Intelligent distribution method and system for open block traffic flow

Abstract

The application provides an open block traffic flow intelligent distribution method and system, and relates to the technical field of data processing.The method comprises the following steps: processing a basic triangular unit to define the interaction range of multi-mode traffic flow; according to the interaction range, mapping point cloud data on a virtual analysis plane to a corresponding actual road network area to extract multi-mode traffic flow interference parameters in a branch network, and generating a Gaussian mixture model initial parameter set; based on the Gaussian mixture model initial parameter set, iteratively optimizing and distributing open block OD pairs, dynamically adjusting path selection probability, obtaining a final traffic distribution scheme, and outputting real-time traffic distribution proportions of each path.The application can extract key features from massive trajectory data, accurately locate traffic conflict areas in space, and quantify interference intensity.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an open block traffic flow intelligent allocation method and system. BACKGROUND

[0002] With the continuous development of urban road network and the popularization of open block mode, the traffic flow composition carried by the branch network in the block is becoming increasingly complex. Various traffic modes such as motor vehicles, non-motor vehicles and pedestrians share, interweave and compete in the limited road space of the block. Especially in the morning peak and other specific periods, the demand for short-distance travel such as commuting and going to school is highly concentrated, which leads to the intensification of local road traffic conflicts and the decline of traffic efficiency. How to accurately perceive, analyze and intelligently optimize the allocation of multi-mode traffic flow in the open block has become a key technical problem to improve the vitality of the block and ensure traffic safety and smoothness.

[0003] Traditional traffic flow allocation methods are mostly oriented to urban trunk roads or expressway networks. Their models are usually based on the assumption of uniform and continuous traffic flow, and focus on a single mode of motor vehicles. When applied to the open block scene, these methods have some limitations:

[0004] Firstly, the traditional method relies on fixed-position detectors (such as coils and cameras) to obtain cross-section flow or average speed, which is difficult to comprehensively and finely capture the continuous trajectories of dynamic targets such as pedestrians and non-motor vehicles and their micro-interaction behaviors with motor vehicles, and there are data blind spots and mode fragmentation problems. Secondly, existing methods lack effective quantitative means for the complex mutual interference caused by mixed traffic and interlaced paths in the branch network. Usually, a simple speed-flow relationship or an empirical reduction coefficient is used to estimate the impact, which cannot reflect the non-uniform distribution of interference in space and the dynamic evolution characteristics over time. Finally, the path selection probability and traffic allocation scheme generated based on the above simplified analysis model often deviates greatly from the actual traffic conditions in the block, making it difficult to support fine control requirements such as inducing non-motor vehicles and pedestrians to use specific paths and balancing the load of each road network. SUMMARY

[0005] The technical problem to be solved by the present application is to provide an open block traffic flow intelligent allocation method and system, which can extract key features from massive trajectory data, accurately locate traffic conflict areas in space, and quantify interference intensity, thereby providing dynamic decision-making basis for traffic management and planning in open blocks.

[0006] To solve the above technical problems, the technical solutions of the present application are as follows:

[0007] In a first aspect, an open block traffic flow intelligent allocation method is provided, the method comprising:

[0008] The multi-source heterogeneous traffic data of the open street block is collected, the position and trajectory information of the traffic entity contained in the traffic data are converted into discrete spatial point cloud data, a virtual analysis plane is constructed based on the spatial distribution of the point cloud data and the road topological structure of the open street block, the point cloud data is processed in time and space alignment and fusion, and the correlation mapping relationship of the motor vehicle and non-motor vehicle flow data is established on the virtual analysis plane to generate a street full-mode traffic flow fusion data set;

[0009] Based on the street full-mode traffic flow fusion data set, the travel time distribution characteristics and travel mode composition characteristics of the short distance OD pair are analyzed on the virtual analysis plane, the walking and non-motor vehicle dominant characteristics of the school commuting passenger flow in the morning peak period are identified, the key analysis area is determined for the branch network on the virtual analysis plane according to the dominant characteristics, and the basic triangular unit is determined in the key analysis area; the basic triangular unit is processed to define the interaction range of the multi-mode traffic flow; the point cloud data on the virtual analysis plane is mapped to the corresponding actual road network area according to the interaction range, and the mutual interference parameters of the multi-mode traffic flow in the branch network are extracted to generate a Gaussian mixture model initial parameter set;

[0010] Based on the Gaussian mixture model initial parameter set, the OD pair of the open street block is iteratively optimized and distributed, the path selection probability is dynamically adjusted, the final flow distribution scheme is obtained, and the real-time flow distribution proportion of each path is output.

[0011] In a second aspect, an open street block traffic flow intelligent distribution system includes:

[0012] The acquisition module is configured to collect multi-source heterogeneous traffic data of an open street block, convert position and trajectory information of a traffic entity contained in the traffic data into discrete spatial point cloud data, construct a virtual analysis plane based on spatial distribution of the point cloud data and road topological structure of the open street block, process the point cloud data in time and space alignment and fusion, and establish a correlation mapping relationship of motor vehicle and non-motor vehicle flow data on the virtual analysis plane to generate a street full-mode traffic flow fusion data set.

[0013] The processing module is configured to analyze travel time distribution characteristics and travel mode composition characteristics of a short distance OD pair on a virtual analysis plane based on a street full-mode traffic flow fusion data set, identify walking and non-motor vehicle dominant characteristics of school commuting passenger flow in a morning peak period, determine a key analysis area for a branch network on the virtual analysis plane according to the dominant characteristics, and determine a basic triangular unit in the key analysis area; the basic triangular unit is processed to define the interaction range of the multi-mode traffic flow; the point cloud data on the virtual analysis plane is mapped to the corresponding actual road network area according to the interaction range, and the mutual interference parameters of the multi-mode traffic flow in the branch network are extracted to generate a Gaussian mixture model initial parameter set.

[0014] The distribution module is configured to perform iterative optimization distribution on the open block OD pairs based on the initial parameter set of the Gaussian mixture model, dynamically adjust path selection probability, obtain a final traffic distribution scheme, and output real-time traffic distribution proportions of each path.

[0015] The above scheme of the present application at least includes the following beneficial effects:

[0016] By uniformly converting multi-source heterogeneous data such as video, radar, mobile signaling into spatial point clouds with space-time attributes, and aligning, mapping and fusing on a unified virtual analysis plane, the present application constructs a full-mode traffic flow fusion data set containing complete trajectories and interaction relationships of motor vehicles, non-motor vehicles and other modes, overcomes the disadvantages of single data source, fragmented modes and missing trajectory information of traditional methods, and provides a high-fidelity data basis for fine analysis.

[0017] Based on the full-mode fusion data, the present application can intelligently identify specific traffic phenomena such as early morning peak school commuting passenger flow which has a certain mode composition (dominated by walking and non-motor vehicles) and space-time regularity by analyzing the travel characteristics of short-distance OD. Further, by setting a basic triangular unit, the present application can automatically and accurately locate the key areas (interaction range) where multi-mode traffic flows interfere with each other most concentratedly from the complex road network, realizing scientific focusing from macro characteristics to micro hot areas.

[0018] The description of traffic interference by traditional methods is often the whole road section or empirical. The present application maps the point cloud data on the virtual plane to the actual road sub-unit after triangulation, realizes the discretization of traffic state parameters (such as speed, density) in space, and area normalization calculation. The parameters such as speed interference intensity and density interweaving coefficient extracted therefrom can finely depict the distribution difference of interference at different positions inside the road.

[0019] The present application extracts the mutual interference parameters in a data-driven manner as the initial parameter set of the Gaussian mixture model, so that the model contains the actual traffic conflict mode from the initialization stage; in the iterative optimization process, the model parameters are continuously updated according to the real-time traffic state, and the path selection probability is dynamically adjusted accordingly, and the final traffic distribution scheme obtained is not only in line with the theoretical principles of random user equilibrium, but also closely matches the complex reality of multi-mode mixed traffic in the block, and the distribution result is more reliable. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of an open block traffic flow intelligent distribution method provided by an embodiment of the present application.

[0021] Figure 2 is a schematic diagram of an open block traffic flow intelligent distribution system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0022] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms without being limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood, and so that the scope of the present disclosure can be completely conveyed to those skilled in the art.

[0023] As Figure 1 shown, the embodiments of the present application propose an open street traffic flow intelligent allocation method, which comprises the following steps:

[0024] Step 100, collect multi-source heterogeneous traffic data of the open street, convert the traffic entity position and trajectory information contained in the traffic data into discrete spatial point cloud data; based on the spatial distribution of the point cloud data and the road topological structure of the open street, construct a virtual analysis plane; perform spatio-temporal alignment and fusion processing on the point cloud data, and establish an associated mapping relationship of motor vehicle and non-motor vehicle flow data on the virtual analysis plane to generate a street full-mode traffic flow fusion data set;

[0025] Step 200, based on the street full-mode traffic flow fusion data set, analyze the travel time distribution characteristics and travel mode composition characteristics of the short-distance OD pair on the virtual analysis plane, and identify the walking and non-motor vehicle dominant characteristics of the school commuting passenger flow in the morning peak period; according to the dominant characteristics, determine the key analysis area for the branch network on the virtual analysis plane, and determine the basic triangular unit in the key analysis area; process the basic triangular unit to define the interaction range of multi-mode traffic flow; according to the interaction range, map the point cloud data on the virtual analysis plane to the corresponding actual road network area to extract the mutual interference parameters of multi-mode traffic flow in the branch network, and generate a Gaussian mixture model initial parameter set;

[0026] Step 300, based on the Gaussian mixture model initial parameter set, iteratively optimize and allocate the open street OD pair, dynamically adjust the path selection probability, obtain the final traffic allocation scheme, and output the real-time traffic allocation proportion of each path.

[0027] In this embodiment, the application overcomes the limitations of single data source (only focusing on motor vehicle data) and multi-mode traffic flow separation statistics in traditional traffic flow analysis by collecting multi-source heterogeneous traffic data and converting it into discrete spatial point cloud data, constructing a virtual analysis plane and establishing an associated mapping relationship between motor vehicle and non-motor vehicle flow. On the one hand, the combination of point cloud data and road topological structure ensures the spatial accuracy of traffic entity location and trajectory information. On the other hand, spatio-temporal alignment and fusion processing realizes the organic linkage of motor vehicle, non-motor vehicle and pedestrian flow data, providing full-mode, full-dimensional and highly correlated data set support for subsequent flow allocation, avoiding analysis bias caused by single data dimension and ensuring the basic reliability of the allocation scheme.

