Urban traffic mode recognition system based on trajectory clustering

Abstract

The invention relates to the technical field of traffic mode recognition, in particular to an urban traffic mode recognition system based on trajectory clustering, which comprises a trajectory credibility rejection module, a trajectory semantic clustering module, a trajectory evolution stage recognition module, a dynamic parking judgment module and an urban traffic mode generation module. According to the method, boundary dispersion rate difference analysis is carried out on traffic trajectory data, continuous cross-grid anomalies are identified, unstable sections in the trajectory are effectively eliminated, the spatial direction distribution concentration degree is taken as a basis, the trajectory coverage rate difference of the clustering center in adjacent time periods is calculated, and the spatial evolution trend of the trajectory is quantified; according to the method, dynamic description of a traffic mode in the time dimension is achieved, the misjudgment risk caused by a single time or space index is avoided in combination with double indexes of the space overlapping degree and the parking persistence between track fragments, and efficient identification of urban area travel function types is achieved through similarity matching between track features and travel features.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic pattern recognition, and in particular to an urban traffic pattern recognition system based on trajectory clustering. Background Art

[0002] The field of traffic pattern recognition technology mainly focuses on the identification, modeling and analysis of the travel behavior, travel rules and travel structure of traffic participants based on traffic data. This field uses trajectory data mining, spatiotemporal data analysis, machine learning, complex network modeling, graph theory and other methods to extract traffic behavior patterns from both individual and group levels, and identify core elements such as travel modes, commuting routes, and transfer behaviors. Traffic pattern recognition has broad application value in traffic demand forecasting, congestion analysis, traffic scheduling optimization, public transportation resource allocation, travel behavior modeling, etc. It can provide data-driven decision support for urban traffic management, traffic planning and operational efficiency improvement, and is one of the basic supporting technologies for realizing refined management and intelligent operation of the transportation system.

[0003] Among them, the urban traffic mode recognition system is a system used to identify the travel behavior patterns of urban residents under various public transportation modes. By processing and analyzing the trajectory data and passenger flow data of travel modes such as taxis, online car-hailing, buses, and subways, it explores the travel activity patterns of residents and identifies the behavioral characteristics of different travel modes and their competitive and cooperative relationships. The system can identify residents' choice preferences and behavior patterns between different transportation modes, analyze the competitive and cooperative relationships between different transportation modes, and then quantify the impact mechanism of the urban built environment on transportation. Its uses include supporting the operation monitoring of urban transportation systems, travel structure optimization, public transportation service capacity assessment, and resource allocation and scheduling.

[0004] Traditional recognition systems mainly rely on the overall morphology and spatiotemporal distribution of trajectory data to model traffic behavior. They lack an effective screening mechanism for trajectory quality, which causes data with abnormal points such as drift and fracture to be directly involved in the analysis, reducing the accuracy of the recognition results. When clustering trajectories, they are often based on distance similarity or morphological similarity indicators, and do not fully consider the aggregation characteristics of the trajectory spatial direction, which can easily lead to mixed clustering of different travel modes and affect the effect of traffic mode discrimination. The trajectory change trend is mostly presented in the form of static graphic clustering, which fails to describe the evolution process of traffic behavior and makes it difficult to identify the dynamic characteristics of the traffic system. When determining regional functions, they are based only on the stay time, and do not integrate the joint judgment of spatial overlap and continuity, resulting in a large number of false identifications of non-real stops. For example, short stops frequently occur in transportation hub areas. If they are determined only based on time, they are easily misidentified as stop areas, thus affecting the accurate description of regional traffic functions. The recognition stability and dynamic adaptability in a multi-source trajectory data environment are not conducive to building a high-resolution urban traffic behavior model. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose an urban traffic pattern recognition system based on trajectory clustering.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an urban traffic pattern recognition system based on trajectory clustering, the system comprising:

[0007] The trajectory credibility elimination module obtains the location points of the traffic trajectory data and divides them into grid units, calculates the boundary discrete rate difference of the trajectory data, determines whether the continuous cross-grid difference exceeds the stability threshold, marks and eliminates abnormal trajectories with more than three unstable boundaries, and generates a credible trajectory data set;

[0008] The trajectory semantic clustering module obtains trajectory direction and spatial distribution characteristics based on the trusted trajectory data set, calculates the concentration degree of spatial direction distribution, semantically clusters the trajectory data according to the degree of spatial concentration, and generates a semantic trajectory clustering set;

[0009] The trajectory evolution stage identification module calls the semantic trajectory clustering set, obtains the trajectory coverage rate of the cluster center, calculates the difference of trajectory coverage rates in adjacent time periods, divides the trajectory coverage rate changes into three stages: stable, diffuse, and contractive, and generates the trajectory evolution stage type;

[0010] The dynamic stop distinguishing module obtains the trajectory segments in the area according to the trajectory evolution stage type, calculates the spatial overlap and stop duration between adjacent trajectory segments, marks the trajectory segments that meet the conditions as the dynamic stop state, and generates a dynamic stop area set.

[0011] The present invention has the following improvements: the credible trajectory data set includes a trajectory number identifier, a discarded trajectory index and a valid trajectory range; the semantic trajectory clustering set includes a trajectory semantic label, a trajectory cluster boundary configuration and a trajectory concentration direction; the trajectory evolution stage type is specifically a steady evolution stage type, a centripetal contraction stage type and a trajectory diffusion stage type; the dynamic parking area set includes an area number, a parking segment index and a spatial overlapping structure.

[0012] The present invention is improved in that the trajectory credibility elimination module comprises:

[0013] The grid division submodule obtains the location points of traffic trajectory data, collects the city boundary coordinate group and the trajectory point location coordinate set, divides the entire domain into equidistant grid units according to the city boundary coordinate group and the latitude and longitude scale ratio, calls the trajectory point location coordinate set to determine the grid number where it is located, establishes the mapping index relationship between the trajectory point and the grid unit, and generates the trajectory grid mapping structure;

[0014] The discrete rate difference calculation submodule extracts the trajectory segments located in the edge grid cells of each trajectory based on the trajectory grid mapping structure, collects the direction vector of each segment, calculates the concentration degree of the direction amplitude vector of each segment in the corresponding grid, obtains the concentration value of the direction vector of the adjacent grid and constructs the direction difference term, using the formula:

[0015] ;

[0016] The boundary discrete rate difference between adjacent trajectory segments is obtained by operation, and it is determined whether the difference exceeds the stability threshold and the stable state of the trajectory segment is marked to obtain the discrete difference value in the boundary direction;

[0017] in, The representative trajectory is , The difference in discreteness between grids, , Represent the trajectories in , The grid , Direction angles, , are the number of trajectory segments in the corresponding grid, , For the , standard deviation of the magnitude of the direction vector within the grid;

[0018] The trajectory validity screening submodule selects the trajectory segments whose discrete rate difference is higher than the stability threshold and whose direction deviation angle is greater than the direction change reference angle in the continuous trajectory segments according to the discrete difference value of the boundary direction, and counts the number of continuous trajectory segments marked as unstable in the same trajectory. When the cumulative number exceeds three segments, the trajectory is determined to be an invalid trajectory, the trajectory is eliminated and the trajectory set is updated to generate a reliable trajectory dataset.

