Intersections delay analysis method and device based on small sample rate floating car trajectory

Abstract

The application provides a kind of intersection delay analysis method and device based on small sample rate floating car trajectory, belong to traffic engineering technical field, including obtaining the stop line of target intersection entrance lane;According to the directional distance information of multiple sampling points in trajectory data, determine the trajectory data of multiple vehicles;Construct space-time trajectory information, analyze the geometric characteristic parameters of space-time trajectory information;Multiple space-time trajectory information is superimposed to the same coordinate system, and multiple vehicle space-time trajectory graph is formed;According to the traffic wave information of multiple vehicle space-time trajectory graph, obtain signal interval information;According to geometric characteristic parameters and signal interval information, determine the delay index information of entrance lane;According to the delay index information of each entrance lane, total delay index information is obtained.The application can stably and accurately analyze the total delay index information of target intersection under the condition of sparse sample and without external data, which is convenient for large-scale deployment in actual road network.

Description

Technical Field

[0001] This application relates to the field of traffic engineering technology, and in particular to a method and apparatus for intersection delay analysis based on floating car trajectories with a small sample rate. Background Technology

[0002] Intersections are key nodes in urban traffic networks, and their traffic efficiency directly determines the operational quality of the entire network. Delay indicators are important indicators for evaluating the effectiveness of traffic signal control and the operational status of intersection networks. Traditional delay analysis methods rely on fixed-point detection devices such as geomagnetic detectors, which have limitations such as high deployment costs, limited detection cross-sections, and susceptibility to weather conditions.

[0003] With the development of mobile internet and intelligent connected vehicle technologies, floating car trajectory data has become an important data source for traffic condition perception and road network operation quality assessment, offering advantages such as wide coverage, all-weather collection, and the ability to simultaneously acquire multi-dimensional information. Currently, existing intersection delay analysis methods typically rely on full vehicle trajectories and high-frequency continuous trajectories, requiring high-quality trajectory data and incorporating external data such as signal timing, high-precision maps, and detectors to jointly analyze traffic delays. However, in real-world road networks, the available floating car trajectory data is often characterized by low penetration, low sampling frequency, and a small number of samples per cycle, failing to meet the stringent requirements for data volume and quality. This is especially true for non-perceptive intersections without deployed detection equipment, where the lack of external data makes it difficult to directly apply traditional methods to obtain reliable delay estimation results.

[0004] Therefore, how to achieve accurate and efficient assessment of intersection delays without relying on external auxiliary data under low-penetration floating car trajectory conditions has become a key problem that urgently needs to be solved in the field of intelligent transportation. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for intersection delay analysis based on floating car trajectories with a small sample rate, in order to solve the problem that traditional analysis methods rely on high-quality trajectory data and external data, making large-scale deployment difficult.

[0006] To achieve the above objectives, this application provides a method for intersection delay analysis based on floating car trajectories with a small sample rate, including: obtaining the stop line of at least one approach lane of the target intersection; Based on the stop line and trajectory data of multiple vehicles passing through the approach lane, the directed distance information of multiple sampling points in the trajectory data is determined; the directed distance information includes the distance and direction information from the sampling point to the stop line of the approach lane. Based on the directed distance information, the spatiotemporal trajectory information corresponding to the vehicle is constructed, and the geometric feature parameters of the spatiotemporal trajectory information are analyzed; Multiple spatiotemporal trajectory information is superimposed onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map; Based on the geometric feature parameters of each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory diagram, traffic wave information of the multi-vehicle spatiotemporal trajectory diagram is obtained, and based on the traffic wave information, the signal interval information of the approach lane is obtained; Based on the geometric feature parameters and the signal interval information, the delay index information of the approach lane is determined. The delay index information includes one or more of the following: parking delay information, queuing delay information, fixed delay information, and operational delay information. Based on the delay index information of each of the aforementioned approach lanes, the total delay index information of the target intersection is obtained. The total delay index information includes one or more of the following: total parking delay information, total queuing delay information, total fixed delay information, and total operational delay information.

[0007] Optionally, obtaining the stop line of at least one approach lane at the target intersection includes: The historical trajectory data of multiple vehicles passing through the target intersection within at least one historical sampling period are preprocessed to obtain preprocessed historical trajectory data. Multiple sampling points with instantaneous speeds below a first preset threshold in each preprocessed historical trajectory data are projected onto the center line of the entrance lane of the corresponding target intersection to form a set of low-speed projection points. Density estimation is performed on the set of low-velocity projection points to obtain the spatial distribution density function corresponding to the set of low-velocity projection points. Extract the maximum points from the spatial distribution density function to form a set of candidate peak points; The stop line of the approach lane corresponding to the target intersection is determined based on the set of candidate peak points.

[0008] Optionally, it also includes: Acquire trajectory data of multiple vehicles passing through the entrance lane within at least one new sampling period following the historical sampling period; Based on the trajectory data within the new sampling period, stop line estimation information is generated; Based on the stop line estimation information, the stop line corresponding to the entrance lane is calibrated.

[0009] Optionally, based on the stop line and trajectory data of multiple vehicles passing through the approach lane, the directed distance information of multiple sampling points in the trajectory data is determined, including: Based on the topological relationship information of the target intersection, calculate the matching degree score from the sampling point to each approach lane of the target intersection, and obtain the first approach lane with the highest matching degree score corresponding to the sampling point; The distance from the projection point of the sampling point on the center line of the first inlet channel to the corresponding stop line of the first inlet channel is determined as the distance information of the sampling point. Based on the upstream and downstream positional relationship of the sampling point relative to the stop line of the first inlet channel, the directional information corresponding to the sampling point is determined.

[0010] Optionally, spatiotemporal trajectory information corresponding to the vehicle is constructed based on the directed distance information, and the geometric feature parameters of the spatiotemporal trajectory information are analyzed, including: Based on the topological relationship information of the target intersection, sampling points belonging to the same vehicle and the same approach lane are associated as the same trip; Based on the time information of sampling points belonging to the same trip and the directed distance information, the spatiotemporal trajectory information is constructed. The spatiotemporal trajectory information contains multiple trajectory points, each of which corresponds one-to-one with a sampling point. Each trajectory point is composed of the time information and directed distance information of the corresponding sampling point. The geometric feature parameters of each trajectory point in the spatiotemporal trajectory information are analyzed. The geometric feature parameters include the multi-scale slope information, curvature information and slope consistency information of the trajectory point. The multi-scale slope information includes multiple local slope information of the trajectory point. The local slope information is the slope between the trajectory point and another trajectory point.

[0011] Optionally, traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained based on the geometric feature parameters of each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, and signal interval information of the approach lane is obtained based on the traffic wave information, including: Based on the geometric feature parameters of the spatiotemporal trajectory information, a set of parking trajectory segments of the spatiotemporal trajectory information is formed; the set of parking trajectory segments includes at least one parking trajectory segment. Based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained, and the traffic wave information includes parking wave function and starting wave function; Based on the parking wave function and the starting wave function, the signal interval information corresponding to the approach lane is obtained.

[0012] Optionally, the delay index information of the approach lane is determined based on the geometric feature parameters and the signal interval information, including: The set of parking trajectory segments is analyzed based on the signal interval information to determine the effective parking trajectory segments, which are parking trajectory segments generated due to signal waiting. Based on the geometric feature parameters of the spatiotemporal trajectory information, calculate the line crossing time points of the spatiotemporal trajectory information; Based on the signal interval information, the effective parking trajectory segment, and the crossing time point, the delay index information corresponding to the approach lane is obtained.

[0013] Optionally, a set of parking trajectory segments of the spatiotemporal trajectory information is formed based on the geometric feature parameters of the spatiotemporal trajectory information, including: Based on the geometric feature parameters of each trajectory point in the spatiotemporal trajectory information, a feature vector of the trajectory point is constructed; The feature vector is analyzed using a first classification model to determine the motion state information of the trajectory point; Merge the trajectory segments between multiple adjacent trajectory points whose motion state information is in the parking state to generate a parking trajectory segment; Based on the parking trajectory segments, the parking trajectory segment set is formed.