[0028] The application is aimed at the core characteristics of high proportion of short-distance OD pairs in open blocks, and early morning peak school commuting passenger flow dominated by walking / non-motor vehicles. By identifying such dominant characteristics and delineating key analysis areas, the application combines basic triangular elements and convex hull algorithm to define the interaction range of multi-mode traffic flow, solving the problem of traditional allocation methods ignoring the dominance of slow traffic in open blocks and the fuzzy positioning of key congestion areas. Specifically, by mapping virtual plane point cloud data to actual road network through triangular area algorithm, the application realizes the quantitative extraction of multi-mode traffic flow interference parameters in branch network, enabling the initial parameter set of Gaussian mixture model to accurately match the actual scene of dense roads in open blocks and strong mixed interference, avoiding the insufficient adaptation of general allocation models to special scenes.

[0029] The application performs OD pair iterative optimization allocation based on the initial parameter set of Gaussian mixture model, dynamically adjusts the path selection probability, and breaks through the rigid defect of traditional static traffic allocation scheme that is difficult to adapt to real-time traffic flow changes. Through the iterative optimization process, the traffic allocation scheme can respond to the tidal changes of early morning peak school commuting passenger flow and the dynamic intensity of multi-mode traffic flow interference in real time, reasonably allocate motor vehicles and slow traffic flow to different paths in the branch network, effectively reduce the mixed congestion of motor vehicles and non-motor vehicles, and improve the traffic efficiency in key areas. Especially for the walking / non-motor vehicle dominant characteristics of early morning peak school commuting passenger flow, the allocation scheme can prioritize the traffic space and safety of slow traffic, while avoiding excessive concentration of motor vehicle flow on a certain branch, achieving the goal of collaborative and efficient allocation of motor vehicles and slow traffic.

[0030] The present application converts the open street traffic flow distribution from experience-based decision-making to data-driven precise management through the intelligent design of the whole process of data fusion-feature recognition-model optimization-dynamic allocation. On the one hand, the fusion of full-way traffic flow data sets and the extraction of mutual interference parameters provide quantitative data support for open street road planning and facility optimization (such as slow system upgrade and intersection channelization); on the other hand, the dynamic traffic distribution scheme can be directly applied to traffic signal timing optimization, real-time traffic guidance and other practical management scenarios, improving the response speed and accuracy of traffic management. Ultimately, by relieving congestion, improving traffic efficiency and protecting the rights and interests of slow traffic, the travel experience of residents (especially school-going crowds) is improved, and the coordinated, safe and efficient operation of the open street traffic system is promoted.

[0031] In a preferred embodiment of the present application, step 100, multi-source heterogeneous traffic data of an open street is collected, and the position and trajectory information of the traffic entities contained in the traffic data are converted into discrete spatial point cloud data, including:

[0032] Step 101, in combination with the characteristics of dense branch roads, mixed traffic of multiple modes, and concentrated school-going passenger flow in the morning peak in open streets, a combination of fixed monitoring, mobile sensing and platform linkage is adopted to comprehensively collect multi-dimensional traffic data covering motor vehicles, non-motor vehicles and pedestrians, ensuring that the data covers no blind area and information is full-dimensional:

[0033] In the traffic flow dense areas of the main and secondary entrances and exits of open streets, branch road intersections, key road sections around schools, and near commercial sites, high-definition video cameras, geomagnetic detectors, and pedestrian counting sensors are arranged. Among them, the high-definition video camera is used to capture the real-time motion state and appearance characteristics of various traffic entities, the geomagnetic detector is used to collect the number of motor vehicles and instantaneous stay information, and the pedestrian counting sensor is used to count the pedestrian flow density in key areas; at the same time, non-motor vehicle dedicated detectors are added at the key nodes of the branch road network to accurately collect the traffic trajectory fragments and flow data of bicycles, electric bicycles and other non-motor vehicles.

[0034] Step 102, through low-altitude unmanned aerial vehicle, the open street full-branch network is timed to cruise shooting, making up for the lack of fixed monitoring equipment in the regional coverage, obtaining the macro distribution of large-scale traffic flow; synchronously accessing the positioning trajectory data of regional shared bicycles, shared electric bicycle operation platform, and the journey trajectory data of local taxis and online car-hailing services, extracting dynamic information such as continuous driving paths and midway stops of non-motor vehicles and motor vehicles; in addition, personnel carrying portable positioning equipment collect pedestrian trajectory sample data along typical walking routes during the morning peak school commute period, supplementing the basic information of pedestrian traffic flow; linking the city traffic control platform, school intelligent management system, and surrounding parking lot management system, obtaining the school commute flow tide law, motor vehicle parking and driving out data, and traffic control temporary information during the morning peak period.

[0035] Step 103, the original traffic data collected from multiple sources is cleaned, standardized and integrated to eliminate data noise and format differences, ensuring the accuracy of subsequent conversion and analysis:

[0036] Invalid data is removed, including abnormal values caused by equipment failure, blurred monitoring data caused by weather or obstruction, and trajectory points deviating from the actual road caused by positioning drift; for a small amount of data missing during the collection process, a trend completion method based on adjacent period valid data is used to ensure the time sequence continuity of the data; at the same time, duplicate data is filtered to avoid data redundancy affecting processing efficiency.

[0037] Step 104, the coordinate system of data from different sources is converted to a unified local plane coordinate system, wherein the image pixel coordinates captured by the video camera are converted to actual geographic coordinates through preset calibration parameters, and the image data captured by the unmanned aerial vehicle is associated to the corresponding spatial position through geographic registration technology; the time stamps of all data are uniformly calibrated to the same timing standard to ensure the consistency of different types of data in the time dimension; in addition, the description fields of various traffic data are standardized and defined to determine the data attribute type, laying the foundation for subsequent data fusion. Combined with the collected traffic entity appearance characteristics, motion speed, and trajectory form, etc. information, the data is classified and labeled to determine the corresponding traffic entity type (such as small motor vehicles, large motor vehicles, bicycles, electric bicycles, pedestrians), and according to the associated platform data, the student group related traffic data during the morning peak school commute period is labeled.

[0038] Step 105, based on the pre-processed standardized data, through discretization processing and attribute packaging, the continuous traffic entity position and trajectory information is converted to structured spatial point cloud data:

[0039] The continuous driving or walking trajectory of each traffic entity is uniformly sampled at fixed time intervals to obtain a series of discrete trajectory nodes; for motor vehicles with high movement speed, the sampling time interval is appropriately reduced to ensure trajectory restoration accuracy, and for pedestrians and non-motor vehicles with low movement speed, the sampling interval can be reasonably increased to improve processing efficiency while ensuring data effectiveness; each discrete sampling trajectory node is assigned a unique spatial coordinate to form a basic spatial point; at the same time, each spatial point is encapsulated with corresponding attribute information, including the type of the traffic entity to which it belongs, the sampling time, the instantaneous movement state (static / moving), the corresponding road segment information, etc., to construct a complete point cloud data unit containing spatial and temporal attributes and entity attributes; all traffic entity point cloud data units are integrated according to spatial position and sampling time to form an initial spatial point cloud dataset covering the entire open street block; and spatial indexing is constructed for the integrated point cloud data.

[0040] In a preferred embodiment of the present application, based on the spatial distribution of point cloud data and the road topology of the open street block, a virtual analysis plane is constructed, including:

[0041] Step 106: Extracting the two-dimensional spatial range covered by the discrete spatial point cloud data as the basic geographical boundary of the virtual plane, including: first, screening the previously generated discrete spatial point cloud data to remove abnormal point clouds that exceed the actual geographical range of the open street block due to device errors and positioning drift (preliminary judgment can be made in combination with the administrative boundaries, land ownership range, etc. of the open street block), and retaining point cloud data that effectively reflect the movement trajectories of traffic entities within the block; for the filtered effective point cloud data, extract the coordinate extreme values in the two-dimensional plane (i.e. the unified local plane coordinate system mentioned earlier), including the maximum and minimum values of the horizontal coordinate and the maximum and minimum values of the vertical coordinate. These extreme values correspond to the easternmost, westernmost, southernmost, and northernmost boundary points of the point cloud data in space; using the four coordinate extreme values as vertices, a minimum rectangular region is constructed, which is the initial basic geographical boundary of the virtual plane, ensuring that all effective point cloud data are completely contained within this boundary, avoiding omission of key traffic data due to a too small boundary, or introduction of irrelevant regional interference in the analysis due to a too large boundary; the initial boundary is fine-tuned in combination with the actual geographical features of the open street block; if there is an area on the edge of the initial rectangular boundary that is not covered by point cloud data but belongs to the core road network of the block (such as newly built branches, edge parking lot entrances, etc.), the boundary range in the corresponding direction is appropriately expanded; if the initial boundary contains a large area of non-street block regions (such as parks, vacant lots, etc. where there is no traffic entity activity), the boundary is contracted to the edge of the actual road and building coverage range of the block, and finally a basic geographical boundary that accurately fits the core area of the traffic activity of the open street block is formed.

[0042] Step 107, according to the actual road network topology of the open street block, the basic geographic boundary is divided into analysis units corresponding to branches, intersections and interest point areas, including: first, collect the actual road network basic data of the open street block, including the direction, length and width of branches and main and secondary roads, the location and type of intersections (such as crossroads and T-shaped intersections), and the specific distribution of interest points (such as schools, community entrances, commercial sites, and bus stops, etc., with a focus on interest points related to early morning peak school traffic flow). At the same time, determine the connection relationship of the road topology, that is, the connection mode of each branch and intersection, and the intersection of roads of different levels.