[0019] The present invention is improved in that the trajectory semantic clustering module comprises:

[0020] The trajectory category screening submodule obtains the trusted trajectory data set, parses the trajectory source label field, classifies the samples according to the service attribute information included in the trajectory data, and screens the trajectory records of bus routes, taxi services and online car-hailing dispatch to obtain a transportation service trajectory set;

[0021] The spatial direction feature extraction submodule calls the traffic service trajectory set, collects the direction vectors and spatial coordinate positions of the continuous trajectory points in each trajectory segment, and projects the direction vector of each trajectory using the formula:

[0022] ;

[0023] The directional concentration value of each trajectory is obtained by operation, and the directional distribution structure is established for each trajectory classification according to the trajectory directional concentration, so as to obtain the trajectory directional concentration index set;

[0024] in, represents the directional concentration of the trajectory, Represents the total number of direction vectors contained in the trajectory, Indicates the trajectory The angle of the direction vector, represents the average value of all direction vector angles in the trajectory, Indicates the trajectory The spatial distance between the direction vector and the center point of the trajectory;

[0025] The clustering structure generation submodule identifies the spatial distribution similarity and directional aggregation features between trajectories according to the trajectory direction concentration index set, performs clustering according to the relative distance of the trajectories in the directional feature space, establishes semantic aggregation relationships between trajectories, and generates a semantic trajectory clustering set.

[0026] The present invention is improved in that the trajectory evolution stage identification module comprises:

[0027] The coverage extraction submodule calls the semantic trajectory cluster set, extracts the boundary position coordinates of the trajectory cluster center point in each time period, combines the coordinate distribution interval of the trajectory samples in the time period, calculates the proportion of trajectory samples falling into the closed polygonal area of ​​the cluster center, and generates the trajectory center coverage value;

[0028] The change amplitude calculation submodule obtains the coverage value in two adjacent time periods according to the trajectory center coverage value, collects the corresponding trajectory number and cluster center moving distance, and uses the formula:

[0029] ;

[0030] Calculate and obtain the trajectory coverage change index;

[0031] in, represents the trajectory coverage change index, represents the trajectory coverage value of the cluster center in the first time period, represents the trajectory coverage value of the cluster center in the second time period, represents the number of trajectory samples in the first time period, represents the number of trajectory samples in the second time period, Represents the spatial moving distance between cluster centers in two time periods.

[0032] The evolution type judgment submodule compares the trajectory coverage change index with the set trajectory evolution difference threshold to determine the change direction and intensity corresponding to the coverage change. If the change index is lower than the upper limit of the threshold interval, it is classified as steady evolution. If it is higher than the upper limit and the coverage increases, it is classified as centripetal contraction. If it is higher than the upper limit and the coverage decreases, it is classified as trajectory diffusion, and the trajectory evolution stage type is established.

[0033] The present invention is improved in that the dynamic parking determination module comprises:

[0034] The trajectory segment extraction submodule selects the region numbers in the contraction and stability stages according to the trajectory evolution stage type, obtains all trajectories in the region and divides them into continuous trajectory segments according to time periods, establishes the relationship between trajectory numbers and segment indexes, and generates a trajectory time segment set;

[0035] The stay degree calculation submodule calls the trajectory time segment set, extracts the trajectory point position set of the trajectory segment in unit time, calculates the number of spatial overlapping points and trajectory duration with the adjacent segment trajectory point set, using the formula:

[0036] ;

[0037] The spatial residence index of the trajectory segment in the local area is obtained by calculation, and the spatial residence evaluation of each segment is established to obtain the trajectory segment residence value;

[0038] in, Indicates the trajectory segment stay value, Indicates the number of overlapping trajectory points between a trajectory segment and its neighboring segments. Indicates the duration of the trajectory segment staying in the area. represents the average moving distance of the trajectory segments, represents the distribution area of ​​the trajectory points of the trajectory segment, represents the distribution area of ​​trajectory points of adjacent trajectory segments;

[0039] The parking state determination submodule compares the trajectory segment stay value with the parking state threshold value, selects the trajectory segments that meet high overlap and low movement in the continuous trajectory segments, classifies them into a dense parking set, marks the area number to which they belong, and establishes a dynamic parking area set.

[0040] The present invention is improved in that the system further comprises:

[0041] The urban traffic mode generation module calls the dynamic parking area set, identifies the dense parking area, calls the trajectory data in the area, extracts the spatiotemporal distribution characteristics, compares the parking characteristics and travel characteristics of the regional trajectory, divides the regional trajectory into traffic modes according to the similarity of the characteristics, defines the division results as commuting type, commercial active type and transportation hub type, and generates urban traffic mode identification information;

[0042] The urban traffic mode identification information includes traffic type labels, regional traffic activity levels and traffic behavior attribute indicators.

[0043] The present invention is improved in that the urban traffic mode generation module comprises:

[0044] The regional feature extraction submodule calls the dynamic stop area set, extracts all trajectory point data in the area, calculates the trajectory density value of each time period according to the time label of the trajectory point, and counts the distribution interval of the trajectory point on the spatial coordinate axis according to the spatial label to obtain the regional trajectory spatiotemporal feature value;

[0045] The behavior feature matching submodule extracts the residence time, activity start and end time, and movement range of the trajectory segment according to the spatiotemporal feature value of the regional trajectory, and extracts the activity radius and path overlap in the trajectory segment using the formula:

[0046] ;

[0047] The pattern matching degree of the regional trajectory on the behavior index is obtained by calculation, and a mapping relationship is established based on the matching value and the similarity scoring rule of each type of behavior template to generate a traffic behavior similarity index;

[0048] in, represents the traffic behavior similarity index, Indicates the number of high-frequency occurrence times during the dwell period in the trajectory segment, Indicates the overlap ratio of the path points of the trajectory segments in the area. represents the spatial discrete value of the trajectory activity path, represents the trajectory activity radius, represents the average activity radius of all trajectories in the area;

[0049] Compared with the prior art, the advantages and positive effects of the present invention are:

[0050] In the present invention, by performing boundary discrete rate difference analysis on traffic trajectory data and identifying continuous cross-grid anomalies, the unstable segments in the trajectory are effectively excluded, the accuracy of subsequent data clustering and analysis is ensured, and the trajectories are semantically clustered based on the concentration degree of spatial direction distribution, the accuracy and rationality of trajectory type discrimination are improved, and the spatial evolution trend of the trajectory is quantified by calculating the trajectory coverage difference of the cluster center in adjacent time periods, so as to realize the dynamic characterization of the traffic mode in the time dimension, and the dynamic parking state is accurately identified by combining the dual indicators of spatial overlap and parking duration between trajectory segments, avoiding the risk of misjudgment caused by a single time or space indicator, extracting the spatiotemporal characteristics of trajectory data in dense parking areas, and realizing efficient identification of travel function types in urban areas through similarity matching between trajectory characteristics and travel characteristics, improving the credibility and accuracy of trajectory data processing, enhancing the generalization and adaptability of traffic mode recognition in the interactive scenarios of multiple travel modes, promoting the fine characterization and function identification of urban traffic structure, and providing high-resolution data support for the optimal allocation of traffic resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a system flow chart of the present invention;