[0014] Optionally, traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, including: Based on the set of parking trajectory segments of multiple spatiotemporal trajectory information, the trajectory points at the start time of the first parking trajectory segment corresponding to the multiple spatiotemporal trajectory information are obtained to form a first trajectory point set; Obtain the trajectory points at the termination time of the last parking trajectory segment corresponding to multiple spatiotemporal trajectory information to form a second trajectory point set; The parking wave function is obtained based on the first set of trajectory points, and the starting wave function is obtained based on the second set of trajectory points.

[0015] Optionally, traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, including: Based on the multiple sets of parking trajectory segments containing the aforementioned spatiotemporal trajectory information, the start time trajectory points and end time trajectory points of all parking trajectory segments are obtained to form multiple sets of start and stop points. The multiple sets of start and stop points are divided into multiple candidate signal periods according to a preset time interval condition. The starting time trajectory points of each parking trajectory segment in each candidate signal period constitute the third trajectory point set of the candidate signal period, and the ending time trajectory points of each parking trajectory segment constitute the fourth time trajectory point set of the candidate signal period. For each candidate signal period, the parking wave function is obtained based on the third trajectory point set corresponding to the candidate signal period, and the starting wave function is obtained based on the fourth trajectory point set corresponding to the candidate signal period.

[0016] Optionally, based on the parking wave function and the starting wave function, the signal interval information corresponding to the approach lane is obtained, including: Based on the parking wave function, the red light activation time corresponding to the multi-vehicle spatiotemporal trajectory map is obtained; Based on the start-up wave function, obtain the green light start time corresponding to the multi-vehicle spatiotemporal trajectory map; Based on the red light start time and green light start time of the same multi-vehicle spatiotemporal trajectory map, determine the red light interval information and red light duration information corresponding to the entrance lane; Based on the red light or green light turn-on time of the multi-vehicle spatiotemporal trajectory map of adjacent sampling periods, the signal cycle information corresponding to the entrance lane is determined; Based on the signal cycle information and the red light duration information, determine the green light ratio information corresponding to the approach lane; The signal interval information includes the signal cycle information, the red light interval information, the red light duration information, and the green light ratio information.

[0017] Optionally, the method further includes: Obtain at least one historical trajectory data of the inlet channel within a historical sampling period that is in the same time period as the current sampling period; Based on the directed distance information of each trajectory point in the historical trajectory data and the directed distance information of each trajectory point in the trajectory data within the current sampling period, the historical trajectory data is aligned with the trajectory data within the current sampling period; An enhanced trajectory data set is formed based on the aligned historical trajectory data and the trajectory data; The enhanced parking wave function and enhanced start wave function of the enhanced trajectory data set are obtained, and the enhanced signal interval information is obtained based on the enhanced parking wave function and the enhanced start wave function. The enhanced signal interval information is used as the signal interval information of the approach lane.

[0018] Optionally, the method further includes: Multiple enhanced trajectory data sets are acquired, wherein the multiple enhanced trajectory data sets are formed by aligning historical trajectory data with trajectory data within the current sampling period using different statistical windows, and the statistical window includes one or more historical sampling periods; Based on multiple sets of enhanced trajectory data, multiple enhanced signal interval information is obtained, and the enhanced signal interval information located at the median is used as the signal interval information of the approach lane.

[0019] Optionally, the set of parking trajectory segments is analyzed based on the signal interval information to determine valid parking trajectory segments, including: Obtain the overlap between the time interval information of the parking trajectory segment and the red light interval information; The parking trajectory segment that meets the preset overlap condition is determined as the valid parking trajectory segment.

[0020] Optionally, calculating the line-crossing time points of the spatiotemporal trajectory information based on the geometric feature parameters of the spatiotemporal trajectory information includes: The first upstream trajectory point and the first downstream trajectory point are obtained from the spatiotemporal trajectory information; the first upstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located upstream of the stop line, and the first downstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located downstream of the stop line. Based on the geometric feature parameters of the first upstream trajectory point and the first downstream trajectory point, either a uniform interpolation model or an acceleration / deceleration interpolation model is selected to calculate the time point of the spatiotemporal trajectory information.

[0021] Optionally, based on the signal interval information, the effective parking trajectory segment, and the crossing time point, the delay index information corresponding to the approach lane is obtained, including: Based on the effective parking trajectory segment, calculate the single-vehicle parking delay information of the spatiotemporal trajectory information, and based on the single-vehicle parking delay information of each spatiotemporal trajectory information, calculate the parking delay information of the entrance lane; Based on the effective parking trajectory segment and the crossing time point, calculate the single-vehicle queuing delay information of the spatiotemporal trajectory information, and calculate the queuing delay information of the entrance lane based on the single-vehicle queuing delay information of each spatiotemporal trajectory information. Based on the signal interval information and the effective parking trajectory segment, calculate the fixed delay information of a single vehicle in the spatiotemporal trajectory information, and calculate the fixed delay information of the approach lane based on the fixed delay information of a single vehicle in each spatiotemporal trajectory information. Based on the single-vehicle queuing delay information and the single-vehicle fixed delay information, the single-vehicle operation delay information of the spatiotemporal trajectory information is calculated, and based on the single-vehicle operation delay information of each of the spatiotemporal trajectory information, the operation delay information of the entrance lane is calculated.

[0022] This application also provides an intersection delay analysis device based on small sample rate floating car trajectories, including: The stop line acquisition module is used to acquire the stop line of at least one approach lane of the target intersection; A directed distance acquisition module is used to determine directed distance information of multiple sampling points in the trajectory data based on the stop line and trajectory data of multiple vehicles passing through the approach lane; the directed distance information includes the distance information and direction information from the sampling point to the stop line of the approach lane; The spatiotemporal trajectory construction module is used to construct the spatiotemporal trajectory information corresponding to the vehicle based on the directed distance information, and to analyze the geometric feature parameters of the spatiotemporal trajectory information; The spatiotemporal trajectory overlay module is used to overlay multiple spatiotemporal trajectory information onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map; The signal interval acquisition module is used to obtain traffic wave information of the multi-vehicle spatiotemporal trajectory map based on the geometric feature parameters of each spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, and to obtain the signal interval information of the approach lane based on the traffic wave information. The delay index analysis module is used to determine the delay index information of the approach lane based on the geometric feature parameters and the signal interval information. The delay index information includes one or more of the following: parking delay information, queuing delay information, fixed delay information, and operation delay information. The total delay index analysis module is used to obtain the total delay index information of the target intersection based on the delay index information of each of the approach lanes. The total delay index information includes one or more of the following: total parking delay information, total queuing delay information, total fixed delay information, and total operational delay information.

[0023] This application also provides an analysis device, including a memory, a processor, and a computer program stored in the memory and executable on the processor; wherein, when the processor executes the program, it implements the above-described method for analyzing intersection delays based on small sample rate floating car trajectories.

[0024] This application also provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps in the above-described method for intersection delay analysis based on small sample rate floating car trajectories.

[0025] This application also provides a computer program product, which includes computer instructions that, when executed by a processor, implement the steps in the above-described method for analyzing intersection delays based on small sample rate floating car trajectories.