[0043] According to the combed road topology information, the basic geographic boundary is divided into three types of analysis units to ensure that the unit division meets the needs of traffic flow analysis:

[0044] According to the actual direction and range of a single branch, extend to the edge of the road red line on both sides of the road center line to form an independent branch analysis unit; if two branches exist continuous connection and no obvious intersection separation, split the unit according to the road turning or slope change point to ensure that each branch unit corresponds to a complete and continuous single branch section; according to the actual control range of the intersection, cover the stop line of each incoming branch to the center area of the intersection, determine the unit range combined with the size of the intersection and the number of turning lanes to ensure that the unit can completely contain the intersection of each traffic flow, turning area, and avoid missing the core area of traffic conflict due to too small unit range; take the actual entrance of the interest point as the core and extend to the corresponding branch or sidewalk range to form an interest point associated analysis unit (such as the analysis unit of the school gate needs to cover the school gate to the adjacent branch walking path, non-motor vehicle parking area); if there are multiple people flow / vehicle flow entrances for the same interest point, separate independent units are divided to ensure accurate capture of the traffic flow characteristics around the interest point; calibrate the boundaries of each type of analysis unit after division to ensure that there is no overlap and no omission between adjacent units, the boundaries of branch units and intersection units are accurately connected to the intersection stop line position, and the boundaries of interest point associated units are consistent with the range of actual sidewalks and non-motor vehicle lanes; at the same time, label a unique identifier for each analysis unit to determine its corresponding actual road name, intersection number or interest point name, which is convenient for subsequent data association and feature analysis.

[0045] Step 108, for each analysis unit, a two-dimensional coordinate grid is allocated, and all point cloud data is mapped to the corresponding grid node according to its spatial coordinates to form a virtual analysis plane, including: in combination with the characteristics of open street branch dense and traffic entity movement space compact, a unified two-dimensional coordinate grid specification is determined. The grid size needs to consider the analysis accuracy and processing efficiency. If the grid is too large, different traffic flow data will be confused; if the grid is too small, the data processing load will be increased. Usually, the grid side length is determined according to the average width of the open street branch, to ensure that a single grid can accurately cover a single lane or sidewalk area of the branch, and each analysis unit contains several continuous grids; taking the basic geographic boundary determined in step 106 as the reference, a corresponding two-dimensional coordinate grid is allocated for each analysis unit; the coordinate system of all grids is consistent with the coordinate system of point cloud data and road topology data (that is, a unified local plane coordinate system), the unique coordinate code of each grid is determined, the coding rule includes the identification of the analysis unit to which it belongs and the relative position of the grid in the unit, to ensure that the grid can be accurately positioned to the corresponding actual area; the spatial coordinates of each valid point cloud data are extracted one by one, and the corresponding two-dimensional coordinate grid is matched according to the coordinate information, and the point cloud data is associated with the node of the corresponding grid. If the coordinates of a single point cloud data are exactly at the boundary of two grids, it is associated with the nearest grid node according to the nearest matching principle; if multiple point cloud data are mapped to the same grid node, the data is classified and summarized according to the type of traffic entity (motor vehicle, non-motor vehicle, pedestrian), to ensure that the point cloud data in the same grid node is classified clearly; after all point cloud data are mapped to the grid, all coordinate grids containing point cloud data and the corresponding analysis unit are integrated to form a complete virtual analysis plane. This plane not only retains the real spatial relationship of the open street road topology, but also realizes the structured distribution of point cloud data through the grid node, which can directly support subsequent short-distance OD pair analysis, key area identification and other operations, and realizes the accurate linkage of virtual space data and real traffic scene.

[0046] In a preferred embodiment of the present application, the point cloud data is processed by spatio-temporal alignment and fusion, and the association mapping relationship between motor vehicle and non-motor vehicle flow data is established on the virtual analysis plane to generate a street full-mode traffic flow fusion data set, including:

[0047] Step 109, all sources of point cloud data are converted to the same standard timestamp to obtain completed spatio-temporal alignment point cloud data, including: first determine the unified time reference standard, sort out the original timestamp information of all source point cloud data, including the time record format and sampling frequency difference of different collection means such as fixed monitoring equipment, unmanned aerial vehicle, shared bicycle platform. For example, the point cloud data with original time as local time is screened out, and is converted to UTC time through time zone conversion; for the scattered point cloud data without determined timestamp (such as manually collected walking trajectory samples), the UTC timestamp is supplemented in combination with the collection log; based on the dynamic change characteristics of open street traffic flow (frequent change of morning peak school traffic flow), the sampling time interval of all point cloud data is unified to a fixed value (such as 0.5 seconds). For point cloud data with sampling frequency higher than the standard interval (such as high-frequency video monitoring data), the point cloud of key time points is screened out in time sequence, and the core data reflecting the continuous motion state of traffic entities is reserved; for point cloud data with sampling frequency lower than the standard interval (such as part of GPS trajectory data), the point cloud information of missing time nodes is supplemented through time interpolation, ensuring the continuity of data in time dimension, and the interpolation process strictly conforms to the actual motion trend of traffic entities, avoiding distortion; combined with the spatial coordinate system of the virtual analysis plane, the spatial consistency of the point cloud data with unified timestamp is checked; check whether the spatial coordinates of each point cloud data are within the basic geographical boundary of the virtual analysis plane and match the corresponding grid node; for point cloud data with time alignment but abnormal spatial coordinates (such as drift points beyond the street range), further screening and elimination is carried out, and finally the time-unified, space-accurate spatio-temporal alignment point cloud data is obtained.

[0048] Step 110, based on the virtual gridding plane, the point cloud data from different detection means and after spatio-temporal alignment is classified and mapped into two dynamic data layers of motor vehicles and non-motor vehicles according to the corresponding traffic entity types, including: based on the classification and labeling results in the previous data preprocessing stage, the type of the point cloud data after spatio-temporal alignment is verified again. Combining the attribute information of the point cloud data (such as motion speed, trajectory shape, and associated platform data), the errors in the preliminary classification are corrected, for example, excluding electric bicycles with too high speed from the non-motor vehicle category, accurately classifying small passenger cars into the motor vehicle category, and determining the separate classification of pedestrian point cloud data; based on the virtual gridding plane, two parallel dynamic data layers, motor vehicle dynamic data layer and non-motor vehicle dynamic data layer, are built. Both data layers follow the grid division rules of the virtual plane, retain the correspondence between each grid node and the analysis unit (branch, intersection, and interest point), and each data layer is pre-set with data classification fields (such as timestamp, traffic entity ID, motion speed, and trajectory node) to provide structured support for the mapping of point cloud data; according to the traffic entity type, the point cloud data from different detection means is mapped to the corresponding dynamic data layer one by one. Among them, the motor vehicle point cloud data (including all types of vehicles) is all mapped to the motor vehicle dynamic data layer, the non-motor vehicle point cloud data (including bicycles and electric bicycles) and the pedestrian point cloud data are mapped to the non-motor vehicle dynamic data layer; during the mapping process, each point cloud data is accurately matched to the corresponding grid node of the virtual plane, and its analysis unit information (such as a motor vehicle point cloud mapped to XX branch-3 grid node) is associated synchronously; the same type of point cloud data in the same grid node is sorted in chronological order to form a structured dynamic data sequence.

[0049] Step 111, the road intersection and the road section are associated nodes, and the associated mapping relationship between the traffic, speed and trajectory conflict points of the motor vehicle data layer and the non-motor vehicle data layer is established on the virtual grid plane, including: taking the intersection analysis unit and the branch analysis unit in the virtual analysis plane as the core, the specific range of the associated node is determined. The intersection associated node covers all the grid areas of the intersection analysis unit, and focuses on the core grid of the traffic flow intersection (such as the grid corresponding to the turning lane); the road section associated node covers the key road section of the branch analysis unit (such as the school surrounding branch, the adjacent branch of commercial network), and several sub-associated nodes are evenly divided according to the length of the road section, so as to ensure that the interaction state of the traffic flow in the road section can be fully captured; at the same time, a unique identifier is assigned to each associated node to determine its corresponding actual road position; at each associated node, the core parameters are extracted from the motor vehicle and non-motor vehicle dynamic data layer respectively, including traffic flow (the number of traffic entities passing through the node per unit time), speed (the instantaneous speed of the traffic entity passing through the node), and trajectory node (the specific grid coordinates of the traffic entity passing through the node). Through time stamp association, the corresponding relationship between the parameters of the two types of data layers in the same time dimension is established, for example, the motor vehicle traffic and non-motor vehicle traffic of a certain intersection associated node at 7:40 in the morning peak, and the trajectory intersection point (i.e. trajectory conflict point) of the motor vehicle and non-motor vehicle in the same grid node.

[0050] Based on the extracted associated parameters, the associated mapping relationship between the two types of data layers is constructed on the virtual grid plane. For traffic association, a four-dimensional mapping relationship of intersection / road section node-time-motor vehicle traffic-non-motor vehicle traffic is established to mark the traffic running state under different traffic combinations; for speed association, the mutual influence relationship between the motor vehicle speed and the non-motor vehicle speed at the same node is constructed to reflect the coordinated running law of the two; for trajectory conflict point, the grid coordinates, conflict occurrence time and involved traffic entity type of the conflict point are marked to form a complete associated mapping relationship network, and all the mapping relationships are accurately bound with the grid nodes of the virtual plane.

[0051] Step 112, according to the associated mapping relationship, the static basic road network layer containing road attributes and restriction information is superimposed and fused to generate a street full-mode traffic flow fusion data set containing time, space, mode, traffic and mutual relationship, including: collecting the static basic road network information of the open street, constructing the static basic road network layer, which mainly contains two types of information, road attribute information (such as branch grade, road width, number of lanes, road surface material) and road restriction information (such as speed limit requirement, motor vehicle prohibited period, non-motor vehicle exclusive lane range, intersection turning restriction). These information is accurately matched to the corresponding analysis unit and grid node of the virtual analysis plane to ensure that the static information and the dynamic data layer are completely aligned in space, for example, the restriction information of XX branch speed limit 30km / h is associated with all the grid nodes of the branch analysis unit.