[0052] Figure 2 It is a flow chart of the trajectory credibility elimination module of the present invention;

[0053] Figure 3 It is a flow chart of the trajectory semantic clustering module of the present invention;

[0054] Figure 4 It is a flow chart of the trajectory evolution stage identification module of the present invention;

[0055] Figure 5 It is a flow chart of the dynamic parking determination module of the present invention;

[0056] Figure 6 This is a flow chart of the urban traffic mode generation module of the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0059] See also Figure 1 The present invention provides a technical solution: an urban traffic mode recognition system based on trajectory clustering, the system comprising:

[0060] The trajectory credibility elimination module obtains the location points of the traffic trajectory data and divides them into grid units, calculates the boundary discrete rate difference of the trajectory data, determines whether the continuous cross-grid difference exceeds the stability threshold, marks and eliminates abnormal trajectories with more than three unstable boundaries, and generates a credible trajectory data set;

[0061] The trajectory semantic clustering module obtains trajectory direction and spatial distribution characteristics based on the trusted trajectory dataset, screens bus route trajectory, taxi operation trajectory and online car-hailing trajectory data, calculates the spatial direction distribution concentration of trajectory data, and semantically clusters trajectory data according to the spatial concentration level to generate a semantic trajectory clustering set;

[0062] The trajectory evolution stage identification module calls the semantic trajectory clustering set, obtains the trajectory coverage rate of the cluster center, calculates the difference in trajectory coverage rates between adjacent time periods, determines whether the coverage rate difference exceeds the trajectory evolution difference threshold, divides the trajectory coverage rate change into three stages: stable, diffuse, and contractive, and generates the trajectory evolution stage type;

[0063] The dynamic stop discrimination module obtains the trajectory segments in the area according to the trajectory evolution stage type, calculates the spatial overlap and stop duration between adjacent trajectory segments, determines whether the spatial overlap and stop duration of the trajectory segments exceed the stop state threshold at the same time, marks the trajectory segments that meet the conditions as the dynamic stop state, and generates a dynamic stop area set;

[0064] The urban traffic mode generation module calls the dynamic parking area set, identifies the dense parking area, calls the trajectory data in the area, extracts the spatiotemporal distribution characteristics, compares the parking characteristics and travel characteristics of the regional trajectory, divides the regional trajectory into traffic modes according to the similarity of characteristics, and defines the division results as commuting type, commercial active type and transportation hub type, generating urban traffic mode identification information;

[0065] The trusted trajectory dataset includes trajectory number identification, excluded trajectory index and valid trajectory range. The semantic trajectory clustering set includes trajectory semantic label, trajectory cluster boundary configuration and trajectory concentration direction. The trajectory evolution stage type is specifically the steady evolution stage type, centripetal contraction stage type and trajectory diffusion stage type. The dynamic stop area set includes area number, stop segment index and spatial overlapping structure. The urban traffic mode recognition information includes traffic type label, regional traffic activity level and traffic behavior attribute indicator.

[0066] See also Figure 2 ,The trajectory credibility elimination module includes:

[0067] The grid division submodule obtains the location points of traffic trajectory data, collects the city boundary coordinate group and the trajectory point location coordinate set, divides the entire domain into equidistant grid units according to the city boundary coordinate group and the latitude and longitude scale ratio, calls the trajectory point location coordinate set to determine the grid number where it is located, establishes the mapping index relationship between the trajectory point and the grid unit, and generates the trajectory grid mapping structure;

[0068] The grid division submodule obtains the location points of traffic trajectory data. First, the original trajectory data set containing the vehicle or pedestrian positioning information is read. Each trajectory consists of multiple longitude and latitude coordinates marked with timestamps. The samples are sampled every 10 seconds to form a sequence point set. The sampling time period is set from 7:00 am to 9:00 pm. The trajectories with a data density greater than 5 points / minute are selected for subsequent processing. At the same time, the city boundary coordinate group is collected. The boundary information is extracted from the basic data of the city map. The longitude and latitude range is set to 121.30° to 121.80° east longitude and 31.00° to 31.40° north latitude. According to the boundary range, the longitude and latitude coordinates are divided into 50 sections in the vertical and horizontal directions, forming 2500 equidistant grid units. The longitude span of each grid is 0.01° and the latitude span is 0.008°. Then the longitude and latitude of each trajectory point are numbered and assigned. For example, the longitude of the trajectory point The latitude is 121.45° and the latitude is 31.16°, so the grid number is (15,20). The system maps the track point to the grid cell at row 15 and column 20, completing the index establishment between the track point and the grid cell. By continuously performing the mapping operation on the full track data, the corresponding grid number is updated point by point, and an index set from the track number to its corresponding multiple grid numbers is established. The data structure of the mapping structure is a one-to-many dictionary structure, in which the key is the track number and the value is an array of multiple grid numbers corresponding to the track, forming a unique mapping pair between the track number and the grid path. For example, the mapping path of track T001 is [(12,15), (13,15), (13,16), (14,16)]. In actual operation, a single track usually corresponds to 20 to 50 grid cells. The mapping structure is dynamically updated as the track grows, and finally a track grid mapping structure is generated.

[0069] The discrete rate difference calculation submodule extracts the trajectory segments located in the edge grid cells of each trajectory based on the trajectory grid mapping structure, collects the direction vector of each segment, and calculates the concentration degree of the direction amplitude vector of each segment in the corresponding grid, obtains the concentration value of the direction vector of the adjacent grid and constructs the direction difference term, using the formula:

[0070] ;

[0071] The boundary discrete rate difference between adjacent trajectory segments is obtained by operation, and it is determined whether the difference exceeds the stability threshold and the stable state of the trajectory segment is marked to obtain the discrete difference value in the boundary direction;

[0072] in, The representative trajectory is , The difference in discreteness between grids, , Represent the trajectories in , The grid , Direction angles, , are the number of trajectory segments in the corresponding grid, , For the , standard deviation of the magnitude of the direction vector within the grid;

[0073] The discrete rate difference calculation submodule extracts the trajectory segments located in the edge grid units of each trajectory based on the trajectory grid mapping structure. The judgment criteria for edge grid units are as follows: in the trajectory corresponding grid sequence, if there are grids in the four neighbors of a grid that are not passed by the trajectory, then the grid is marked as an edge grid. For example, the trajectory passes through (12,15), (13,15), (14,15), (15,15), where (12,15) and (15,15) are the starting and ending boundaries. The corresponding trajectory points are extracted to form segments, and then the direction vector of each segment is collected. The direction vector is calculated by the coordinate difference between adjacent trajectory points. For example, point A (121.4500, 31.1500) and point B (121.4510, 31.1515), then its direction angle is , all angles are in degrees, ranging from [0°, 180°]. The average value and standard deviation of all direction angles in each segment are calculated. The standard deviation is used as the discrete degree of the direction amplitude vector. Suppose the five direction angles in grid i are [45°, 50°, 55°, 52°, 53°], with an average value of 51.0° and a standard deviation S i =3.5 °, the direction angles in adjacent grids j are [70°, 68°, 72°, 66°, 69°], with a mean of 69.0° and a standard deviation of S j =2.2 ° , the parameters into the formula are:

[0074] ;

[0075] in, represents the difference in trajectory direction discreteness between grid i and grid j, , is the direction angle value of the trajectory in the corresponding grid, in degrees, , is the number of trajectory segments in the two grids, both 5 in this example, , are the standard deviations of the direction angles in grids i and j, respectively, representing the fluctuation intensity of the local motion direction. The standard deviation value is obtained by taking the square root of the mean difference of the direction angle set. This result shows that there is a large deviation between the trajectory motion directions of i and j.