[0026] At least one of the above technical solutions in the specific embodiments of this application has the following beneficial effects: The method and apparatus for intersection delay analysis based on floating car trajectories with a small sample rate provided in this application converts the original trajectory data of floating cars obtained through a satellite positioning system into a sequence of trajectory points measured along the approach lane with the stop line as the zero point by constructing spatiotemporal trajectory information. This avoids directly judging stopping, starting, and crossing events on the original planar trajectory, and enables sparse trajectory data of different vehicles and different cycles to be compared under the same coordinate system, providing a stable basis for delay calculation under small sample conditions. Furthermore, this application uses the geometric feature parameters of spatiotemporal trajectory information to infer the signal interval information of the target intersection. It can still stably and accurately analyze the total delay index information of the target intersection without relying on external data such as signal timing or roadside detection equipment, which is convenient for large-scale deployment in actual traffic networks. Attached Figure Description

[0027] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is one of the flowcharts illustrating an intersection delay analysis method based on small sample rate floating car trajectories provided in an embodiment of this application; Figure 2 This is a second flowchart illustrating an embodiment of the intersection delay analysis method based on small sample rate floating car trajectories provided in this application. Figure 3 The third flowchart illustrates the intersection delay analysis method based on small sample rate floating car trajectories provided in one embodiment of this application. Figure 4 The fourth flowchart illustrates a method for intersection delay analysis based on floating car trajectories with a small sample rate, provided in one embodiment of this application. Figure 5 The fifth flowchart illustrates a method for analyzing intersection delays based on floating car trajectories with a small sample rate, provided in one embodiment of this application. Figure 6 This is the sixth flowchart illustrating an embodiment of the intersection delay analysis method based on small sample rate floating car trajectories provided in this application. Figure 7 This is a schematic diagram of spatiotemporal trajectory information provided in one embodiment of this application; Figure 8 This is a schematic diagram of a multi-vehicle spatiotemporal trajectory map provided in one embodiment of this application; Figure 9 This is a schematic diagram of a multi-vehicle spatiotemporal trajectory diagram with signal interval information provided in one embodiment of this application; Figure 10 This is a schematic diagram illustrating the calculation of the time point when crossing the line according to one embodiment of this application; Figure 11 This is a schematic diagram of the structure of an intersection delay analysis device based on a small sample rate floating car trajectory provided in one embodiment of this application. Detailed Implementation

[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Exemplary embodiments of this application will be described in more detail with reference to the accompanying drawings.

[0029] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for intersection delay analysis based on floating car trajectories with a small sample rate, provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step 101: Obtain the stop line of at least one approach lane of the target intersection.

[0030] In one embodiment of the present invention, such as Figure 2 As shown, step 101 includes: Step 1011: Preprocess the historical trajectory data of multiple vehicles passing through the target intersection within at least one historical sampling period to obtain preprocessed historical trajectory data.

[0031] It should be noted that the vehicles mentioned in this embodiment refer to floating cars equipped with satellite positioning devices such as GPS or Beidou, such as urban operating vehicles (taxis, ride-hailing vehicles, buses, and logistics trucks, etc.). Their trajectory data refers to the sequence of sampling points collected and transmitted back by the floating car at fixed time intervals (such as a few seconds to tens of seconds) during its operation. Each sampling point usually includes information such as latitude and longitude position, direction angle, timestamp, and vehicle identification.

[0032] In this embodiment, the scope of the target intersection is the road segment starting from a certain distance (which can be set to 400 meters) upstream of the stop line of any approach lane of the target intersection, along the center line of the approach lane, passing the stop line, crossing the target intersection area, and ending at the extension line of the stop line of the exit lane in that direction of travel. Among them, the section between a certain distance upstream of the stop line of the approach lane and the stop line is the main section where traffic delays occur, and it is also the main section analyzed in this embodiment.

[0033] Specifically, historical trajectory data of multiple vehicles passing through the target intersection within a certain time range, including at least one sampling period, is collected and preprocessed. First, high-error sampling points with a horizontal dilution of precision (DIP) greater than 5 are removed. The DIP is a core indicator for measuring the positioning accuracy of global navigation satellite systems (such as GPS and BeiDou). The higher the DIP value, the lower the positioning accuracy. At the same time, drift sampling points with speed changes greater than 200 km / h between adjacent sampling points are removed. Then, noise is smoothed through Kalman filtering to obtain the preprocessed historical trajectory data.

[0034] Step 1012: Project multiple sampling points in each preprocessed historical trajectory data whose instantaneous speed is lower than the first preset threshold onto the center line of the corresponding entrance lane of the target intersection to form a set of low-speed projection points.

[0035] Specifically, the instantaneous velocity of each sampling point in each historical trajectory data is first calculated: ; in, For the first The planar coordinates of each sampling point in the projected coordinate system of the target intersection For the first Sampling time of each sampling point For the first +1 sampling point in the plane coordinates of the target intersection projection coordinate system For the first +1 sampling time, For the first Instantaneous velocity at each sampling point, in meters per second.

[0036] Extraction speed is lower than the first preset threshold The low-speed sampling points are obtained and projected onto the corresponding inlet centerline to obtain the set of low-speed projection points along the centerline. ,in, For the first The distance from the projection point of each low-speed sampling point to the geometric center of the target intersection is a one-dimensional position, expressed in meters.

[0037] Step 1013: Perform density estimation on the set of low-velocity projection points to obtain the spatial distribution density function corresponding to the set of low-velocity projection points.

[0038] Specifically, for the set of projection points of low-speed points The purpose of density estimation is to transform discrete points into a continuous density curve. In this embodiment, an adaptive bandwidth kernel density estimation algorithm is used for density estimation. First, a trial density estimation is performed with a fixed bandwidth. Performing a simple kernel density estimate on the set of low-velocity projection points yields the trial density: ; Where x is the one-dimensional positional variable along the centerline of the corresponding inlet lane, and n is the total number of low-speed projection points. For the first The one-dimensional position of a low-speed projection point Let x be the trial density value. For Gaussian kernel function: ; in, The standardized distance variable is the Gaussian kernel function.

[0039] Initial fixed bandwidth Determined according to Silverman's rule of thumb: ; in, For the set of low-speed projection points The sample standard deviation is used to determine the bandwidth order of the mean square error when the sample is approximately unimodal, which serves as a benchmark for subsequent adaptive adjustment.

[0040] Then, adaptive bandwidth adjustment is performed, scaling the bandwidth of each low-speed projection point inversely according to the trial density at each point: ; in, To be assigned to the Adaptive bandwidth for low-speed projection points; To test the density value at this point; The geometric mean of the test density at all low-velocity projection points is used as the benchmark for dividing the high-density and low-density areas. , is the bandwidth sensitivity index, typically set to 0.5. It degrades to a fixed bandwidth at times. The larger the bandwidth, the more sensitive it is to local density. This step allows for adjustments in high-density areas (…). )make Maintain the spatial resolution of the main peak at the stop line; in low-density regions, make This expands the smoothing range and suppresses sparse pseudo-peaks.

[0041] Finally, adaptive density estimation is performed, and the density is re-estimated using the adaptive bandwidth at each point: ; in, Let be the spatial distribution density function of the low-velocity projection points along the centerline. The significant peak corresponds to the habitual parking section of the vehicle group.

[0042] Step 1014: Extract the maximum points in the spatial distribution density function to form a set of candidate peak points.

[0043] Specifically, the spatial distribution density function is first subjected to local maximum point detection to obtain the local maximum points that satisfy the local maximum condition. Then, remove the maxima that do not satisfy the significance condition, i.e., those that do not satisfy the significance condition. The maximum points are identified to eliminate insignificant maximum points caused by noise. The remaining maximum points are then used as candidate peak points to form a set of candidate peak points.

[0044] in, The one-dimensional location coordinates of the candidate peak points. , These are the first and second derivatives of the spatial density distribution function, respectively. and These are the mean and standard deviation of the spatial density distribution function, respectively.

[0045] Step 1015: Determine the stop line of the approach lane corresponding to the target intersection based on the set of candidate peak points.

[0046] Specifically, the candidate peak point set usually contains multiple candidate peak points, such as the main peak formed by the first row of cars parked at the stop line, the secondary peak formed by the upstream secondary queue (the tail of the queue that has not dissipated in the previous cycle), and isolated peaks formed by bus stops or roadside temporary parking. Since the queue always extends upstream from the stop line, the first row of cars parked is closest to the stop line and occurs most frequently. Therefore, the one-dimensional position closest to the center of the target intersection is selected from the candidate peak point set. The smallest candidate peak point is the stop line of the inlet.

[0047] In one embodiment of the present invention, step 101 further includes: Step 1016: Obtain trajectory data of multiple vehicles passing through the entrance lane within at least one new sampling period after the historical sampling period.

[0048] Step 1017: Based on the trajectory data within the new sampling period, generate stop line estimation information.

[0049] Specifically, in this embodiment, after a predetermined period of time, the stop lines of the target intersection are calibrated and updated online based on trajectory data from multiple new sampling periods following the historical sampling period. The trajectory data from each new sampling period, or the trajectory data from all new sampling periods, are grouped together, and the resulting multiple stop line estimation information is used to form a stop line estimation sequence. , where k is the sampling period number or group number.