[0052] The motor vehicle dynamic data layer, the non-motor vehicle dynamic data layer and the static basic road network layer are superimposed and fused. In the fusion process, the grid nodes of the virtual grid plane are taken as the core correlation carriers, the dynamic data (flow, speed, trajectory) corresponding to the same grid node is bound with the static information (road attribute, restriction requirement), and meanwhile, the correlation mapping relationship established in the integration step 111 is integrated, the flow, speed and trajectory conflict correlation information of the motor vehicle and the non-motor vehicle is supplemented to the corresponding grid node, and an integrated data structure of grid node-time-traffic mode-dynamic parameter-static attribute-correlation relationship is formed.

[0053] The superimposed and fused data is optimized and processed, the repeated data and invalid correlation information (such as grid node data without actual traffic flow) are removed, and the data is classified and sorted according to the time dimension, the space dimension and the traffic mode dimension. The finally generated block full-mode traffic flow fused data set needs to completely contain the core information of time (uniform UTC timestamp), space (grid node, analysis unit, actual road position), mode (motor vehicle, non-motor vehicle, pedestrian), flow (unit time flow of each mode) and mutual relationship (flow correlation, speed correlation, trajectory conflict relationship), and the data format is adapted to the operation needs of subsequent short-distance OD pair analysis and Gaussian mixture model parameter extraction.

[0054] In this embodiment, by unifying the timestamp reference and the standardized time interval, the time deviation caused by the multi-source collection means (fixed monitoring, mobile sensing, etc.) is eliminated; the spatial anomaly data is removed by combining the virtual analysis plane to complete the time and space double verification, ensuring that all point cloud data are accurately aligned in the time and space dimensions, avoiding the analysis deviation caused by different data synchronization; the point cloud data is classified and mapped to the motor vehicle and the non-motor vehicle two dynamic data layers according to the traffic entity type, the data flow of different traffic modes is clearly distinguished, and the problem of mixed and disordered multi-source data is solved; meanwhile, relying on the spatial correlation of the virtual grid plane, the data is accurately bound with the actual road network unit, which facilitates the subsequent targeted extraction of the flow characteristics of specific areas and specific traffic modes, greatly improves the data retrieval and utilization efficiency; the correlation mapping relationship of the motor vehicle and non-motor vehicle flow, speed and trajectory conflict is established with the intersection and the road section as the core correlation nodes, the interaction state of the open block multi-mode traffic flow is quantified, the limitation of the traditional data processing in the motor vehicle and non-motor vehicle data separation analysis is broken through, the mixed interference core area and the key period can be accurately identified, and key interaction relationship data is provided for the subsequent extraction of mutual interference parameters and the optimization of flow distribution.

[0055] In a preferred embodiment of the present application, step 200, based on the street-wide traffic flow fusion dataset, the travel time distribution characteristics and travel mode composition characteristics of short-distance OD pairs are analyzed on the virtual analysis plane to identify the walking and non-motorized vehicle dominant characteristics of school commuting passenger flow during the morning peak period, including:

[0056] Step 201, based on the street-wide traffic flow fusion dataset, short-distance OD pairs with travel distances below a preset threshold are screened out on the virtual analysis plane, and their corresponding point cloud trajectory data is extracted, including: considering the scene characteristics of short-distance travel in open streets, referring to the street branch network density and the general short-distance travel cognition (such as the reachable range within 15 minutes of walking), the travel distance preset threshold (for example, 1.5 kilometers) is set; this threshold can be flexibly adjusted according to the actual size of different open streets to ensure that most of the residents' daily short-distance travel scenarios are covered; relying on the street-wide traffic flow fusion dataset, all complete OD pairs (i.e. traffic travel units containing determined starting point, ending point and complete travel trajectory) are identified on the virtual analysis plane; through the spatial coordinate correlation of the virtual analysis plane, the grid node positions corresponding to the starting point and ending point of each OD pair are matched to determine their actual geographical distance, and it is judged whether the distance is below the preset short-distance threshold; all short-distance OD pairs with travel distances below the preset threshold are screened out, and their unique identification information (such as OD pair number, starting and ending grid node code) is recorded; the complete point cloud trajectory data corresponding to these short-distance OD pairs is extracted from the fusion dataset, including all grid nodes on the trajectory, the time stamp corresponding to each node, the type of traffic entity, etc., to ensure that the extracted trajectory data completely covers the entire travel process from the starting point to the ending point; the validity of the extracted short-distance OD pair point cloud trajectory data is checked, and invalid data such as incomplete trajectories (e.g. missing starting or ending point data) and trajectory nodes deviating from the virtual analysis plane range are removed; the trajectory data that passes the verification is classified and archived according to the unique identification of the OD pair to form a structured short-distance OD pair trajectory dataset.

[0057] In step 202, the point cloud trajectory data is processed to count the average travel time and travel time variance of each short-distance OD pair in different time periods to form a travel time distribution feature set, including: for the point cloud trajectory data of each short-distance OD pair, extracting the time stamp corresponding to the starting point of the trajectory (the start time of the trip) and the time stamp corresponding to the end point (the end time of the trip), and calculating the actual duration of a single trip; if there are multiple trip trajectories for the same OD pair, the travel time of each trajectory is calculated to form a set of original travel times of the OD pair; in combination with the conventional time period characteristics of urban traffic travel, the whole day analysis period is divided into multiple continuous sub-periods (for example, each 30 minutes is a sub-period, covering the 06:00-22:00 core travel period); each sub-period corresponds to a time interval identifier, which is convenient for subsequent period-based statistical characteristics; according to the divided sub-periods, the original travel time set of each short-distance OD pair is classified and counted; for each sub-period, the average travel time of all trips of the OD pair in the period is calculated, and the dispersion degree of the travel time (i.e. the travel time variance) is also calculated, which intuitively reflects the stability of the travel time of the OD pair in the period; the average travel time and travel time variance of each OD pair in each sub-period are bound with the corresponding sub-period identifier to form the time distribution feature of a single OD pair; the time distribution features of all short-distance OD pairs are aggregated and sorted by the unique identifier of the OD pair to form a unified travel time distribution feature set; the OD pair information, sub-period, average travel time and travel time variance corresponding to each record in the feature set are determined for subsequent quick retrieval and analysis.

[0058] In step 203, based on the set of travel time distribution characteristics, the analysis period is focused on the morning peak period, and based on the short-distance OD pairs, the number of trips and the proportion of trips by walking, non-motor vehicles and motor vehicles in the morning peak period are counted, and a set of morning peak travel mode composition characteristics is formed, including: in combination with the travel rules of open block school commuting passenger flow, the range of conventional urban morning peak period is referred to, and the morning peak analysis period (for example, 7:30-8:30, which can be flexibly adjusted according to the class time of different school blocks) is defined; from the set of travel time distribution characteristics, all short-distance OD pair records with travel start time falling within the morning peak period are screened out; for the screened short-distance OD pairs in the morning peak period, relying on the traffic entity type information in the block full-mode traffic flow fusion data set, the travel mode (i.e. walking, non-motor vehicle, motor vehicle) corresponding to each OD pair and each travel track is matched; if there are multiple mode switches in a single track (such as walking to the intersection and then taking a non-motor vehicle), the mode that dominates the travel time is taken as the core mode of this trip; group by OD pair unique identifier, count the total number of trips of each short-distance OD pair in the morning peak period; at the same time, count the number of trips by walking, non-motor vehicle and motor vehicle respectively, and calculate the proportion of the number of trips by each mode to the total number of trips (i.e. travel mode proportion); bind the total number of trips, the number of trips by each mode and the corresponding proportion of each morning peak period short-distance OD pair with the OD pair unique identifier, and collect all OD pair records that meet the conditions to form a set of morning peak travel mode composition characteristics; the set of characteristics clearly distinguishes the travel mode distribution differences of different OD pairs.

[0059] In step 204, in combination with the preset school-related POI location information, the OD pairs with a school as the core and showing an aggregated flow direction on the virtual analysis plane are screened from the early morning peak travel mode composition feature set to form a school commuting candidate OD set, including: collecting the location information of all schools (including primary schools and middle schools) in the open block, and determining the specific geographic coordinates of each school; mapping the school location information to the virtual analysis plane, determining the grid node range corresponding to each school, forming a school commuting POI coordinate set, and labeling the name of each school, the school starting time and other auxiliary information; screening the short-distance OD pairs with the origin or destination falling within the grid node range corresponding to the school-related POI (school) from the early morning peak travel mode composition feature set; that is, determining the OD pairs with the travel origin being the school periphery (such as the grid within 50 meters outside the school gate) or the destination being the school, and including them in the preliminary set of school-related OD pairs; on the virtual analysis plane, the travel flow direction (from the origin grid to the destination grid) of each OD pair in the preliminary set is visually presented; analyzing the flow direction distribution of these OD pairs, and screening the OD pairs showing the aggregated flow direction characteristics, for example, the destinations of multiple OD pairs all pointing to the same school, or the origins of multiple OD pairs converging to the periphery grid of the same school from different directions, forming a radial or converging flow direction with the school as the core; summarizing all the OD pairs meeting the school-related and aggregated flow direction conditions to form the school commuting candidate OD set; labeling the corresponding school information and flow direction type (converging type / radiation type) of each OD pair in the candidate OD set, facilitating the subsequent dominant feature determination.

[0060] In step 205, for the school commuting candidate OD set, the total proportion of walking and non-motor vehicle travel modes of each OD pair is calculated according to the corresponding early morning peak travel mode composition characteristics. If the total proportion exceeds the preset dominant threshold, it is determined that the passenger flow of the OD pair has the walking and non-motor vehicle dominant characteristics, including: combining the conventional travel characteristics of open block school commuting passenger flow, the walking and non-motor vehicle dominant threshold (for example, 70%) is preset. The threshold can be flexibly adjusted according to the type of school in the block and the degree of perfection of the surrounding slow traffic facilities to ensure that the slow and motor dominant school commuting passenger flow can be accurately distinguished. From the early morning peak travel mode composition characteristics, the walking travel proportion and non-motor vehicle travel proportion corresponding to each OD pair in the school commuting candidate OD set are extracted. The walking proportion and non-motor vehicle proportion of the same OD pair are added to obtain the total proportion of walking and non-motor vehicle travel modes of the OD pair, which quantitatively reflects the dominant degree of slow travel. The total proportion of each school commuting candidate OD pair is compared with the preset dominant threshold. If the total proportion of a certain OD pair exceeds the preset threshold, it is determined that the passenger flow corresponding to the OD pair has the walking and non-motor vehicle dominant characteristics. If it does not exceed the threshold, it is determined to be non-slow dominant characteristics (such as motor dominant or no obvious dominant mode). The information of all OD pairs determined to have walking and non-motor vehicle dominant characteristics is recorded, including OD pair identification, associated school, total proportion, early morning peak travel times, etc. These OD pairs are summarized to form the early morning peak school commuting slow dominant OD set, providing accurate passenger flow guidance basis for subsequent key analysis area determination.