[0076] The trajectory validity screening submodule selects the trajectory segments whose discrete rate difference in the continuous trajectory segments is higher than the stability threshold and whose direction deviation angle is greater than the direction change reference angle according to the discrete difference value of the boundary direction. The number of continuous trajectory segments marked as unstable in the same trajectory is counted. When the cumulative number exceeds three segments, the trajectory is judged as invalid, the trajectory is eliminated and the trajectory set is updated to generate a reliable trajectory dataset.

[0077] The trajectory validity screening submodule screens the continuous trajectory segments in the trajectory according to the discrete difference values ​​in the boundary direction, and retrieves the grids corresponding to the continuous trajectory segments in the grid path of each trajectory. If the value is higher than the set stability threshold, it is marked as an unstable segment. The stability threshold is set to 20. According to the distribution of trajectory direction continuity in the sample trajectory set, 90% of the stable segments are selected. The upper limit of the value distribution is 20, which is used as the judgment critical value, and it does not change with the change of urban areas to ensure the consistency of the judgment benchmark. At the same time, the system judges whether the direction deviation angle between adjacent trajectory segments is greater than the direction change reference angle, which is set to 30°. That is, if the difference between the mean direction angles corresponding to grids i and j is greater than 30°, the direction deviation angle meets the judgment condition. The setting basis is that the minimum significant deflection angle presented by the actual driving trajectory when changing lanes or turning at the intersection is about 28° to 35°, so 30° is selected as the unified judgment value. For example, if a certain trajectory continuous segment corresponds to The angles of the first three segments are 22.6, 25.1, 21.4, and 17.8, and the angular deviations are 32°, 36°, 29°, and 28°. The first three segments meet the unstable marking conditions. The system counts the number of consecutive unstable segments in the same trajectory. If the cumulative number exceeds 3, it is determined to be an invalid trajectory. The trajectory removal action is performed, the trajectory number is removed from the trajectory set, and the valid trajectory list is updated to obtain the final credible trajectory data set.

[0078] See also Figure 3 ,The trajectory semantic clustering module includes:

[0079] The trajectory category screening submodule obtains the trusted trajectory data set, parses the trajectory source label field, and classifies the samples according to the service attribute information included in the trajectory data. It screens the trajectory records from bus routes, taxi services, and online car-hailing dispatch, and obtains the transportation service trajectory set.

[0080] The trajectory category filtering submodule obtains the trusted trajectory data set and parses the source label field in each trajectory record. This field comes from the "service type" identifier preset in the original trajectory data collection phase. The identifier field is of string type and is divided into four types of labels: "public transportation", "rental", "online booking", and "private car". During the screening process, the label field corresponding to each trajectory record is first read and compared with the target value in the service trajectory category dictionary one by one. If the value is "public transportation" or "rental" or "online booking", the trajectory record is retained. The screening step is looped according to the trajectory number to filter out records with the source of "private car" or "unlabeled". Then, the service attribute field information contained in the trajectory itself is further called for the filtered trajectory set. The field structure includes auxiliary information such as the operating unit number, vehicle number, order number (for online booking) and the time period mark, which is used to further verify the legitimacy and integrity of its trajectory service attributes. For example, the trajectory ID is T_003. The service label is "taxi", the operating unit number is "TX001", the track point time is distributed between 08:10 and 08:45, the track length is 12.5 kilometers, and the data meets the logic of urban taxi operation. The system formally classifies the track into the "taxi service" track subset. In the classification screening, the track service attribute information must meet two conditions. The first is that the source label matches the three service values ​​of bus, taxi, and online booking. The second is that its vehicle number must have a matching item in the registered operation number database of the Transportation Bureau. The system verifies the index of the operation number to ensure that the track record is truly derived from a valid transportation service system. If the vehicle number is missing, abnormal, or not in the registration table, the track is excluded, and finally a transportation service track set is established. This set only retains the track data that passes the above two screening checks. The set data structure still uses the track number as the key, with track points, timestamp sequences, service labels and other contents as the value to form a structure, and finally a transportation service track set is obtained.

[0081] The spatial direction feature extraction submodule calls the traffic service trajectory set, collects the direction vectors and spatial coordinate positions of the continuous trajectory points in each trajectory segment, and projects the direction vector of each trajectory using the formula:

[0082] ;

[0083] The directional concentration value of each trajectory is obtained by operation, and the directional distribution structure is established for each trajectory classification according to the trajectory directional concentration, so as to obtain the trajectory directional concentration index set;

[0084] in, represents the directional concentration of the trajectory, Represents the total number of direction vectors contained in the trajectory, Indicates the trajectory The angle of the direction vector, represents the average value of all direction vector angles in the trajectory, Indicates the trajectory The spatial distance between the direction vector and the center point of the trajectory;

[0085] The spatial direction feature extraction submodule calls the traffic service trajectory set, collects the direction vectors and spatial coordinate positions of the continuous trajectory points in each trajectory segment, and performs direction angle calculation for all point pairs. It is defined as the angle between the vector formed by two adjacent points and the due east direction. The direction angle value range is [0°, 180°]. For example, if the trajectory point P1 is (121.4550, 31.2450) and P2 is (121.4600, 31.2475), then the direction angle is In trajectory T_045, there are F=8 direction vectors with angle values ​​of [63.4°, 64.0°, 62.1°, 63.8°, 63.5°, 62.9°, 64.2°, 63.7°], and the average direction angle is , calculate the cosine value between the direction angle and the average direction angle to get the cosine value array , , , , , , , , the spatial distance corresponding to each direction vector The calculation method is: extract the Euclidean distance between the midpoint of each direction vector and the center point of the trajectory, calculate in meters, set the center point to (121.4575, 31.2460), and after coordinate conversion calculation, get Value is , , , , , , , , each item is entered into the formula for calculation:

[0086]

[0087] After calculating each term separately:

[0088] ;

[0089] In the formula, is the direction concentration value, which measures the overall direction consistency of the trajectory. The value range is (0,1). The closer the value is to 1, the more concentrated the direction is. is the number of direction vectors, For the direction vector angle, is the average direction angle, Used to measure the degree of deviation from the direction. is the spatial distance between the direction vector and the center point, by adding Controls the influence weight of the direction away from the center point. In this example, the trajectory direction concentration is 0.2380, which is recorded in the direction concentration index set corresponding to the current trajectory classification.