[0050] Then, outlier estimates are removed using the median absolute deviation method. Estimates that satisfy the following formula are identified as outliers and removed from the calibration process: ; in, To estimate the median of the sequence for the stop line, For each estimated value and The median of the absolute values ​​of the differences, 1.4826, is the value that makes... The consistency constant is on the same scale as the standard deviation of the normal distribution, and 3 is the outlier determination factor. This formula is used to prevent erroneous estimates caused by abnormal situations such as accidents blocking roads or temporary traffic organization from affecting the calibration of individual windows.

[0051] Step 1018: Based on the stop line estimation information, calibrate the stop line corresponding to the inlet lane.

[0052] Specifically, based on the stop line estimates obtained through outlier detection, the original stop lines are calibrated online by sequentially applying exponentially weighted moving averages in chronological order: ; ; in, For the first The stop line position after the next update For the first The historical stop line position before the next update; It is an adaptive smoothing coefficient; The baseline smoothing coefficient is typically 0.2. The quality coefficient for this estimation is calculated by combining the estimated peak significance z with the total number of low-velocity projection points within the current sampling period or group. Joint normalization yields, where: ; in, The density corresponding to the estimated information for this stop line. and These represent the mean and standard deviation of the spatial density distribution function, respectively. The higher the peak significance z of the stop line estimation information, the greater the total number of low-velocity projection points. The more data available, the faster the updates; conversely, fewer data points will automatically slow down the updates, relying more on stop lines derived from historical trajectory data.

[0053] The method provided in this embodiment automatically obtains the target stop line by performing density estimation on the trajectory data within the historical sampling period of the target intersection and calibrating it with the trajectory data within the new sampling period. This eliminates the need for manual stop line calibration using high-precision maps, facilitating large-scale deployment.

[0054] In one embodiment of the present invention, the method further includes detecting whether the difference between the estimated stop line value and the historical stop line position exceeds a drift threshold, i.e.: ; in, The drift threshold can be set to 5 meters. When the above formula is continuous When the window is established, pause the calibration of the stop line. It can be set to 3. At this time, the stop line may actually move due to road reconstruction, triggering a road reconstruction warning and transferring to manual review. This prevents the mixed update of the estimated stop line information before and after the reconstruction into a false stop line position between the two.

[0055] Step 102: Based on the stop line and the trajectory data of multiple vehicles passing through the entrance lane, determine the directed distance information of multiple sampling points in the trajectory data; the directed distance information includes the distance information and direction information from the sampling point to the stop line of the entrance lane.

[0056] In one embodiment of the present invention, such as Figure 3 As shown, step 102 includes: Step 1021: Based on the topological relationship information of the target intersection, calculate the matching score from the sampling point to each approach lane of the target intersection, and obtain the first approach lane with the highest matching score corresponding to the sampling point.

[0057] Specifically, the topological relationship information is the road network topology map and / or approach lane index table of the target intersection. Based on the topological relationship information, a coordinate system is established with the center of the target intersection as the origin. According to the latitude and longitude of the sampling points, each sampling point is matched to the corresponding position in the coordinate system. In this embodiment, directional consistency constraints are introduced on the basis of distance detection, and the matching score between each trajectory point and each candidate approach lane is calculated: ; in, For the point pair of this trajectory Matching score for each import lane; For the trajectory point to the th The perpendicular distance between the centerlines of the two entrance lanes, in meters; This is a distance attenuation scale parameter, which can be set to 10 meters to control the degree to which distance affects the score. The vehicle's heading angle at that point can be obtained via GPS messages or the location of adjacent track points. For the first The forward direction angle of the entrance lane, through The cosine term is truncated to be non-negative, so that the road score for the angle between the driving direction and the road direction exceeds 90° (such as oncoming or lateral driving) is zero, so as to avoid reverse driving being mistakenly matched as negative correlation. Finally, the first entrance road with the highest matching degree score is taken as the corresponding entrance road for that point.

[0058] Step 1022: Determine the distance information of the sampling point as the distance from the projection point of the sampling point on the center line of the first inlet channel to the stop line corresponding to the first inlet channel along the center line.

[0059] Step 1023: Determine the direction information corresponding to the sampling point based on the upstream and downstream positional relationship of the sampling point relative to the stop line of the first inlet channel.

[0060] Specifically, after determining the entrance lane to which the sampling point belongs, calculate the projection of the sampling point on the centerline of the corresponding entrance lane, and the directed distance from the centerline to the corresponding stop line: ; in, For the first The directed distance between points on the trajectory, in meters; Let this be the projection of the point onto the center line of the entrance lane; For along the center line from to The distance is the distance information in the directed distance information of the sampling point. For curved road segments, since the center line of the approach road is usually represented by a discrete node sequence in the road network topology map, this embodiment approximates the center line as a broken line formed by connecting adjacent nodes in sequence. When calculating the directed distance, first determine the broken line segment where the projection point is located, and then successively accumulate the partial length from the projection point to the endpoint of the broken line segment, the length of the complete broken line segment between the endpoint and the stop line, and the partial length within the broken line segment where the stop line is located to obtain an approximate value of the arc length along the line from the projection point to the stop line, which is used as the distance information in the directed distance information of the sampling point. The direction information of the sampling point is determined by the position of the sampling point relative to the stop line. If the sampling point is upstream of the stop line, the direction information is positive; if the sampling point is downstream of the stop line, the direction information is negative. When the sampling point is exactly at the stop line, the direction information is zero. The distance information and the direction information together constitute the directed distance information.

[0061] Step 103: Construct the spatiotemporal trajectory information corresponding to the vehicle based on the directed distance information, and analyze the geometric feature parameters of the spatiotemporal trajectory information.

[0062] In one embodiment of the present invention, such as Figure 4 As shown, step 103 includes: Step 1031: Based on the topological relationship information of the target intersection, associate sampling points belonging to the same vehicle and the same approach lane as the same journey.

[0063] Specifically, after determining the entrance lane corresponding to each sampling point, sampling points belonging to the same vehicle but to different entrance lanes are divided into different journeys and assigned different journey identifiers.

[0064] Step 1032: Construct the spatiotemporal trajectory information based on the time information of sampling points belonging to the same journey and the directed distance information. The spatiotemporal trajectory information contains multiple trajectory points, each of which corresponds one-to-one with a sampling point. Each trajectory point is composed of the time information and directed distance information of the corresponding sampling point.

[0065] Specifically, based on the time information and directed distance information of the sampling points, trajectory points corresponding to the sampling points are constructed. The time information is determined by the timestamp of the sampling time of the sampling point. Trajectory points corresponding to multiple sampling points belonging to the same vehicle and the same journey are sorted according to the order of their time information, forming a point sequence composed of multiple trajectory points. This point sequence is the spatiotemporal trajectory information. Since each trajectory point is composed of time information and directed distance information, it possesses definite two-dimensional coordinate attributes. Therefore, the spatiotemporal trajectory information essentially represents the geometric distribution of the vehicle in the two-dimensional space of "time information - directed distance information." In other words, whether it is a sequence of trajectory points arranged according to time information or a discrete point graph displayed on a two-dimensional coordinate plane with time information as the horizontal axis and directed distance information as the vertical axis, it is all spatiotemporal trajectory information. By analyzing the spatiotemporal trajectory information, the geometric feature parameters corresponding to the vehicle can be obtained. These geometric feature parameters can reflect the vehicle's motion state. Figure 7 , Figure 7This is a schematic diagram of spatiotemporal trajectory information constructed according to an embodiment of this application. In the diagram, black circles represent trajectory points of this spatiotemporal trajectory information, and white circles represent the line-crossing points corresponding to this spatiotemporal trajectory information, calculated through subsequent steps. It can be seen that the slope between the first and second trajectory points is not 0, reflecting that the corresponding vehicle is in a moving state between the first and second trajectory points. The magnitude of the slope between the two trajectory points also reflects the vehicle's speed along the approach lane. The slope between the second and third trajectory points is 0 or approximately 0, reflecting that the corresponding vehicle is in a stationary state between the second and third trajectory points. The slope between the third and fourth trajectory points is different from the slope between the fourth and fifth trajectory points, reflecting a change in the vehicle's speed. By analyzing the curvature of the corresponding trajectory points, the acceleration and deceleration intensity of the corresponding vehicle within this interval can be reflected. It should be noted that, to more intuitively demonstrate the relationship between spatiotemporal trajectory information and vehicle motion state... Figure 7 By connecting the trajectory points in the spatiotemporal trajectory information into a curve, in the actual calculation process, it is not necessary to actually draw a continuous curve. Instead, the corresponding geometric characteristic parameters such as slope and curvature, or other information such as the time point of crossing the line and the parking trajectory segment can be calculated directly based on the spatiotemporal trajectory information in the form of trajectory point sequence or discrete point graph.