[0061] In this embodiment, OD is a core term in traffic planning and flow analysis, which means Origin-Destination, specifically refers to the pairing relationship between the origin and destination of a complete trip (i.e. the origin-destination pair); the short-distance OD pair in step 200 refers to the origin-destination combination within an open block with a travel distance below a preset threshold (such as the travel origin-destination combination from home to a nearby school or community commercial site); the school commuting OD pair in step 200 is a specific type of short-distance OD pair, specifically refers to the origin-destination combination with a school (school interest point) as the origin or destination, and the travel flow direction presents aggregation (such as multiple origins converging to the same school), and the core corresponds to the origin and destination of the early morning peak school commuting passenger flow. In short, OD is essentially a refinement of the core information of "who travels from where and where to go", and is the basis for subsequent analysis of travel characteristics, delineation of key areas, and optimization of flow distribution.

[0062] The short-distance OD pair is screened by a preset threshold value, and the corresponding trajectory data is extracted, the interference of long-distance travel data is eliminated, the subsequent analysis focuses on the high-frequency travel scene of the open block, the invalid data processing cost is reduced, the pertinence of feature analysis is improved, the early morning peak period is focused based on the travel time distribution characteristics, the core travel window of school travel passenger flow is accurately matched, accurate time dimension support is provided for subsequent school travel passenger flow identification and mode composition analysis, key feature dilution caused by full-period analysis is avoided, school candidate OD pairs are accurately locked by combining the school interest point and the aggregated flow direction screening of the virtual analysis plane, the limitation of relying only on distance to determine school travel passenger flow in the traditional analysis is overcome, the interference of non-school travel passenger flow is reduced, and the accuracy of passenger flow identification is improved, and the slow travel dominant feature of school travel passenger flow is quantified by calculating the total proportion of walking and non-motor vehicles and comparing the preset threshold value, clear target orientation is provided for subsequent key analysis area demarcation of branch network and extraction of mixed interference parameters, and the subsequent analysis is accurately matched with the core demand of the school travel scene of the open block.

[0063] In a preferred embodiment of the application, according to the dominant feature, the key analysis area is determined for the branch network on the virtual analysis plane, and the basic triangular unit is determined in the key analysis area; the basic triangular unit is processed to define the interaction range of multi-mode traffic flow, including:

[0064] In step 206, according to the determined OD pairs with walking and non-motor vehicle dominant characteristics, the origin and destination points and the travel paths of the OD pairs are positioned on the virtual analysis plane, and the continuous road segment area carrying the main traffic and composed of branches is determined as the key analysis area, including: calling the early morning peak school travel slow travel dominant OD set determined in the foregoing, matching the grid nodes corresponding to the origin and destination of each OD pair on the virtual analysis plane one by one, extracting the complete travel path (composed of continuous grid nodes passed through by trajectory data) of each OD pair, and associating the path with the branch network in the virtual plane to mark the specific branch name and road segment range covered by the path; based on the street full mode traffic flow fusion data set, the traffic data of each road segment on the travel path of each slow travel dominant OD pair is counted, and the branches and main roads (excluding the main road segments contained in the path) are distinguished; the proportion of each branch road segment in the total traffic of the corresponding OD pair is calculated, and the branch road segments with a proportion exceeding a preset proportion (such as 60%) are screened out and determined as the core carrying branch of the OD pair; the core carrying branches carrying the traffic of multiple slow travel dominant OD pairs in the same area are integrated, if the multiple core carrying branches are connected to form a continuous road segment network, and the total traffic in the network accounts for more than a preset threshold (such as 75%) of the total traffic of all slow travel dominant OD pairs, the spatial range (mapped to the grid area of the virtual analysis plane) corresponding to the continuous branch network is determined as the key analysis area; each key analysis area is assigned a unique identifier, and the branch name covered thereby, the associated slow travel dominant OD pair information is marked; the preliminary determined key analysis area is checked to ensure that the region boundary completely covers the origin to the destination of the core carrying branch, and does not include non-branch areas (such as parks and open spaces); if there is an overlap between adjacent key analysis areas, they are merged into a unified key analysis area to avoid repeated analysis.

[0065] Step 207, in each of the key analysis areas, a corresponding intersection is selected as a core point, and the adjacent intersections in the three key branch directions directly connected with the intersection are taken as vertices, and the basic triangular unit is connected on the virtual analysis plane, comprising: in each key analysis area, combining the branch network topology and traffic data, the intersection with the most dense traffic flow and the most number of connected branches is selected as the core point; the core intersection needs to meet two conditions of directly connecting at least three branches and being the intersection of at least two slow travel dominant OD pair paths, to ensure that it is the core hub of the regional traffic flow; taking the core intersection as the center, combing all branches directly connected with it, three key branches with the largest traffic flow and belonging to the slow travel dominant OD pair core bearing branch are selected; the intersections directly adjacent to the core intersection on the three key branches (i.e. the next intersection from the core intersection of each key branch) are located, and the three adjacent intersections are taken as the vertices of the triangular unit; on the virtual analysis plane, the core intersection and the three vertex intersections are connected in turn by straight lines, and the three vertex intersections are connected, to form a closed basic triangular unit; it is ensured that the triangular unit is completely located within the key analysis area, and can cover the core intersection and the core intersection section of the three key branches; if the range of a single key analysis area is large, multiple non-overlapping basic triangular units can be constructed according to the above method to ensure full coverage of the area; for each basic triangular unit, the corresponding key analysis area identifier, core intersection name, three vertex intersection name and associated key branch information are labeled, to form a structured unit information table.

[0066] Step 208, the vertices of the base triangular unit and the core point are taken as a set of spatial point position collection; the spatial point position collection is processed to calculate and generate a minimum convex polygon region containing all points of the spatial point position collection, including: for each base triangular unit, extracting the spatial coordinates of its core intersection and three vertex intersection (based on the unified coordinate system of the virtual analysis plane), and integrating the four points into a complete set of spatial point position collection; if there are multiple base triangular units in the same key analysis area, arrange the point position collection by unit grouping to avoid confusion of point positions of different units; perform convex hull analysis on each set of spatial point position collection, the process is: first determine the point with the minimum horizontal coordinate in the point position collection as the starting reference point, then take this point as the center and connect the other points in anticlockwise direction, remove the points inside the connected line segment (i.e. the points of non-convex vertices), and keep the points on the convex hull boundary; connect the boundary points retained after convex hull analysis in anticlockwise direction to form a closed polygon region, which is the minimum convex polygon containing all points of the set of spatial point position collection; ensure that each edge of the polygon fits the actual spatial boundary of the key branch or intersection, and does not cross the non-traffic area (such as buildings, green belts); check whether the generated minimum convex polygon completely contains the corresponding base triangular unit and the core intersection section, if there is point omission or boundary deviation, adjust the boundary determination standard of convex hull analysis until the generated convex polygon accurately covers the target point position and the core traffic area.

[0067] Step 209, the minimum convex polygon region is defined as the interaction range of the actual speed change, trajectory interweaving and flow competition of the multi-mode traffic flow in the key analysis region, including: calling the full-mode traffic flow fusion data set of the block, extracting the traffic flow data in the minimum convex polygon region, focusing on analyzing the speed change, trajectory interweaving and flow competition characteristics, including the speed drop section of motor vehicles and non-motor vehicles, the intersection point of trajectories of different traffic modes, and the competition area of multi-mode flow in the same lane; if the traffic flow data in the minimum convex polygon region meets the conditions of containing at least 3 speed drop points, there are not less than 5 groups of different mode trajectory interweaving and the proportion of the multi-mode flow competition time length of the core section to the morning peak time is more than 80%, it is determined that the convex polygon region can accurately cover the core interaction area of the multi-mode traffic flow, and the convex polygon region is formally defined as the multi-mode traffic flow interaction range in the corresponding key analysis region; the boundaries of each interaction range are clearly marked on the virtual analysis plane, and the corresponding key analysis region, basic triangular unit and traffic flow interaction feature data are associated; the information of all interaction ranges is summarized and archived to form the associated data table of key region-convex polygon range-interaction feature, which provides clear spatial positioning basis for subsequent extraction of multi-mode traffic flow interference parameters; the defined interaction range is globally checked, if the interaction ranges of adjacent key analysis regions overlap, the boundary ownership of the overlapping area is adjusted combined with the actual traffic flow interaction intensity, to ensure that each traffic flow interaction point only belongs to one interaction range, and to avoid repeated interference parameter extraction.

[0068] In this embodiment, the key analysis region is determined based on the flow bearing characteristics of the slow travel dominant OD pair, avoiding indiscriminate analysis of the entire branch network, focusing on the core branch area where the school traffic flow is concentrated, greatly improving the efficiency and accuracy of subsequent analysis, and solving the problems of generalization and fuzzy focus of traditional analysis range; through the basic triangular unit construction, the complex branch network is converted into a quantifiable and structured unit for analysis, which adapts to the spatial characteristics of open block branch intensive and complex intersection, provides a stable benchmark framework for subsequent convex hull algorithm processing, and avoids analysis disorder caused by chaotic road network; the interaction range is defined based on the minimum convex polygon, which can accurately frame the core area of multi-mode traffic flow speed change and trajectory interweaving, overcoming the limitations of traditional range division based on administrative boundaries or road sections, and ensuring that the subsequently extracted mutual interference parameters can truly reflect the actual traffic flow interaction state.