[0090] The clustering structure generation submodule identifies the spatial distribution similarity and directional aggregation features between trajectories according to the trajectory direction concentration index set, clusters the trajectories according to their relative distances in the directional feature space, establishes semantic aggregation relationships between trajectories, and generates a semantic trajectory clustering set.

[0091] The clustering structure generation submodule identifies the differences in directional patterns between different trajectories based on the aforementioned directional concentration index set. The system first constructs a directional feature space matrix of all trajectories. Each row of the matrix corresponds to a trajectory, which is listed as its directional concentration. , average direction angle , standard deviation of direction angle The three-dimensional vector represents the trajectory directional feature. For example, the directional feature of trajectory T_045 is (0.2380, 63.45, 0.65). Then, the Euclidean distance calculation between trajectories is performed on all trajectory feature vectors. The directional space distance between any two trajectories is defined as the square root of the sum of the differences of the three feature terms. If the directional features of T_045 and T_121 are (0.2380, 63.45, 0.65) and (0.2420, 62.75, 0.70), their directional space distance is:

[0092] ;

[0093] All trajectories are paired to calculate the directional spatial distance, and clustering is performed according to the set distance threshold. The threshold value is set to 0.75. The setting basis is that in the historical samples of urban traffic trajectories, more than 90% of the trajectory pairs with directional semantic consistency have a directional feature space distance less than 0.75. Therefore, this value is selected as the upper limit of the distance for semantic trajectory clustering. If the distance between two trajectories is less than the threshold, they are included in the same cluster. Finally, all trajectories are combined according to the directional spatial proximity relationship to form a set of trajectories with consistent directional semantics. The system assigns a semantic number to each cluster and maps the original trajectory number to the cluster number to complete the generation of the semantic trajectory cluster set.

[0094] See also Figure 4 ,The trajectory evolution stage identification module includes:

[0095] The coverage extraction submodule calls the semantic trajectory clustering set, extracts the boundary position coordinates of the trajectory cluster center point in each time period, combines the coordinate distribution interval of the trajectory samples in the time period, calculates the proportion of trajectory samples falling into the closed polygonal area of ​​the cluster center, and generates the trajectory center coverage value;

[0096] The coverage extraction submodule calls the semantic trajectory clustering set. First, the time window interval is divided into hourly time segments. Each hour is used as an independent analysis cycle to extract the center point position coordinates of the trajectory clustering structure in each time period. The center point of the trajectory cluster is obtained by calculating the average longitude and latitude of the coordinates of the midpoints of all trajectories in the cluster. For example, a cluster contains trajectories T01~T08, and its center point coordinates are (121.4553,31.2411). Then, all trajectory points of all trajectories in the cluster in the current time period are called to obtain the minimum circumscribed rectangular boundary of the trajectory point set, calculate the area of ​​its boundary area, and divide the number of trajectory points contained in the boundary of the area by the total number of trajectory points of the cluster in the time period to obtain the trajectory point falling into the cluster center boundary. The proportion of the area. Assuming that the cluster contains 180 trajectory points in the time period of 8:00~9:00, of which 150 are inside the circumscribed rectangular area, the coverage rate in this time period is 150 / 180=0.833. The coverage rate value range is defined between [0,1]. The closer the value is to 1, the more concentrated the distribution of trajectory samples is, and the more representative the cluster center is. The boundary area of ​​the cluster center surrounds the boundary of the trajectory point distribution in a convex hull manner. Non-closed shapes or point sets distributed outside the boundary line are not counted in the number of internal trajectory points. The accuracy of the trajectory points used in the actual calculation must meet more than five decimal places (that is, the coordinate accuracy is within 1 meter) to ensure the accuracy of the spatial boundary calculation, and finally generate the trajectory center coverage value corresponding to each cluster in each time period.

[0097] The change amplitude calculation submodule obtains the coverage value in two adjacent time periods based on the trajectory center coverage value, collects the corresponding trajectory number and cluster center moving distance, and uses the formula:

[0098] ;

[0099] Calculate and obtain the trajectory coverage change index;

[0100] in, represents the trajectory coverage change index, represents the trajectory coverage value of the cluster center in the first time period, represents the trajectory coverage value of the cluster center in the second time period, represents the number of trajectory samples in the first time period, represents the number of trajectory samples in the second time period, Represents the spatial moving distance between cluster centers in two time periods.

[0101] The change amplitude calculation submodule extracts the coverage values ​​of the same cluster in two adjacent time periods in turn according to the above coverage values, and obtains the number of trajectory samples and the spatial moving distance of the cluster center. Assume that the coverage rate in the first time period is 0.833, the corresponding number of trajectories is 180, and the coverage rate in the second time period is 0.785, the corresponding number of trajectories is 200. The coordinates of the cluster center points in the two time periods are (121.4553, 31.2411) and (121.4571, 31.2405), respectively. The spatial distance is calculated by coordinate conversion: Meters (Note: longitude and latitude coordinates are converted at 111km per degree), and the parameters are substituted into the trajectory coverage change index calculation formula:

[0102] ;

[0103] In the formula, is the coverage change index, which indicates the weighted change intensity of the coverage change range under the influence of the number of trajectory samples and spatial movement in two time periods. and is the coverage value, and is the sample size, is the moving distance of the cluster center in meters. The moving distance is obtained by converting the latitude and longitude coordinates. The index value result is a numerical indicator of dimensional consistency and is not distorted by changes in the number of trajectories. The result value of 0.2773 indicates that the coverage rate has a deviation trend in the current time period.

[0104] The evolution type judgment submodule compares the trajectory coverage change index with the set trajectory evolution difference threshold to determine the change direction and intensity corresponding to the coverage change. If the change index is lower than the upper limit of the threshold interval, it is classified as stable evolution. If it is higher than the upper limit and the coverage increases, it is classified as centripetal contraction. If it is higher than the upper limit and the coverage decreases, it is classified as trajectory diffusion, and the trajectory evolution stage type is established.

[0105] The evolution type judgment submodule identifies the trajectory evolution pattern according to the aforementioned coverage change index. The system sets the trajectory evolution difference threshold to 0.15. The threshold is set based on the statistical analysis of the evolution of 10,000 traffic service trajectory samples in 10 cities in continuous hourly segments. The average value of the change amplitude index in 90% of the intervals is about 0.15. This value is selected as the reference point for dividing the stable and changing states. This value remains unchanged with the change of trajectory source or urban structure. The judgment step first compares the current coverage change index with 0.15. If , then the trajectory evolution is classified as “steady evolution”, if and , that is, the coverage rate increases, and the trajectory is judged to be "centripetal contraction". and , that is, the coverage rate decreases, and it is judged as "trace diffusion". In the calculation results, ,at the same time Therefore, the evolution type is judged to be trajectory diffusion. The system uses the cluster number and time period index as identifiers to generate a trajectory evolution stage type record. The record contains structured field information such as evolution type, time period start and end, cluster center coordinate change, and coverage change for subsequent trajectory behavior trend modeling.