[0066] Step 1033: Analyze the geometric feature parameters of each trajectory point in the spatiotemporal trajectory information. The geometric feature parameters include the multi-scale slope information, curvature information, and slope consistency information of the trajectory point. The multi-scale slope information includes multiple local slope information of the trajectory point, and the local slope information is the slope between the trajectory point and another trajectory point.

[0067] Specifically, in this embodiment, the geometric feature parameters include multi-scale slope information, curvature information, and slope consistency information. To suppress the influence of satellite system sampling noise on the trajectory point slope, a joint estimation is performed on each trajectory point at multiple time scales to obtain the multi-scale slope information of the trajectory point. ; in, For the first A trajectory point in the window width The local slope information is the slope of the line connecting the trajectory point and another trajectory point, which physically represents the speed of the vehicle along the line, in meters per second. , The first The trajectory point moves backwards to the th... Directed distance and time information for each trajectory point.

[0068] In one embodiment of this application, respectively take The final output multi-scale slope information is as follows: ; The multi-scale slope information has a dimension of 4, and different window widths can also be set according to actual needs. This application does not impose specific limitations here, and further establishes slope consistency information based on the degree of dispersion of multi-scale slope information: ; in, , These are the standard deviation and mean of the multi-scale slope information at that point, respectively. To prevent the constant from having a denominator of zero, it can be set to 0.1 m / s; The closer the value is to 1, the more consistent the velocity estimates across different scales and the more stable the motion. A low value indicates that the point is in the transition period between acceleration and deceleration.

[0069] Furthermore, the curvature information of each trajectory point on the spatiotemporal trajectory is calculated, and its physical meaning is the absolute value of the equivalent acceleration: ; in, For the first Curvature information of each trajectory point; , The minimum scale slope information for two adjacent points, i.e., within the window width Local slope information when the value is 1; The equivalent acceleration at that point is expressed in meters per second². Based on the magnitude of the curvature and the direction of the slope change, the motion state of each trajectory point and the degree of acceleration or deceleration can be determined.

[0070] Step 104: Superimpose multiple spatiotemporal trajectory information onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map.

[0071] Specifically, the spatiotemporal trajectory information of multiple vehicles belonging to the same entrance and within the same sampling period is superimposed onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map. See [link / reference]. Figure 8 , Figure 8 This is a schematic diagram of a multi-vehicle spatiotemporal trajectory diagram according to an embodiment of this application. The spatiotemporal trajectory information of different vehicles is superimposed on the same two-dimensional coordinate system of "time information - directed distance information" to obtain the multi-vehicle spatiotemporal trajectory diagram, which can further display and analyze the overall traffic flow pattern. It should be noted that... Figure 8The purpose of connecting the trajectory points of each vehicle into a curve is to facilitate the display of the relationship between spatiotemporal trajectory information and vehicle motion state. In actual processing, it is not necessary to connect them into a curve. Instead, a multi-vehicle spatiotemporal trajectory map can be constructed and processed in the form of a point sequence or discrete point graph. In this way, the queuing evolution process of traffic flow can be analyzed through the distribution pattern of trajectory points.

[0072] Step 105: Based on the geometric feature parameters of each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, obtain the traffic wave information of the multi-vehicle spatiotemporal trajectory map, and based on the traffic wave information, obtain the signal interval information of the approach lane.

[0073] In one embodiment of the present invention, such as Figure 5 As shown, step 105 includes: Step 1051: Based on the geometric feature parameters of the spatiotemporal trajectory information, form a set of parking trajectory segments of the spatiotemporal trajectory information; the set of parking trajectory segments includes at least one parking trajectory segment.

[0074] In one embodiment of the present invention, step 1051 includes: Based on the geometric feature parameters of each trajectory point in the spatiotemporal trajectory information, a feature vector of the trajectory point is constructed; The feature vector is analyzed using a first classification model to determine the motion state information of the trajectory point; Merge the trajectory segments between multiple adjacent trajectory points whose motion state information is in the parking state to generate a parking trajectory segment; Based on the parking trajectory segments, the parking trajectory segment set is formed.

[0075] It should be noted that the trajectory segment described in the embodiments of this application refers to a continuous sequence of points in spatiotemporal trajectory information. Specifically, a trajectory segment consists of at least two trajectory points with continuous time information. In a two-dimensional coordinate system of "time information - directed distance information", the trajectory segment represents the local motion path of the vehicle within the corresponding time period. The feature vector of the i-th trajectory point... At least its multi-scale slope information Slope consistency information curvature information The above data can be obtained from step 1033. In addition, other information may be included, such as displacement and instantaneous velocity relative to adjacent trajectory points. Directed distance This information is used to facilitate a more comprehensive analysis of the motion pattern of the trajectory point in subsequent steps.

[0076] The first classification model can be a sequence model, such as a conditional random field model or a bidirectional long short-term memory network, or a point-by-point classifier model, such as a random forest model or a gradient boosting decision tree model. The sequence model is then smoothed. The motion state information of each trajectory point in the trajectory data corresponding to the real parking event is labeled in advance. The motion state information includes at least driving state, parking state and transition state. A historical labeled trajectory training set is obtained. The model is trained on the historical labeled trajectory training set to obtain a first classification model that can determine the corresponding motion state of the trajectory point based on its feature vector.

[0077] The feature vectors of all constructed trajectory points are used as input to the first classification model. The first classification model determines the motion state information of each trajectory point and outputs it. The trajectory segments between multiple adjacent trajectory points belonging to the same parking type are aggregated to obtain multiple parking trajectory segments. Then, short parking trajectory segments with a total parking time of less than a preset time threshold (which can be set to 5 seconds) are filtered to obtain the parking trajectory segment set for each vehicle. Where s is the sequence number of the parking trajectory segment. , Let the start and end times of the s-th parking trajectory segment be given. The first parking time of each vehicle can be obtained from the set of parking trajectory segments. With the first parking location This is used for calculating subsequent bicycle delay information.

[0078] Step 1052: Based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, obtain the traffic wave information of the multi-vehicle spatiotemporal trajectory map, wherein the traffic wave information includes a parking wave function and a starting wave function.

[0079] In one embodiment of the present invention, step 1052 includes: Based on the set of parking trajectory segments of multiple spatiotemporal trajectory information, the trajectory points at the start time of the first parking trajectory segment corresponding to the multiple spatiotemporal trajectory information are obtained to form a first trajectory point set; Obtain the trajectory points at the termination time of the last parking trajectory segment corresponding to multiple spatiotemporal trajectory information to form a second trajectory point set; The parking wave function is obtained based on the first set of trajectory points, and the starting wave function is obtained based on the second set of trajectory points.

[0080] Specifically, on the multi-vehicle spatiotemporal trajectory map, for the j-th vehicle, the trajectory point at the start time of its first stopping segment is denoted as... The starting time trajectory points of the first parking trajectory segments of all vehicles are collected into a first trajectory point set. Similarly, for the j-th vehicle, the ending time trajectory point of its last parking trajectory segment is denoted as... The trajectory points at the end time of the last stop of all vehicles are collected into a second trajectory point set.

[0081] The parking wave function is obtained by performing least-squares linear regression on the first set of trajectory points: ; in, Let be the slope of the parking wave function, which physically represents the speed at which the parking wave propagates upstream along the approach lane. This represents the average of the time information of the trajectory points at each starting time. This represents the average of the directed distance information of the trajectory points at each starting time.