[0069] In a preferred embodiment of the present application, according to the interaction range, the point cloud data on the virtual analysis plane is mapped to the corresponding actual road network area to extract the mutual interference parameters of the multi-mode traffic flow in the branch network, and generate the initial parameter set of the Gaussian mixture model, including:

[0070] Step 210, on the virtual analysis plane, the actual branch network profile covered by the interaction range is extracted as a polygonal road area, including: calling the multi-mode traffic flow interaction range (minimum convex polygon area) defined in the foregoing, combining the basic geographic data of the open block actual road network (including branch direction, road red line, intersection profile, etc.), establishing the spatial correlation of the interaction range and the actual branch network on the virtual analysis plane, and clearly defining the specific branch road section, intersection and auxiliary traffic area (such as non-motor vehicle lane, sidewalk) covered by the former; based on the unified coordinate system of the virtual analysis plane, the coordinates of the outer contour boundary points of the actual branch network in the interaction range covered range are extracted, and these boundary points need to accurately correspond to the turning points of the actual road red line, the corner points of the intersection and other key positions; at the same time, the boundary points corresponding to the non-road area (such as building occupied land, green belt, parking lot, etc.) in the contour are removed, so as to ensure that the extracted contour only reflects the road passing area; the extracted actual road network contour boundary points are connected in sequence in the clockwise direction to form a closed polygonal road area; it is ensured that the polygon can completely cover all the actual branch passing areas in the interaction range, and the boundary is completely matched with the actual road red line; each polygonal road area is labeled with the corresponding interaction range identifier and the covered branch name to form a structured area correlation table; the generated polygonal road area is superimposed on the interaction range of the virtual analysis plane, and the spatial fitting degree of the two is checked; if there are polygons exceeding the interaction range or not completely covering the road in the interaction range, the boundary point coordinates are adjusted for correction; at the same time, the closure and rationality of the polygon are checked to avoid problems such as boundary intersection and gap.

[0071] Step 211, triangulate the polygonal road area on the virtual analysis plane into a plurality of non-overlapping triangular sub-units, and calculate the area of each triangular sub-unit, including: in combination with the characteristics of the open block branch network being dense and the intersections being many, develop a triangulation rule: first, the triangular sub-units after triangulation need to completely cover the polygonal road area, and there is no overlap and no gap between adjacent sub-units; second, the vertices of the triangular sub-units preferentially select key positions such as road intersection corner points and branch centerline turning points, to ensure that the sub-unit boundary fits the actual road direction; third, the edge length of a single triangular sub-unit should not be too long, to avoid insufficient data mapping accuracy due to the sub-unit being too large; based on the developed rule, triangulate each polygonal road area, starting from one vertex of the polygon, sequentially connecting other non-adjacent vertices, to decompose the polygon into a plurality of independent triangular sub-units; for a complex polygonal area containing intersections, take the intersection center as the core, and radiate to the peripheral road red line turning points, to realize accurate correspondence between the sub-units after triangulation and the flow directions of the intersections; based on the coordinate system of the virtual analysis plane, calculate the virtual area of each sub-unit through the triangular area algorithm; at the same time, in combination with the scale of the virtual analysis plane and the actual geographic space, convert the virtual area into the corresponding actual road area; assign a unique identifier to each triangular sub-unit, record its virtual area, actual area, and the identifier of the polygonal road area to which it belongs, to form a sub-unit area information table; check whether the total area of all triangular sub-units is consistent with the area of the corresponding polygonal road area, to ensure that there is no area omission or repeated calculation; at the same time, check whether the sub-units exist across non-road areas, if so, readjust the triangulation nodes until all sub-units are completely located within the actual road traffic area.

[0072] Step 212, according to the street full mode traffic flow fusion dataset, all point cloud data falling within the interaction range is obtained, and according to its spatial coordinates, it is mapped to the corresponding triangular subunit by calculating its inclusion relationship with each triangular subunit, including: from the street full mode traffic flow fusion dataset, all point cloud data whose spatial coordinates fall within the interaction range (the smallest convex polygon region) is screened out, and these data are the core data reflecting the multi-mode traffic flow state of the region; At the same time, the abnormal points (such as drift points deviating from the road area, repeated data) in the screened data are removed to ensure the effectiveness of the mapping data; The inclusion relationship judgment logic of point cloud data and triangular subunit is formulated: taking the three vertices of a single triangular subunit as the benchmark, it is judged whether the spatial coordinates of the point cloud data are located inside or on the boundary of the subunit; For the point cloud data located on the boundary of the subunit, according to the principle of nearest attribution, it is mapped to the adjacent triangular subunit corresponding to the more matched road flow direction; The point cloud data coordinates after screening are read one by one, and according to the determined judgment logic, the spatial matching is carried out with each triangular subunit, and the point cloud data is accurately mapped to the corresponding subunit; At the same time, the mapping results of each point cloud data are recorded, including the identification of the triangular subunit, the corresponding actual road area position, the traffic entity type and other information; The point cloud data in the same triangular subunit is classified and archived according to the traffic entity type, which is convenient for subsequent parameter extraction; The total amount of point cloud data mapped to the triangular subunit is counted, and compared with the total amount of the target point cloud data screened out to ensure that there is no data omission; At the same time, part of the point cloud data is randomly extracted, and it is checked whether the subunit mapped by the point cloud data matches the actual road position, and the mapping accuracy is verified.

[0073] Step 213, based on the mapping relationship, the number of different traffic mode point clouds in each triangular subunit, the average speed and the speed variance are counted, and the speed interference intensity and the density interweaving coefficient of multi-mode traffic flow in the triangular subunit are calculated by combining the area of the triangular subunit, as mutual interference parameters, including: for each triangular subunit, based on the mapped point cloud data, the core features are counted respectively according to the traffic entity type (motor vehicle, non-motor vehicle, pedestrian): first, the number of point clouds (indirectly reflecting the traffic scale of the corresponding traffic mode); second, the average speed (reflecting the traffic efficiency of the corresponding traffic mode in the region); third, the speed variance (reflecting the stability of the speed of the corresponding traffic mode, the greater the variance, the more obvious the disturbance); combined with the statistical basic features and the actual area of the triangular subunit, two types of core mutual interference parameters are calculated: first, the speed interference intensity, which is calculated by the difference of the average speed of different traffic modes and the weighted speed variance, reflecting the interference degree of the speed difference between motor vehicles and slow traffic (non-motor vehicles, pedestrians) on the traffic; second, the density interweaving coefficient, which is calculated by the ratio of the number of point clouds of different traffic modes in the unit actual area, reflecting the interweaving density of multi-mode traffic flow in space, the greater the ratio, the more intense the interweaving, the stronger the mutual interference; the mutual interference parameters calculated are reasonably checked, and the abnormal parameter values caused by too little point cloud data are eliminated; for the triangular subunits with parameter values exceeding the reasonable range, combined with the actual traffic flow observation situation (such as actual investigation, video monitoring playback), the parameters are corrected to ensure that the parameters can truly reflect the multi-mode traffic flow interference state of the region; the interference parameters after checking and correcting are bound with the corresponding triangular subunit identifier to form a mutual interference parameter set.

[0074] Step 214, according to the statistical distribution characteristics of the mutual interference parameters extracted from all the triangular sub-units, determine the number of clusters of the Gaussian mixture model, and take the mean and covariance of the interference parameters of each sub-unit as the initial values to generate the initial parameter set of the Gaussian mixture model, including: statistically analyzing the mutual interference parameters (velocity interference strength, density interweaving coefficient) of all the triangular sub-units, identifying the distribution characteristics of the parameters through frequency distribution statistics, trend fitting, etc., including the central interval, dispersion degree, peak value number, etc. of the parameters; for example, if the velocity interference strength presents multiple obvious peak value distributions, it indicates that there are multiple typical interference states in the region; according to the number of distribution peaks of the mutual interference parameters, determine the number of clusters of the Gaussian mixture model; the clustering needs to ensure that the interference parameters in each cluster have similar distribution characteristics, corresponding to a typical multi-mode traffic flow interference state; for example, if the parameter distribution presents three obvious peaks, it indicates that there are three typical interference states, corresponding to setting the number of clusters of the Gaussian mixture model to 3; taking the mutual interference parameters in each cluster as a sample set, calculate the mean and covariance of each sample set as the initial mean and initial covariance of the corresponding component of the Gaussian mixture model; at the same time, statistically analyze the proportion of the number of triangular sub-units in each cluster as the initial weight of the corresponding component; the mean, covariance and weight together constitute the core initial parameters of the Gaussian mixture model; arrange the extracted core initial parameters according to the format requirements of the Gaussian mixture model to generate the initial parameter set of the Gaussian mixture model; perform a reasonableness check on the initial parameter set to ensure that the initial mean of each component is within the reasonable distribution interval of the corresponding interference parameter, and the covariance can reflect the dispersion characteristics of the parameter; after the check is passed, associate and archive the initial parameter set with the corresponding triangular sub-unit and the information of the interaction range.

[0075] In this embodiment, the mapping of point cloud data from the virtual analysis plane to the actual road area is completed through the triangular area algorithm, solving the problem of disconnection between traditional virtual data and actual road network, ensuring that the interference parameters extracted subsequently can be accurately associated with the actual road area, and improving the practical application value of the parameters; through the extraction of parameters such as velocity interference strength and density interweaving coefficient, the originally abstract multi-mode traffic flow mutual interference phenomenon is converted into quantifiable indicators, overcoming the limitation that traditional analysis can only qualitatively describe interference and cannot be quantified; based on the real distribution characteristics of the interference parameters, the number of clusters and the initial parameters of the Gaussian mixture model are determined, ensuring that the initial parameter set can accurately match the multi-mode traffic flow interference state of the open street block network, avoiding the problem of slow model convergence and large deviation of optimization results caused by traditional empirical setting of initial parameters, and improving the efficiency and accuracy of subsequent iterative optimization.