[0106] See also Figure 5 , the dynamic parking discrimination module includes:

[0107] The trajectory segment extraction submodule selects the region numbers in the contraction and stability stages according to the trajectory evolution stage type, obtains all trajectories in the region and divides them into continuous trajectory segments by time period, establishes the relationship between trajectory number and segment index, and generates a trajectory time segment set;

[0108] The trajectory segment extraction submodule selects the area numbers in the "centripetal contraction" or "stable evolution" stage according to the trajectory evolution stage type. The system reads the trajectory evolution type record generated in the previous stage, extracts the entries whose evolution type field value is "contraction" or "stable", and obtains the corresponding area numbers, such as area numbers A101, A102, B034, etc., and then obtains the trajectory number set corresponding to these area numbers, and divides them into several segments according to the time series of each trajectory. The unit time of each trajectory segment is 15 minutes. Set the duration of trajectory T011 to 07:00 to 08:00, then divide it into 4 trajectory segments T011-1 to T011-4, each segment saves all trajectory points in this time period At the same time, a mapping relationship between the trajectory number and its fragment index is established. The mapping relationship is implemented using a two-dimensional array structure. Each primary key is the trajectory number, and the value is the corresponding time fragment number sequence. For example, the fragment corresponding to T011 is [T011-1, T011-2, T011-3, T011-4]. Each fragment object contains the trajectory point position coordinates, timestamp sequence, corresponding area number, and time period to facilitate subsequent spatial analysis operations. During the division process, the trajectory duration must be greater than the fragment unit time and the number of trajectory points must be no less than 3 points / minute. If the trajectory has less than 45 trajectory points in any fragment, the fragment will not be included in the subsequent analysis to ensure that the data density meets the requirements of subsequent overlap and stay analysis, and finally generate a trajectory time fragment set.

[0109] The stay calculation submodule calls the trajectory time segment set, extracts the trajectory point position set of the trajectory segment in unit time, and calculates the number of spatial overlap points and trajectory duration with the adjacent segment trajectory point set using the formula:

[0110] ;

[0111] The spatial residence index of the trajectory segment in the local area is obtained by calculation, and the spatial residence evaluation of each segment is established to obtain the trajectory segment residence value;

[0112] in, Indicates the trajectory segment stay value, Indicates the number of overlapping trajectory points between a trajectory segment and its neighboring segments. Indicates the length of time that a trajectory segment stays in the area. represents the average moving distance of the trajectory segments, represents the distribution area of ​​the trajectory points of the trajectory segment, represents the distribution area of ​​trajectory points of adjacent trajectory segments;

[0113] The stay calculation submodule calls the above trajectory time segment set, aggregates the segments by region number, and performs spatial overlap and stay calculations on a region-by-region basis. First, the trajectory point position set in each segment is extracted, and the trajectory point coordinates are discretized into integer grid points with a spatial accuracy of 2 meters. For example, the trajectory point coordinates in trajectory segment T011-2 are (121.45702, 31.24133), which are mapped to (121457, 312413) after coordinate conversion and discretization. Then, the adjacent segments in the same time period are cross-compared to determine the number of intersections between the discretized point sets of the two segments, which is recorded as , if the number of intersection trajectory points of T011-2 and T012-1 is 65, then , trajectory duration is the time span of the segment, 15 minutes is converted to 900 seconds, and the average moving distance of the trajectory segment By summing and averaging the Euclidean distances between trajectory points, the average distance between trajectory points in a certain segment is set to 4.2 meters. There are 60 segments in total, so the average moving distance is 252 meters. The distribution area of ​​trajectory points is Distribution area with adjacent fragments By estimating the area of ​​the minimum circumscribed polygon, set , , put it into the formula:

[0114] ;

[0115] In the formula, is the retention value of the trajectory fragment, which comprehensively evaluates whether the trajectory points have repeated appearance behavior in the same area. The higher the value, the more significant the spatial retention phenomenon of the fragment. The first term of the formula is the standardized measure of the spatial overlap of the trajectory points, and the second term is the normalized proportional difference of the point distribution area difference. The combination of the two represents the spatial consistency and stagnation characteristics between the fragment and the adjacent fragments. The retention value calculated in this example is 0.1635. The system records this value in the fragment attribute structure for status screening in the next stage.

[0116] The parking state determination submodule compares the stop value of the trajectory segment with the parking state threshold, selects the trajectory segments that meet high overlap and low movement in the continuous trajectory segments, classifies them into the dense parking set, marks the area number to which they belong, and establishes a dynamic parking area set.

[0117] The parking state determination submodule compares the dwell value of the aforementioned trajectory segment with the set parking state threshold. The system sets the parking state determination threshold to 0.15. This threshold is derived from the distribution statistics of the dwell value samples of the known parking behavior trajectory segments. 95% of the segments that are judged as "parked" with a dwell value greater than 0.15 are selected. Therefore, this value is used as the determination boundary for identifying dense parking segments. The determination operation is performed on all segments. If a segment , it is determined to meet the parking state, and the system aggregates and judges the continuous segments of the trajectory number. If there are 3 or more continuous segments under a certain trajectory number that all meet the parking state, then its spatial location area number will be recorded as a dense parking area. For example, segments T011-1 to T011-3 of trajectory T011 are all determined to be parking segments, and the corresponding area number is A102. The system adds A102 to the current parking area set. The set uses the area number as the primary key, and the value is the sequence of all parking trajectory numbers and segment numbers in the area, which is continuously updated to finally establish a dynamic parking area set.

[0118] See also Figure 6 ,The urban traffic pattern generation module includes:

[0119] The regional feature extraction submodule calls the dynamic stop area set to extract all trajectory point data in the area, calculates the trajectory density value of each period according to the time label of the trajectory point, and counts the distribution interval of the trajectory point on the spatial coordinate axis according to the spatial label to obtain the spatiotemporal feature value of the regional trajectory;

[0120] The regional feature extraction submodule calls the dynamic parking area set, reads the trajectory segment set corresponding to each area number in turn, extracts the latitude and longitude coordinates and timestamp information of all trajectory points in the set, divides all trajectory points into time periods according to the timestamp, counts the number of trajectory points in each time period and divides it by the area of ​​the area to obtain the trajectory density value. The density unit is "points / square kilometer". For example, the number of trajectory points in area number A204 from 08:00 to 09:00 is 420, and the area is 0.35 square kilometers. The density value of this period is 1200 points / km 2At the same time, the system counts the distribution range of all trajectory points in the area on the spatial coordinate axis (longitude and latitude), extracts the maximum and minimum longitude, maximum and minimum latitude, and determines the horizontal and vertical distribution spans of the trajectory points. For example, the longitude range of the trajectory points is [121.4520, 121.4605], the latitude range is [31.2401, 31.2446], the horizontal span is 0.0085 degrees, and the vertical span is 0.0045 degrees. Combined with the trend of trajectory point density changes and the spatial distribution boundary, the system constructs the trajectory spatiotemporal eigenvalue structure of the area, which contains the trajectory density of each time period, the maximum span of the trajectory on the spatial coordinate axis, the distribution range of the regional boundary and other fields as input parameters for subsequent behavior analysis.