[0082] Performing least-squares linear regression on the second trajectory point set yields the initiation wave function: ; in, The slope of the initiation wave function represents the physical velocity of the initiation wave propagating upstream along the inlet channel. This represents the average of the time information of the trajectory points at each termination time. This represents the average of the directed distance information of the trajectory points at each termination time.

[0083] Further, based on the extracted queuing features, the trajectory point with the largest directed distance along the time information coordinate axis, located upstream of the stop line and in a stopped state, is obtained. The absolute value of its directed distance information is determined as the maximum queue length within that sampling period. Combined with floating car penetration rate The observed number of vehicles in queues is amplified to an estimate of the actual number of vehicles in queues: ; in, Floating vehicle penetration rate, which is the proportion of floating vehicles to total traffic, is determined by historical traffic data or provided by the operator. This is an estimate of the actual number of vehicles in the queue. This represents the number of vehicles observed to be queuing for parking within the sampling period.

[0084] In one embodiment of the present invention, step S1052 further includes: Based on the multiple sets of parking trajectory segments containing the aforementioned spatiotemporal trajectory information, the start time trajectory points and end time trajectory points of all parking trajectory segments are obtained to form multiple sets of start and stop points. The multiple sets of start and stop points are divided into multiple candidate signal periods according to a preset time interval condition. The starting time trajectory points of each parking trajectory segment in each candidate signal period constitute the third trajectory point set of the candidate signal period, and the ending time trajectory points of each parking trajectory segment constitute the fourth time trajectory point set of the candidate signal period. For each candidate signal period, the parking wave function is obtained based on the third trajectory point set corresponding to the candidate signal period, and the starting wave function is obtained based on the fourth trajectory point set corresponding to the candidate signal period.

[0085] Specifically, for intersections with high traffic volume and long queues, vehicles may need to make multiple stops and starts to pass through the intersection. If the parking wave function and starting wave function are calculated only using the first and last parking trajectory segments of each trajectory's spatiotemporal trajectory information, multiple actual signal intervals may be merged into one signal interval when calculating the signal interval information, causing serious errors in subsequent analysis. Therefore, for such intersections, this embodiment first obtains the start and end time trajectory points of all parking trajectory segments based on the set of parking trajectory segments of multiple spatiotemporal trajectory information, forming a set of start and stop points. The set of start and stop points corresponds one-to-one with the parking trajectory segments, and includes the start and end time trajectory points. For the first parking trajectory segment within the same approach lane... For any parking trajectory segment of the vehicle, if the time information of the trajectory point at the start time of that parking trajectory segment is... The time information of the trajectory point at the termination time is And satisfy > Then the starting and ending trajectory points of the parking trajectory segment are taken as a set of starting and stopping points for the vehicle.

[0086] For the set of all stop and start points of multiple vehicles within the same entrance lane, sort them according to their starting and ending trajectory points. The starting and stopping points are sorted sequentially, and then divided into different candidate signal cycles based on the time interval between adjacent first stopping times, a preset minimum / maximum signal cycle range, or the candidate cycle length obtained from historical statistics. Multiple starting time trajectory points belonging to the same candidate signal cycle constitute the third trajectory point set of that cycle, and multiple ending time trajectory points belonging to the same candidate signal cycle constitute the fourth trajectory point set of that cycle.

[0087] For each candidate signal period, a parking wave function is fitted based on the third trajectory point set corresponding to the candidate signal period, and a start wave function is fitted based on the fourth trajectory point set corresponding to the candidate signal period. For abnormal parking trajectory segments that span multiple periods, lack start time trajectory points, lack end time trajectory points, or have time information of the end time trajectory point that is earlier than the time information of the end time trajectory point, they are not included in the traffic wave function fitting for that candidate signal period.

[0088] Step 1053: Obtain the signal interval information corresponding to the approach lane based on the parking wave function and the starting wave function.

[0089] In one embodiment of the present invention, step 1053 includes: Based on the parking wave function, the red light activation time corresponding to the multi-vehicle spatiotemporal trajectory map is obtained; Based on the start-up wave function, obtain the green light start time corresponding to the multi-vehicle spatiotemporal trajectory map; Based on the red light start time and green light start time of the same multi-vehicle spatiotemporal trajectory map, determine the red light interval information and red light duration information corresponding to the entrance lane; Based on the red light or green light turn-on time of the multi-vehicle spatiotemporal trajectory map of adjacent sampling periods, the signal cycle information corresponding to the entrance lane is determined; Based on the signal cycle information and the red light duration information, determine the green light ratio information corresponding to the approach lane; The signal interval information includes the signal cycle information, the red light interval information, the red light duration information, and the green light ratio information.

[0090] Specifically, see Figure 9 , Figure 9 This is a schematic diagram of a multi-vehicle spatiotemporal trajectory diagram with signal interval information according to an embodiment of this application. Figure 9 The black square dots represent either the first or third trajectory point set, and the dashed line corresponding to each black square dot represents the parking wave function. The white triangle dots represent either the second or fourth trajectory point set, and the dashed line corresponding to each white square dot represents the starting wave function. The parking wave function propagates upstream from the stop line when the red light turns on, thus extending to the stop line axis. The time intercept of the starting wave function is the moment the red light turns on. Similarly, the starting wave function propagates upstream from the moment the green light turns on, starting from the stop line. Therefore, the time intercept of the starting wave function line on the stop line axis is the moment the green light turns on. Figure 9 As shown in the diagram, the black pentagram indicates when the red light turns on, and the white pentagram indicates when the green light turns on. Therefore, we can conclude that: ; ; in, This is when the red light turns on; When the green light turns on. Let be the slope of the parking wave function. This represents the average of the time information of the trajectory points at the initial moment. This represents the mean of the directed distance information of the trajectory points at the initial time. To determine the slope of the initial wave function, This represents the average of the time information of the trajectory points at each termination time. This represents the average of the directed distance information of the trajectory points at each termination time. The red light start time and the green light start time are uniformly given by the same geometric framework, which is the key to the robustness of back-calculation even with small samples.

[0091] The red light activation time and green light activation time are calculated for adjacent sampling periods, thus obtaining the red light activation time sequence for each period. Sequence of green light activation times ,in, Using the cycle number, the signal interval information of the target intersection can be analyzed: ; ; ; in, Red light duration; This is the signal cycle information, which is the difference between the times when two adjacent red lights turn on; The green light ratio, which represents the proportion of green light duration within a cycle, can be used to further determine the red light intervals for the approach lanes. The area between the red light activation time and the green light activation time within the same cycle is the red light interval information. , Figure 9 The dotted filled areas in the graph represent the red light interval information corresponding to the multi-vehicle spatiotemporal trajectory. Further, the green light interval information corresponding to the approach lanes can be obtained. The area between the green light start time of the current cycle and the red light start time of the next cycle represents the green light interval information. The signal interval information includes at least the signal period information, the red light duration information, the red light interval information, and the green light ratio information.

[0092] In one embodiment of the present invention, step 1053 further includes: Obtain at least one historical trajectory data of the inlet channel within a historical sampling period that is in the same time period as the current sampling period; Based on the directed distance information of each trajectory point in the historical trajectory data and the directed distance information of each trajectory point in the trajectory data within the current sampling period, the historical trajectory data is aligned with the trajectory data within the current sampling period; An enhanced trajectory data set is formed based on the aligned historical trajectory data and the trajectory data; The enhanced parking wave function and enhanced start wave function of the enhanced trajectory data set are obtained, and the enhanced signal interval information is obtained based on the enhanced parking wave function and the enhanced start wave function. The enhanced signal interval information is used as the signal interval information of the approach lane.

[0093] Specifically, to prevent insufficient trajectory data within a single sampling period due to a small sample rate from causing unstable regression, the signal interval information can be calibrated based on historical data. Historical trajectory data from the same time period as the current sampling period (e.g., the same peak period over the past 30 working days) are collected, aligned with the trajectory data of the current sampling period according to phase, and then aggregated to form an enhanced trajectory data set. Further, the enhanced parking wave function and enhanced starting wave function of the enhanced trajectory data set are obtained using the aforementioned method. Based on the enhanced parking wave function and the enhanced starting wave function, enhanced signal interval information is obtained and used as the signal interval information for the approach lane. This makes the signal interval information more accurate.