[0076] In a preferred embodiment of the present application, step 300, based on the initial parameter set of the Gaussian mixture model, iteratively optimizes the distribution of open street OD pairs, dynamically adjusts the path selection probability, obtains the final traffic allocation scheme, and outputs the real-time traffic allocation proportion of each path, including:

[0077] Step 301, based on the initial parameter set of the Gaussian mixture model, generates the initial path selection probability of all feasible paths within the interaction range for each open street OD pair on the virtual analysis plane, including: for each open street OD pair on the virtual analysis plane (including the early morning peak school commuting dominant OD pair and other regular OD pairs identified earlier), combining the open street branch network topology, the static basic road network layer road restriction information (such as no entry, restricted entry requirements), filtering out all feasible paths from the starting point to the ending point; the feasible path must consist of branches or allowed road segments, avoiding non-passable areas and restricted road segments, while ensuring path connectivity and integrity; comb through each feasible path one by one to determine whether it passes through or covers the multi-modal traffic flow interaction range defined earlier, record the association between the path and each interaction range (such as the interaction range identifier, the length of the road within the range); at the same time, call the interference parameter distribution characteristics of the corresponding interaction range in the initial parameter set of the Gaussian mixture model as the core basis for setting the path selection probability; based on the speed interference intensity, density interlacing coefficient and other interference parameters in the initial parameter set of the Gaussian mixture model, evaluate the traffic friendliness of each feasible path, the path with less interaction range and lower interference parameter value (i.e. less traffic interference and higher efficiency) is given a higher initial selection probability; otherwise, a lower probability is given; ensure that the sum of the initial selection probabilities of all feasible paths for the same OD pair is 1, form the initial path selection probability set for the OD pair, and label the probability value of each path and the corresponding passing area information; check the generated initial path selection probability, exclude unreasonable cases with probability of 0 or abnormally high (such as single path probability exceeding 90%); if there are unreasonable probabilities, adjust the probability distribution according to the actual traffic conditions of the path (such as road width, slow walking facility perfection), ensure that the initial probability can preliminarily reflect the actual traffic advantage difference of the path.

[0078] Step 302, according to the initial path selection probability, the total traffic demand of the OD pair is allocated to each feasible path to form an initial flow allocation scheme; calculate the real-time traffic state parameters of each interaction range under the initial flow allocation scheme, including: obtaining the total traffic demand of each OD pair (based on the historical and real-time flow statistical data of the block full mode traffic flow fusion data set), according to the initial path selection probability generated in step 301, the total demand is proportionally allocated to each feasible path to form an initial flow allocation scheme; The scheme clearly records the allocated flow of each OD pair on each feasible path, the path passing through the road section and the corresponding interaction range; combined with the open block traffic control demand, define the types of real-time traffic state parameters to be calculated, including the path flow density (the number of traffic entities per unit length) in each interaction range, the average speed, the traffic flow queue length, the flow proportion of different traffic modes, etc. These parameters need to accurately reflect the traffic flow running state and the degree of mutual interference; based on the initial flow allocation scheme, extract the allocated flow data of all passing paths in each interaction range, combined with the road attribute information such as the length of the path in the range and the number of lanes, calculate the real-time traffic state parameters of each interaction range; for example, calculate the flow density according to the allocated flow and the road section area, calculate the average speed according to the correlation between the flow and the interference parameters, and form the real-time traffic state parameter set of each interaction range; check whether the real-time traffic state parameters of each interaction range are complete, if there are missing parameters (such as the queue length cannot be calculated due to too small flow in a range), combine the parameter characteristics of adjacent ranges and the actual road conditions to make reasonable supplement; at the same time, eliminate the abnormal parameters obviously beyond the reasonable range, ensure that the parameters can truly reflect the traffic running state under the initial allocation scheme.

[0079] Step 303, iteratively update the parameters of the Gaussian mixture model with the mutual interference parameters as actual observations and real-time traffic state parameters as model predictions, and recalculate the generalized travel cost of each path based on the updated model, including: taking the extracted mutual interference parameters (speed interference intensity, density interweaving coefficient) as actual observations and the real-time traffic state parameters calculated in step 302 as the predictions of the Gaussian mixture model; by comparing the differences between the two, determine the update direction of the model parameters, if the prediction value and the observation value deviate greatly, adjust the mean, covariance and other parameters of the model, so that the model prediction result is closer to the actual traffic interference state; based on the deviation between the prediction value and the observation value, update the parameters of the Gaussian mixture model using iterative optimization logic; during the update process, always aim to reduce the prediction deviation, gradually adjust the mean, covariance and weight of each cluster to ensure that the updated model can more accurately depict the mutual interference rules of multi-mode traffic flow; record the specific content of each parameter update and the deviation change; define the constituent dimensions of the generalized travel cost, including the actual travel time of the path, the delay cost caused by traffic interference, the travel safety cost (based on interference parameter quantification), etc.; based on the updated Gaussian mixture model parameters, combined with the distribution flow of each path and the interference state of the interaction range, calculate the generalized travel cost of each feasible path, the longer the travel time, the more intense the interference, the higher the safety risk, the higher the generalized travel cost of the path; compare the changes of the generalized travel cost of the same path before and after the update to ensure that the cost change trend is consistent with the model parameter update direction (such as the model predicting that the interference is increasing, the corresponding path cost should also increase); if there is an abnormal change in the cost of a path, recheck the parameter update process and cost calculation logic and correct the deviation.

[0080] Step 304, based on the updated generalized travel cost, dynamically adjust the selection probability of each path and redistribute the traffic flow to generate an updated traffic distribution scheme, including: taking the generalized travel cost as the core basis to adjust the selection probability of each OD pair on the feasible path, the lower the generalized travel cost, the higher the travel efficiency and the smaller the interference, the corresponding selection probability is adjusted upward; the higher the cost, the lower the selection probability; after adjustment, it is still necessary to ensure that the sum of the selection probability of all feasible paths of the same OD pair is 1, forming an updated path selection probability set; based on the adjusted path selection probability, redistribute the total traffic demand of each OD pair to the corresponding feasible path to generate an updated traffic distribution scheme; the new scheme needs to focus on optimizing the paths where the traffic is excessively concentrated in the early distribution, causing excessive interference, and diverting part of the traffic to paths with lower cost and less interference to achieve balanced distribution of road network traffic; compare the differences between the updated traffic distribution scheme and the previous scheme, including the change of the distribution flow of each path and the change of the flow of each interaction range, record the specific values and distribution areas of the differences; if the traffic in a certain area changes too much, check the probability adjustment logic to avoid new traffic congestion caused by excessive adjustment.

[0081] Step 305, steps 303 to step 304 are repeatedly executed until the traffic allocation scheme generated by adjacent two iterations has a difference less than a preset convergence threshold, and the obtained scheme is the final traffic allocation scheme, and the real-time traffic allocation proportion of each path is output, including: combining the accuracy requirement of open street traffic control, the convergence threshold is preset, which is the maximum allowed difference value between the traffic allocation schemes generated by adjacent two iterations (such as the maximum value of the change rate of the traffic of each path is not more than 5%); the threshold needs to consider the optimization accuracy and iteration efficiency, so as to avoid that the threshold is too small to cause too many iterations and low efficiency, or the threshold is too large to cause that the scheme does not reach the optimal state; steps 303 to step 304 are repeatedly executed, and after each iteration is completed, the difference value (such as the absolute difference value and the change rate of the traffic of each corresponding path) between the traffic allocation schemes of adjacent two times is calculated and compared with the preset convergence threshold; if the difference value is less than or equal to the convergence threshold, it indicates that the traffic allocation scheme has tended to be stable, the optimization requirement is met, and the iteration is stopped; if the difference value is greater than the convergence threshold, the iteration is continued; the traffic allocation scheme obtained after the iteration converges is taken as the final scheme, and the allocation traffic, the passing road section, the corresponding interaction range and the traffic state parameter of each OD pair to each feasible path in the scheme are recorded; the feasibility of the final scheme is checked to ensure that the scheme meets the road restriction requirement, the traffic distribution is balanced, and there is no excessive congestion risk; based on the final traffic allocation scheme, the real-time traffic allocation proportion (the percentage of the allocation traffic of a path to the total demand of the OD pair) of each OD pair to each feasible path is calculated; the allocation proportion is arranged according to the type of OD pair, the region to which the path belongs and the like, to form a structured output report, which can directly support the actual control work such as traffic guidance and signal timing optimization.

[0082] In this embodiment, the model parameters and the path selection probability are updated through multiple iterations, so that the traffic allocation scheme can respond to the change of the traffic flow interference state in real time, solves the problem that the traditional static allocation scheme is rigid and cannot adapt to the tidal passenger flow and real-time interference, and greatly improves the flexibility and adaptability of the allocation scheme; the path selection probability is adjusted based on the generalized travel cost, the excessively concentrated traffic is distributed to the low-interference and high-efficiency path, the load of the open street branch network is effectively balanced, the mixed congestion of motor vehicles and slow traffic is reduced, and in particular, the travel efficiency and safety of the early morning peak school slow passenger flow can be guaranteed, and the overall road network operation quality is improved; the Gaussian mixture model is iteratively optimized by taking the actual mutual interference parameters as observation values, so that the model can accurately depict the operation law of the multi-mode traffic flow in the open street, the path selection probability and the traffic allocation scheme calculated based on the model are more in line with the actual scene, the deviation caused by the empirical allocation is avoided, and the reliability of the scheme is improved.

[0083] As shown in Figure 2 The embodiment of the present application also provides an open street traffic flow intelligent allocation system, which comprises:

[0084] The acquisition module is configured to collect multi-source heterogeneous traffic data of an open block, convert position and trajectory information of traffic entities contained in the traffic data into discrete spatial point cloud data, construct a virtual analysis plane based on spatial distribution of the point cloud data and road topological structure of the open block, perform spatio-temporal alignment and fusion processing on the point cloud data, and establish an associated mapping relationship of motor vehicle and non-motor vehicle flow data on the virtual analysis plane to generate a block-wide traffic flow fusion data set.

[0085] The processing module is configured to analyze travel time distribution characteristics and travel mode composition characteristics of short-distance OD pairs on the virtual analysis plane based on the block-wide traffic flow fusion data set, identify a walking and non-motor vehicle dominant feature of school commuting passenger flow in an early morning peak period, determine a key analysis region for a branch network on the virtual analysis plane according to the dominant feature, and determine a basic triangular unit in the key analysis region; process the basic triangular unit to define an interaction range of multi-mode traffic flow; map the point cloud data on the virtual analysis plane to a corresponding actual road network region according to the interaction range to extract a mutual interference parameter of multi-mode traffic flow in the branch network, and generate an initial parameter set of a Gaussian mixture model.

