current feature trajectory subchain. , , They are respectively , , The weight, , For the feature trajectory points in the current feature trajectory subchain The offset angle, For the feature trajectory points in the current feature trajectory subchain The vertical Euclidean distance between the original trajectory and the original trajectory. This represents the number of feature trajectory points in the current feature trajectory subchain. , D , N These are the offset angle threshold, spatial distance threshold, and trajectory segment number threshold, respectively; any feature trajectory point in the current feature trajectory subchain. To the original trajectory spatial distance The specific definition is as follows: In the formula, This indicates the number of feature trajectory points in the current feature trajectory subchain; express To the original trajectory The vertical coordinates of are calculated using the following formula: In the formula, , , These represent the first and second digits of the original trajectory, respectively. i Each trajectory point is based on the same geographic coordinate system, with its latitude and longitude converted into two-dimensional plane coordinates, including the horizontal and vertical coordinates and the recorded timestamp. , These represent the first and last points of the original trajectory, respectively; Represents the original trajectory The degree of internal correlation between the various trajectory segments contained in the original trajectory is the ratio of the sum of the products of the length of each trajectory segment and the importance of its corresponding segment to the total number of trajectory points in the original trajectory. Represents the original trajectory The total number of trajectory points contained in it. Indicates the importance of the proportion of trajectory segments. This represents the length of a trajectory segment calculated based on the latitude and longitude of the preceding and following adjacent original trajectory points; Point Location coordinates and x or y coordinates The product of, where include: , , , , include: , Any feature trajectory point in the current feature trajectory subchain To the original trajectory Time distance The specific definition is as follows: in, , These represent the first and second digits of the original trajectory, respectively. i Each trajectory point is based on the same geographic coordinate system, after converting latitude and longitude into two-dimensional plane coordinates, resulting in horizontal and vertical coordinates. In the formula, Indicates the first The x-coordinate of each time location point; Indicates the first The ordinate of each time location point; This represents the time discrete interval state parameter of any feature trajectory point in the current feature trajectory subchain; This represents the total number of state parameters at discrete time intervals. Indicates the current feature trajectory subchain; Indicates the current feature subchain In total The Middle The values ​​of state parameters at discrete time intervals; Indicates the current feature subchain number 1 The mean value of state parameters at discrete time intervals; This represents the time discrete interval state step size difference between the preceding and following adjacent feature trajectory points in the current feature trajectory subchain; and These represent the upper and lower time bounds of the original trajectory, respectively. and These represent the upper and lower time bounds of the current feature trajectory subchain, respectively; Represents the first in the original trajectory i The timestamps of each trajectory point are recorded.

2. The trajectory compression method as described in claim 1, characterized in that, Trajectory points in any trajectory offset angle The specific definition is as follows: In the formula, i ≥2, , They represent Adjacent trajectory points before and after, , , Trajectory points , , The timestamp of the record.

3. The trajectory compression method as described in claim 1, characterized in that, An improved dynamic programming (DP) algorithm is used to simplify and compress the original trajectory into line segments. The improved DP algorithm is as follows: First, the offset angle is searched from the original trajectory. Greater than the preset offset angle threshold All trajectory points are searched, and these trajectory points, together with the first and last trajectory points of the original trajectory, form a preliminary compressed trajectory; then, the DP algorithm is used to further compress the preliminary compressed trajectory to obtain the final compressed trajectory; where any trajectory point in the original trajectory... offset angle The specific definition is as follows: in, i ≥2, , They represent Adjacent trajectory points before and after, , , Trajectory points , , The timestamp of the record.

4. The trajectory compression method as described in claim 1, characterized in that, The threshold number of trajectory segments N The number of trajectory segments obtained by compressing the original trajectory using a trajectory compression algorithm based on the MDL model.

5. The trajectory compression method as described in claim 1, characterized in that, The global optimization is performed using the particle swarm optimization algorithm.

6. The trajectory compression method as described in claim 1, characterized in that, , The normalization process is based on the min-max normalization method. The standardization process is based on the z-score standard score.

7. A trajectory compression device, comprising a memory and a processor for storing and executing computer programs respectively, characterized in that, The computer program is used to implement the trajectory compression method as described in any one of claims 1 to 6.

Images