[0094] In one embodiment of the present invention, step 1053 further includes: Multiple enhanced trajectory data sets are acquired, wherein the multiple enhanced trajectory data sets are formed by aligning historical trajectory data with trajectory data within the current sampling period using different statistical windows, and the statistical window includes one or more historical sampling periods; Based on multiple sets of enhanced trajectory data, multiple enhanced signal interval information is obtained, and the enhanced signal interval information located at the median is used as the signal interval information of the approach lane.

[0095] Specifically, historical trajectory data from different statistical windows can be aggregated with trajectory data from the current sampling period to obtain multiple enhanced signal interval information. The statistical window includes one or more historical sampling periods. Based on the multiple sets of enhanced trajectory data, multiple enhanced signal interval information are obtained through the above method. The enhanced signal interval information located at the median is taken as the final signal interval information to suppress random deviations in individual windows.

[0096] Step 106: Determine the delay index information of the approach lane based on the geometric feature parameters and the signal interval information. The delay index information includes one or more of the following: parking delay information, queuing delay information, fixed delay information, and operation delay information.

[0097] In one embodiment of the present invention, such as Figure 6 As shown, step 106 includes: Step 1061: Analyze the set of parking trajectory segments based on the signal interval information to determine the valid parking trajectory segments, which are parking trajectory segments generated due to signal waiting.

[0098] In one embodiment of the present invention, step 1061 includes: Obtain the overlap between the time interval information of the parking trajectory segment and the red light interval information; The parking trajectory segment that meets the preset overlap condition is determined as the valid parking trajectory segment.

[0099] Specifically, because floating cars involve a large number of non-signal stops such as passenger pick-up / drop-off and roadside temporary stops, including these delays would significantly overestimate the delay parameters of the target intersection. Therefore, the overlap ratio between the stop segment and the red light segment information is calculated: ; in, For the first The overlap ratio between parking trajectory segments and red light interval information This indicates the length of the overlapping interval between interval A and interval B, that is, the duration of the intersection between the parking trajectory segment and the red light interval information. , This refers to the start and end times of the parking trajectory segment; This refers to the red light interval information for the current parking section within a given cycle.

[0100] like If the signal is detected as a stop, the stopping trajectory segment is recorded as a valid stopping trajectory segment and included in the delay calculation. The overlap determination threshold can be set to 0.5. If the vehicle stops entirely within the green light zone, it is considered an unsignaled stop, such as temporary passenger pick-up / drop-off, temporary parking, or illegal parking. It is then removed from the sample and not included in the calculation of delay index information.

[0101] like In the critical case, a spatial rationality check is performed. If the directed distance information of the parking trajectory segment satisfies... This is also included in the sample used for calculating the delay index information. For the maximum queue length, To allow for a buffer, a margin of 20 meters can be set. The purpose of this step is to treat parking segments that extend far beyond the end of the queue as non-signaled parking even if they fall during red light periods.

[0102] Step 1062: Calculate the time point of the spatiotemporal trajectory information crossing the line based on the geometric feature parameters of the spatiotemporal trajectory information.

[0103] In one embodiment of the present invention, step 1062 includes: The first upstream trajectory point and the first downstream trajectory point are obtained from the spatiotemporal trajectory information; the first upstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located upstream of the stop line, and the first downstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located downstream of the stop line. Based on the geometric feature parameters of the first upstream trajectory point and the first downstream trajectory point, either a uniform interpolation model or an acceleration / deceleration interpolation model is selected to calculate the time point of the spatiotemporal trajectory information.

[0104] Specifically, the interval where the directed distance information of adjacent trajectory points changes from positive to negative uniquely corresponds to the interval in which the vehicle passes the stop line. That is, the moment of crossing the line must be located within this interval. Since the directed distance information at the moment of crossing the line is always zero, it is only necessary to calculate the time information of the moment of crossing the line to obtain the moment of crossing the line. This step involves locating the adjacent trajectory points in the directed distance information of each vehicle where the sign changes, and recording the last trajectory point with a positive directed distance value before the change as the first upstream trajectory point. The first trajectory point with a negative directed distance after the jump is the first downstream trajectory point. ,in, , These are the time information for the first upstream trajectory point and the first downstream trajectory point, respectively. , The directed distance information of the first upstream trajectory point and the first downstream trajectory point are respectively. Based on the geometric feature parameters of the first upstream trajectory point and the first downstream trajectory point, the geometric feature parameters are the multi-scale slope information, slope consistency information and curvature information obtained in the above steps, and the interpolation model is adaptively selected.

[0105] See Figure 10 , Figure 10 This is a schematic diagram illustrating the calculation of the crossing time point according to an embodiment of this application. The white circular dots represent the crossing time points. If, based on the geometric feature parameters of the first upstream trajectory point and the first downstream trajectory point, it is determined that the vehicle is crossing the line at a constant speed, then the time information of the crossing time point is calculated according to the constant speed motion interpolation model. ; ; in, This refers to the time information at the point when the line crosses the line. It represents the average speed of the trajectory segment crossing the line, that is, the average speed of the trajectory segment between the first upstream trajectory point and the first downstream trajectory point.

[0106] If, based on the geometric characteristic parameters of the first upstream trajectory point and the first downstream trajectory point, it is determined whether the vehicle is accelerating or decelerating when crossing the line, then the time information of the crossing point is calculated according to the acceleration / deceleration interpolation model, and the value falling within the interval is taken. Internal, envoy The real roots that are zero are used as the time information of the time point when the line crosses. The acceleration / deceleration interpolation model is as follows: ; ; in, for Directed distance information at any given moment; , The velocity information of the first upstream trajectory point and the first downstream trajectory point are respectively determined by the multi-scale slope information of the first upstream trajectory point and the first downstream trajectory point. It is the equivalent acceleration of the trajectory segment crossing the line.

[0107] Step 1063: Based on the signal interval information, the effective parking trajectory segment, and the crossing time point, obtain the delay index information corresponding to the approach lane.

[0108] In one embodiment of the present invention, step 1063 includes: Based on the effective parking trajectory segment, calculate the single-vehicle parking delay information of the spatiotemporal trajectory information, and based on the single-vehicle parking delay information of each spatiotemporal trajectory information, calculate the parking delay information of the entrance lane; Based on the effective parking trajectory segment and the crossing time point, calculate the single-vehicle queuing delay information of the spatiotemporal trajectory information, and calculate the queuing delay information of the entrance lane based on the single-vehicle queuing delay information of each spatiotemporal trajectory information. Based on the signal interval information and the effective parking trajectory segment, calculate the fixed delay information of a single vehicle in the spatiotemporal trajectory information, and calculate the fixed delay information of the approach lane based on the fixed delay information of a single vehicle in each spatiotemporal trajectory information. Based on the single-vehicle queuing delay information and the single-vehicle fixed delay information, the single-vehicle operation delay information of the spatiotemporal trajectory information is calculated, and based on the single-vehicle operation delay information of each of the spatiotemporal trajectory information, the operation delay information of the entrance lane is calculated.

[0109] It should be noted that the delay index information described in this embodiment refers to the time lost by a vehicle during its journey due to interference from other vehicles beyond the driver's control or obstruction by traffic control facilities. Depending on the cause and analysis perspective, it can be categorized into four types: parking delay information, queuing delay information, fixed delay information, and operational delay information. In actual deployment, one or more of these types can be output as needed. Parking delay information refers to the time delay caused by a vehicle coming to a standstill for some reason, i.e., the sum of the duration during which the vehicle speed drops to zero (or below the parking threshold). Queuing delay information is the delay time caused by queuing, which is the sum of the queuing time and the time spent traveling at a free-flowing speed. The time difference in queuing sections refers to the time from when a vehicle first stops to when it crosses the stop line, and the queuing section refers to the distance from the first stopping section to the stop line. Fixed delay information refers to delays caused by traffic control devices (traffic lights) that are unrelated to the amount of traffic or interference from other vehicles. Its physical essence is the time lost by a vehicle even under ideal conditions with no preceding vehicles and no queues, simply because it arrives at a red light and has to wait until the green light turns on. Operational delay information refers to delays caused by interference from various traffic components, such as waiting to cross, traffic congestion, continuous stops, and time lost due to pedestrians and turning vehicles. Its physical meaning is the additional time lost by a vehicle due to being behind other vehicles, being affected by the lag in the propagation of the starting wave, and being affected by oversaturation overflow.