[0086] The distribution module is configured to perform iterative optimization distribution on the open block OD pairs based on the initial parameter set of the Gaussian mixture model, dynamically adjust path selection probability, obtain a final flow distribution scheme, and output real-time flow distribution proportions of each path.

[0087] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the scope of protection of the present application.

Claims

1. An intelligent method for distributing traffic flow in an open street block, characterized in that, The method comprises: Collecting multi-source heterogeneous traffic data of an open street, converting the position and trajectory information of the traffic entities contained in the traffic data into discrete spatial point cloud data; based on the spatial distribution of the point cloud data and the road topology structure of the open street, a virtual analysis plane is constructed; the point cloud data is processed by spatio-temporal alignment and fusion, and the correlation mapping relationship between motor vehicle and non-motor vehicle flow data is established on the virtual analysis plane to generate a street full-mode traffic flow fusion data set; Based on the street full-mode traffic flow fusion data set, the travel time distribution characteristics and travel mode composition characteristics of the short-distance OD pair are analyzed on the virtual analysis plane, and the walking and non-motor vehicle dominant characteristics of the school commuting passenger flow in the morning peak period are identified; according to the dominant characteristics, the key analysis area is determined for the branch network on the virtual analysis plane, and the basic triangular unit is determined in the key analysis area; the basic triangular unit is processed to define the interaction range of multi-mode traffic flow; according to the interaction range, the point cloud data on the virtual analysis plane is mapped to the corresponding actual road network area to extract the mutual interference parameters of multi-mode traffic flow in the branch network, and an initial parameter set of Gaussian mixture model is generated; Based on the initial parameter set of the Gaussian mixture model, the OD pair of the open street is iteratively optimized and distributed, the path selection probability is dynamically adjusted, the final flow distribution scheme is obtained, and the real-time flow distribution proportion of each path is output.

2. The method for intelligent distribution of traffic flow in open street according to claim 1, characterized in that, Based on the spatial distribution of the point cloud data and the road topology structure of the open street, a virtual analysis plane is constructed, comprising: According to the discrete spatial point cloud data, the two-dimensional spatial range covered thereby is extracted as the basic geographical boundary of the virtual plane; According to the actual road network topology structure of the open street, the interior of the basic geographical boundary is divided into analysis units corresponding to branches, intersections and interest point areas; Each analysis unit is assigned a two-dimensional coordinate grid, and all point cloud data are mapped to the corresponding grid nodes according to their spatial coordinates to form a virtual analysis plane.

3. The open block traffic flow intelligent distribution method of claim 2, wherein, The point cloud data is processed by spatio-temporal alignment and fusion, and the correlation mapping relationship between motor vehicle and non-motor vehicle flow data is established on the virtual analysis plane to generate a street full-mode traffic flow fusion data set, comprising: All sources of point cloud data are uniformly converted to the same standard timestamp to obtain point cloud data that has completed spatio-temporal alignment; Based on the virtual gridded plane, the point cloud data from different detection methods and which has completed spatio-temporal alignment is classified and mapped to the motor vehicle and non-motor vehicle dynamic data layers according to the corresponding traffic entity types; Based on the virtual gridded plane, the point cloud data from different detection methods and which has completed spatio-temporal alignment is classified and mapped to the motor vehicle and non-motor vehicle dynamic data layers according to the corresponding traffic entity types; According to the correlation mapping relationship, the static basic road network layer containing road attributes and restriction information is superimposed and fused to generate a street full-mode traffic flow fusion data set containing time, space, mode, flow and mutual relationship.

4. The method for intelligent distribution of traffic flow in open street according to claim 3, characterized in that, Based on the street-wide traffic flow fusion dataset, the travel time distribution characteristics and travel mode composition characteristics of short-distance OD pairs are analyzed on the virtual analysis plane, the walking and non-motor vehicle dominant characteristics of school commuting passenger flow in the morning peak period are identified, including: Based on the street-wide traffic flow fusion dataset, the short-distance OD pairs with travel distance less than a preset threshold are screened out on the virtual analysis plane, and the corresponding point cloud trajectory data is extracted; The point cloud trajectory data is processed, the average travel time and travel time variance of each short-distance OD pair in different time periods are counted, and a travel time distribution characteristic set is formed; Based on the travel time distribution characteristic set, the analysis period is focused on the morning peak period, and based on the short-distance OD pairs, the number of trips and the proportion of each OD pair through walking, non-motor vehicle and motor vehicle in the morning peak period are counted, and a morning peak travel mode composition characteristic set is formed; Combined with the preset school interest point location information, the OD pairs with school as the core and showing an aggregated flow direction on the virtual analysis plane are screened out from the morning peak travel mode composition characteristic set to form a school commuting candidate OD set; For the school commuting candidate OD set, according to the corresponding morning peak travel mode composition characteristics, the total proportion of walking and non-motor vehicle travel modes of each OD pair is calculated; if the total proportion exceeds a preset dominant threshold, it is determined that the passenger flow of the OD pair has walking and non-motor vehicle dominant characteristics.

5. The method for intelligent distribution of traffic flow in open street according to claim 4, characterized in that, According to the dominant characteristics, the key analysis areas are determined for the branch network on the virtual analysis plane, and the basic triangular units are determined in the key analysis areas; The basic triangular units are processed to define the interaction range of multi-mode traffic flow, including: According to the determined OD pairs with walking and non-motor vehicle dominant characteristics, the origin and destination points and the passing path of the OD pairs are located on the virtual analysis plane, the continuous road section area bearing the main flow and composed of branches is determined as the key analysis area; In each of the key analysis areas, a corresponding intersection is selected as the core point, and the adjacent intersections in the three key branch directions directly connected to the intersection are selected as the vertices, and the basic triangular units are connected on the virtual analysis plane; The vertices of the basic triangular units and the core point are taken as a set of spatial point sets; the spatial point sets are processed to calculate and generate a minimum convex polygon area containing all points of the spatial point sets; The minimum convex polygon area is defined as the interaction range of multi-mode traffic flow in the key analysis area, where the speed changes, trajectory interweaving and flow competition actually occur.

6. The open block traffic flow intelligent distribution method of claim 5, wherein, According to the interaction range, the point cloud data on the virtual analysis plane is mapped to the corresponding actual road network area to extract the mutual interference parameters of multi-mode traffic flow in the branch network, generate an initial parameter set of the Gaussian mixture model, including: On the virtual analysis plane, the actual branch network profile covered by the interaction range is extracted as a polygon road area; The polygon road area is triangulated into a plurality of non-overlapping triangular sub-units on the virtual analysis plane, and the area of each triangular sub-unit is calculated; According to the street full-mode traffic flow fusion data set, all point cloud data falling within the interaction range is obtained, and according to the spatial coordinates, the point cloud data is mapped to the corresponding triangular subunit by calculating the inclusion relationship with each triangular subunit; Based on the mapping relationship, the number, average speed and speed variance of different traffic mode point clouds in each triangular subunit are counted, and the speed interference intensity and density interlacing coefficient of multi-mode traffic flow in the triangular subunit are calculated as the mutual interference parameters by combining the area of the triangular subunit; According to the statistical distribution characteristics of the mutual interference parameters extracted from all triangular subunits, the number of Gaussian mixture model clusters is determined, and the mean and covariance of the interference parameters of each subunit are used as the initial values to generate the initial parameter set of the Gaussian mixture model.

7. The open street traffic flow intelligent distribution method according to claim 6, wherein, Based on the initial parameter set of the Gaussian mixture model, the open street OD pair is iteratively optimized and distributed, the path selection probability is dynamically adjusted, the final traffic distribution scheme is obtained, and the real-time traffic distribution proportion of each path is output, including: Based on the initial parameter set of the Gaussian mixture model, for each open street OD pair on the virtual analysis plane, the initial path selection probability of all feasible paths in the interaction range is generated; According to the initial path selection probability, the total traffic demand of the OD pair is distributed to each feasible path to form an initial traffic distribution scheme; the real-time traffic state parameters of each interaction range under the initial traffic distribution scheme are calculated; Taking the mutual interference parameters as the actual observation values and the real-time traffic state parameters as the model prediction values, the parameters of the Gaussian mixture model are iteratively updated, and the generalized travel cost of each path is recalculated based on the updated model; Based on the updated generalized travel cost, the selection probability of each path is dynamically adjusted, and the traffic flow is redistributed to generate an updated traffic distribution scheme, which is repeatedly executed until the difference between the traffic distribution schemes generated by adjacent two iterations is less than a preset convergence threshold, and the obtained scheme is the final traffic distribution scheme, and the real-time traffic distribution proportion of each path is output.

8. An open block traffic flow intelligent distribution system, the system implements the method as claimed in any one of claims 1 to 7, characterized in that, Including: An acquisition module is configured to collect multi-source heterogeneous traffic data of an open street, and convert position and trajectory information of traffic entities contained in the traffic data into discrete spatial point cloud data; Based on the spatial distribution of the point cloud data and the road topological structure of the open street, a virtual analysis plane is constructed; the point cloud data is processed by time and space alignment and fusion, and an associated mapping relationship between motor vehicle and non-motor vehicle flow data is established on the virtual analysis plane to generate a street full-mode traffic flow fusion data set; A processing module is configured to analyze the travel time distribution characteristics and travel mode composition characteristics of short-distance OD pairs on the virtual analysis plane based on the street full-mode traffic flow fusion data set, identify the walking and non-motor vehicle dominant characteristics of school commuting passenger flow in the morning peak period, determine a key analysis area for a branch network on the virtual analysis plane according to the dominant characteristics, and determine a basic triangular unit in the key analysis area. The basic triangular unit is processed to define the interaction range of multi-modal traffic flow; according to the interaction range, the point cloud data on the virtual analysis plane is mapped to the corresponding actual road network area to extract the mutual interference parameters of multi-modal traffic flow in the branch network, and an initial parameter set of the Gaussian mixture model is generated; The distribution module is used for iteratively optimizing the distribution of the open block OD pairs based on the initial parameter set of the Gaussian mixture model, dynamically adjusting the path selection probability, obtaining the final traffic distribution scheme, and outputting the real-time traffic distribution proportion of each path.

Images