[0121] The behavior feature matching submodule extracts the residence time, activity start and end time, and movement range of the trajectory segment according to the spatiotemporal feature value of the regional trajectory, and extracts the activity radius and path overlap in the trajectory segment using the formula:

[0122] ;

[0123] The pattern matching degree of the regional trajectory on the behavior index is obtained by calculation, and a mapping relationship is established based on the matching value and the similarity scoring rule of each type of behavior template to generate a traffic behavior similarity index;

[0124] in, represents the traffic behavior similarity index, Indicates the number of high-frequency occurrence times during the dwell period in the trajectory segment, Indicates the overlap ratio of the path points of the trajectory segments in the area. represents the spatial discrete value of the trajectory activity path, represents the trajectory activity radius, represents the average activity radius of all trajectories in the area;

[0125] The behavior feature matching submodule extracts the residence time, activity start and end time, and activity range of each trajectory segment based on the above trajectory spatiotemporal feature values. The residence time is defined as the length of time that the trajectory point exists continuously in the area, the activity start and end time is the time difference between the earliest and latest timestamps in the segment, the activity range is the length of the diagonal of the minimum circumscribed rectangle formed by the trajectory points, and the path overlap degree is defined as the path overlap degree. Indicates the ratio of the path of the trajectory points in the trajectory segment to the historical trajectory points in this area, and the activity radius Indicates the maximum radius from the trajectory point to the center point of the segment, and the average activity radius is the total number of all trajectory segments in the region Take the average value, and assume that the analysis result of trajectory segment Z302 is as follows: the frequency of occurrence in the dwell period is 5 times, recorded as , the path overlap ratio is 0.72, recorded as , the trajectory path dispersion is calculated by the standard deviation of the Euclidean distance from the path point to the center point, set as Meters, the trajectory activity radius is meters, and the average activity radius of the area is m, into the formula:

[0126] ;

[0127] In the formula, is a traffic behavior similarity index, which measures the closeness between the trajectory segment and the typical behavior template of the area. is the number of times the trajectory segment appears in the target area during high-frequency periods, which can be obtained by counting the concentrated distribution interval of the timestamp. is the overlap rate between the trajectory path and the historical path points of the region, which is calculated by dividing the number of intersections between the trajectory path points and the historical trajectory points of the region by the total number of trajectory path points. is the discrete degree of trajectory segment path, , are the average activity radius of individuals and regions, respectively. The result shows that trajectory Z302 is weak in spatial activity and behavioral repeatability, and its behavioral similarity value is 0.1074.

[0128] The mode label generation submodule determines the mode label corresponding to the interval of the similarity index based on the traffic behavior similarity index and the preset standard interval of traffic function type, and matches the commuting type, commercial active type and traffic hub type labels to establish urban traffic mode identification information.

[0129] The mode label generation submodule classifies the types according to the above traffic behavior similarity index. The system presets the similarity threshold intervals of three types of traffic function types: commuting type is 0.08 to 0.18, commercial active type is 0.18 to 0.30, and transportation hub type is above 0.30. This classification standard is obtained by large-scale trajectory clustering statistics. Based on the high-frequency trajectory areas of the city center, subway station periphery and business district, the sample distribution is extracted, and the similarity index means of their typical trajectory segments are calculated respectively. The relative intervals of the three types of behavior patterns are obtained, and the alignment with the traffic facilities is verified by manual verification. The classification boundaries are repeatable and representative. For the trajectory segment Z302, its behavior similarity value is 0.1074, which falls into the commuting type index interval. Therefore, the system determines it as a commuting type mode. The system outputs the judgment result in a structured form. The structure includes the trajectory number, the area number, the matching behavior label, the matching interval number, and the original behavior similarity value, and finally forms an urban traffic mode recognition information set, which is used as an input basis for the subsequent layout of traffic facilities or the formulation of traffic flow adjustment strategies.

[0130] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An urban traffic pattern recognition system based on trajectory clustering, characterized in that: The system comprises: The trajectory credibility elimination module obtains the location points of the traffic trajectory data and divides them into grid units, calculates the boundary discrete rate difference of the trajectory data, determines whether the continuous cross-grid difference exceeds the stability threshold, marks and eliminates abnormal trajectories with more than three unstable boundaries, and generates a credible trajectory data set; The trajectory semantic clustering module obtains trajectory direction and spatial distribution characteristics based on the trusted trajectory data set, calculates the concentration degree of spatial direction distribution, semantically clusters the trajectory data according to the degree of spatial concentration, and generates a semantic trajectory clustering set; The trajectory evolution stage identification module calls the semantic trajectory clustering set, obtains the trajectory coverage rate of the cluster center, calculates the difference of trajectory coverage rates in adjacent time periods, divides the trajectory coverage rate changes into three stages: stable, diffuse, and contractive, and generates the trajectory evolution stage type; The dynamic stop distinguishing module obtains the trajectory segments in the area according to the trajectory evolution stage type, calculates the spatial overlap and stop duration between adjacent trajectory segments, marks the trajectory segments that meet the conditions as the dynamic stop state, and generates a dynamic stop area set.

2. The urban traffic pattern recognition system based on trajectory clustering according to claim 1 is characterized in that: The trusted trajectory dataset includes a trajectory number identifier, a discarded trajectory index, and a valid trajectory range; the semantic trajectory clustering set includes a trajectory semantic label, a trajectory cluster boundary configuration, and a trajectory concentration direction; the trajectory evolution stage type is specifically a steady evolution stage type, a centripetal contraction stage type, and a trajectory diffusion stage type; the dynamic parking area set includes an area number, a parking segment index, and a spatial overlapping structure.

3. The urban traffic pattern recognition system based on trajectory clustering according to claim 1 is characterized in that: The trajectory credibility elimination module includes: The grid division submodule obtains the location points of traffic trajectory data, collects the city boundary coordinate group and the trajectory point location coordinate set, divides the entire domain into equidistant grid units according to the city boundary coordinate group and the latitude and longitude scale ratio, calls the trajectory point location coordinate set to determine the grid number where it is located, establishes the mapping index relationship between the trajectory point and the grid unit, and generates the trajectory grid mapping structure; The discrete rate difference calculation submodule extracts the trajectory segments located in the edge grid cells of each trajectory based on the trajectory grid mapping structure, collects the direction vector of each segment, calculates the concentration degree of the direction amplitude vector of each segment in the corresponding grid, obtains the concentration value of the direction vector of the adjacent grid and constructs the direction difference term, using the formula: ； The boundary discrete rate difference between adjacent trajectory segments is obtained by operation, and it is determined whether the difference exceeds the stability threshold and the stable state of the trajectory segment is marked to obtain the discrete difference value in the boundary direction; in, The representative trajectory is , The difference in discreteness between grids, , Represent the trajectories in , The grid , Direction angles, , are the number of trajectory segments in the corresponding grid, , For the , standard deviation of the magnitude of the direction vector within the grid; The trajectory validity screening submodule selects the trajectory segments whose discrete rate difference is higher than the stability threshold and whose direction deviation angle is greater than the direction change reference angle in the continuous trajectory segments according to the discrete difference value of the boundary direction, and counts the number of continuous trajectory segments marked as unstable in the same trajectory. When the cumulative number exceeds three segments, the trajectory is determined to be an invalid trajectory, the trajectory is eliminated and the trajectory set is updated to generate a reliable trajectory dataset.

4. The urban traffic pattern recognition system based on trajectory clustering according to claim 1 is characterized in that: The trajectory semantic clustering module includes: The trajectory category screening submodule obtains the trusted trajectory data set, parses the trajectory source label field, classifies the samples according to the service attribute information included in the trajectory data, and screens the trajectory records of bus routes, taxi services and online car-hailing dispatch to obtain a transportation service trajectory set; The spatial direction feature extraction submodule calls the traffic service trajectory set, collects the direction vectors and spatial coordinate positions of the continuous trajectory points in each trajectory segment, and projects the direction vector of each trajectory using the formula: ； The directional concentration value of each trajectory is obtained by operation, and the directional distribution structure is established for each trajectory classification according to the trajectory directional concentration, so as to obtain the trajectory directional concentration index set; in, represents the directional concentration of the trajectory, Represents the total number of direction vectors contained in the trajectory, Indicates the trajectory The angle of the direction vector, represents the average value of all direction vector angles in the trajectory, Indicates the trajectory The spatial distance between the direction vector and the center point of the trajectory; The clustering structure generation submodule identifies the spatial distribution similarity and directional aggregation features between trajectories according to the trajectory direction concentration index set, performs clustering according to the relative distance of the trajectories in the directional feature space, establishes semantic aggregation relationships between trajectories, and generates a semantic trajectory clustering set.

5. The urban traffic pattern recognition system based on trajectory clustering according to claim 1 is characterized in that: The trajectory evolution stage identification module includes: The coverage extraction submodule calls the semantic trajectory cluster set, extracts the boundary position coordinates of the trajectory cluster center point in each time period, combines the coordinate distribution interval of the trajectory samples in the time period, calculates the proportion of trajectory samples falling into the closed polygonal area of ​​the cluster center, and generates the trajectory center coverage value; The change amplitude calculation submodule obtains the coverage value in two adjacent time periods according to the trajectory center coverage value, collects the corresponding trajectory number and cluster center moving distance, and uses the formula: ； Calculate and obtain the trajectory coverage change index; in, represents the trajectory coverage change index, represents the trajectory coverage value of the cluster center in the first time period, represents the trajectory coverage value of the cluster center in the second time period, represents the number of trajectory samples in the first time period, represents the number of trajectory samples in the second time period, Represents the spatial moving distance between cluster centers in two time periods; The evolution type judgment submodule compares the trajectory coverage change index with the set trajectory evolution difference threshold to determine the change direction and intensity corresponding to the coverage change. If the change index is lower than the upper limit of the threshold interval, it is classified as steady evolution. If it is higher than the upper limit and the coverage increases, it is classified as centripetal contraction. If it is higher than the upper limit and the coverage decreases, it is classified as trajectory diffusion, and the trajectory evolution stage type is established.

6. The urban traffic pattern recognition system based on trajectory clustering according to claim 1 is characterized in that: The dynamic parking determination module comprises: The trajectory segment extraction submodule selects the region numbers in the contraction and stability stages according to the trajectory evolution stage type, obtains all trajectories in the region and divides them into continuous trajectory segments according to time periods, establishes the relationship between trajectory numbers and segment indexes, and generates a trajectory time segment set; The stay degree calculation submodule calls the trajectory time segment set, extracts the trajectory point position set of the trajectory segment in unit time, calculates the number of spatial overlapping points and trajectory duration with the adjacent segment trajectory point set, using the formula: ； The spatial residence index of the trajectory segment in the local area is obtained by calculation, and the spatial residence evaluation of each segment is established to obtain the trajectory segment residence value; in, Indicates the trajectory segment stay value, Indicates the number of overlapping trajectory points between a trajectory segment and its neighboring segments. Indicates the duration of the trajectory segment staying in the area. represents the average moving distance of the trajectory segments, represents the distribution area of ​​the trajectory points of the trajectory segment, represents the distribution area of ​​trajectory points of adjacent trajectory segments; The parking state determination submodule compares the trajectory segment stay value with the parking state threshold value, selects the trajectory segments that meet high overlap and low movement in the continuous trajectory segments, classifies them into a dense parking set, marks the area number to which they belong, and establishes a dynamic parking area set.

7. The urban traffic pattern recognition system based on trajectory clustering according to claim 1 is characterized in that: The system further comprises: The urban traffic mode generation module calls the dynamic parking area set, identifies the dense parking area, calls the trajectory data in the area, extracts the spatiotemporal distribution characteristics, compares the parking characteristics and travel characteristics of the regional trajectory, divides the regional trajectory into traffic modes according to the similarity of the characteristics, defines the division results as commuting type, commercial active type and transportation hub type, and generates urban traffic mode identification information; The urban traffic mode identification information includes traffic type labels, regional traffic activity levels and traffic behavior attribute indicators.

8. The urban traffic pattern recognition system based on trajectory clustering according to claim 7 is characterized in that: The urban traffic mode generation module comprises: The regional feature extraction submodule calls the dynamic stop area set, extracts all trajectory point data in the area, calculates the trajectory density value of each time period according to the time label of the trajectory point, and counts the distribution interval of the trajectory point on the spatial coordinate axis according to the spatial label to obtain the regional trajectory spatiotemporal feature value; The behavior feature matching submodule extracts the residence time, activity start and end time, and movement range of the trajectory segment according to the spatiotemporal feature value of the regional trajectory, and extracts the activity radius and path overlap in the trajectory segment using the formula: ； The pattern matching degree of the regional trajectory on the behavior index is obtained by calculation, and a mapping relationship is established based on the matching value and the similarity scoring rule of each type of behavior template to generate a traffic behavior similarity index; in, represents the traffic behavior similarity index, Indicates the number of high-frequency occurrence times during the dwell period in the trajectory segment, Indicates the overlap ratio of the path points of the trajectory segments in the area. represents the spatial discrete value of the trajectory activity path, represents the trajectory activity radius, represents the average activity radius of all trajectories in the area; The mode label generation submodule determines the mode label corresponding to the interval where the similarity index is located based on the traffic behavior similarity index and the preset standard interval of traffic function type, and matches the commuting type, commercial active type and traffic hub type labels to establish urban traffic mode identification information.

Images