[0110] In one embodiment of the present invention, the single-vehicle parking delay information is calculated by using the time lost due to the vehicle remaining stationary, i.e., the total time corresponding to the effective parking trajectory segment of the vehicle. ; in, For the first For each vehicle's single-vehicle parking delay information, sum the duration of all valid parking trajectory segments for that vehicle, where s is the sequence number of the valid parking trajectory segment. , Let be the start and end times of the s-th valid parking trajectory segment.

[0111] Based on the single-vehicle parking delay information of all trajectory data within at least one sampling period, calculate its mean, median, and 95th percentile, which are used as the parking delay information of the approach lane within the corresponding sampling period.

[0112] In one embodiment of the present invention, single-vehicle queuing delay information is calculated by the difference between the queuing time and the time required to pass through the same road segment at free-flow speed: ; in, For the first Information on bicycle queue delays; This provides the time information for the vehicle's crossing point. This is the moment the vehicle first stops. This refers to the vehicle's initial parking position, i.e., the directed distance information at the time of the initial parking. The free travel speed of the vehicle on the entrance lane is determined by either the speed limit of the road section or the 85th percentile driving speed from offline large-sample statistics.

[0113] Based on the single-vehicle queuing delay information of all trajectory data within at least one sampling period, calculate its mean, median, and 95th percentile, which are used as the queuing delay information of that entrance lane within the corresponding sampling period.

[0114] In one embodiment of the present invention, the fixed delay information for a single vehicle is a delay component caused by signal control and unrelated to traffic flow. Accurately calculating the fixed delay information for each vehicle requires a determination based on the matching relationship between the arrival time of unobstructed traffic and the red light interval, which is complex to implement. To reduce computational complexity, this embodiment adopts a simplification strategy. Typically, the arrival times of all vehicles at the target intersection are approximately uniformly distributed within the signal cycle, and the average waiting time for vehicles arriving within the red light interval is approximately equal to half the red light duration. Therefore, for vehicles containing effective stopping trajectory segments, their fixed delay information is taken as half the duration of the corresponding red light interval information. ; ; in, For the first Single-vehicle fixed delay information for vehicles containing valid parking trajectory segments. This provides information on the duration of the red light when the vehicle was stopped. , This refers to the times when the red light turns on and the times when the green light turns on.

[0115] For vehicles that do not contain effective parking trajectory segments, their single-vehicle fixed delay information is 0. The simplified steps in this embodiment decouple the single-vehicle fixed delay information from the specific arrival time, and calculate it based solely on the red light duration information. The calculation is simple, and this step does not change the physical property that the fixed delay is unrelated to traffic volume. Under the premise of simplifying the calculation and facilitating deployment, the accurate estimation of single-vehicle fixed delay is achieved.

[0116] Based on the fixed delay information of each vehicle in all trajectory data within at least one sampling period, calculate its mean, median, and 95th percentile, which are used as the fixed delay information of the approach lane in the corresponding sampling period.

[0117] In one embodiment of the present invention, the vehicle operation delay information is the delay caused by the mutual interference of various traffic components. The vehicle operation delay information is calculated from the remainder after deducting the fixed delay information of a single vehicle from the vehicle queuing delay information. ; in, For the first Information on vehicle operation delays.

[0118] In one embodiment of the present invention, the physical meaning of the single-vehicle operation delay information includes two parts. The first part is the waiting time generated during the queuing vehicle start-up and release process. That is, after the green light turns on, not all vehicles in the queue start simultaneously. Instead, the starting state gradually propagates from the vehicles at the stop line to the upstream vehicles. Vehicles located upstream in the queue need to wait for the starting wave to propagate to their location before they can begin to move. This waiting time can be approximated as... ; The velocity of the initiation wave is equal to the slope of the initiation wave function. The second part is the waiting time across cycles caused by oversaturation. That is, when one green light time is insufficient for the vehicle to pass through the intersection, the vehicle needs to continue to wait for one or more subsequent signal cycles. This additional waiting time can be determined based on the number of green light start points experienced by the vehicle from the first stop to the crossing of the line and the length of the signal cycle.

[0119] The overflow cycle number, i.e. the extra waiting cycle, is determined by the number of green light starting points crossed by the vehicle during the period from its first stop to crossing the line: ; in, For the first The number of overflow cycles experienced by the vehicle; For the first The moment when the green light turns on in each cycle; This represents the number of green light cycles during which the vehicle is in the queue. Normally, a vehicle crosses exactly one green light cycle's starting point during its queue period, meaning it is allowed to proceed during this cycle. .

[0120] when This indicates that the vehicle experienced at least one green light but failed to pass through the intersection, meaning a single green light could not clear the queue, thus indicating that it had experienced saturation. Based on this, the proportion of oversaturated vehicles in the current cycle or several cycles can be further calculated; that is, the percentage of oversaturated vehicles in the total number of vehicles. This is used to further evaluate the traffic quality of that approach lane. At this point, the time information of the vehicle's crossing point is crucial. Compared to the first parking time The delay is postponed to a subsequent signal cycle. The single-vehicle queuing delay information and the single-vehicle operation delay information obtained by subtracting the fixed single-vehicle delay information already include the additional delay caused by waiting across cycles; this additional delay can be approximately understood as... The relevant cross-cycle waiting component, where C is the signal cycle duration.

[0121] Based on the single-vehicle running delay information of all trajectory data within at least one sampling period, calculate its mean, median, and 95th percentile, which are used as the running delay information of the approach lane within the corresponding sampling period.

[0122] In one embodiment of the present invention, due to GPS errors or the end of a vehicle's journey, within a sampling period, there may be trajectory data where a vehicle enters a queue and stops, but its trajectory is interrupted before crossing the line. That is, the trajectory data only has trajectory points in a portion upstream of the stop line. For such trajectory data, the time of its first stop is... With the first parking location Having been obtained through the above steps, this embodiment further predicts the line crossing time point using signal interval information and the starting wave function: ; in, To predict the time information of the crossing point; The green light illuminates at the start of the cycle in which the vehicle first stops; This is the time required for the wave function to propagate from the stop line to the vehicle's initial stopping position. The number of overflow cycles is calculated based on the number of times a vehicle that has not crossed the line "crossed the green light start point but still did not start" during the observation period. For signal periodic information, replace The system calculates the single-vehicle queuing delay information, single-vehicle fixed delay information, and single-vehicle running delay information of the trajectory data to expand the sample size and calculate the delay index information of the target intersection together with other trajectory data.

[0123] In one embodiment of the present invention, the delay index information further includes: the maximum queue length obtained through step 1052. and the estimated number of vehicles actually queuing The oversaturated vehicle ratio obtained in step 1063, and the red light duration information, signal cycle information, and green light ratio information obtained in step 1053.

[0124] Step 107: Based on the delay index information of each of the aforementioned approach lanes, obtain the total delay index information of the target intersection. The total delay index information includes one or more of the following: total parking delay information, total queuing delay information, total fixed delay information, and total operational delay information.

[0125] Specifically, after obtaining the delay index information for all approach lanes at the target intersection, the delay index information for all approach lanes is summarized to obtain the total delay index information for the target intersection. The total delay index information includes at least one or more of the total stopping delay information, the total queuing delay information, the total fixed delay information, and the total running delay information. In addition, it may also include the maximum queue length for each approach lane obtained through step 1052 above. and the estimated number of vehicles actually queuing The oversaturated vehicle ratio of each approach lane obtained in step 1063, and the red light duration, signal cycle, and green light ratio information of each approach lane obtained in step 1053 are summarized and incorporated into the total delay index information.

[0126] The above embodiments construct spatiotemporal trajectory information, converting the original trajectory data of floating cars obtained by the satellite positioning system into a sequence of trajectory points containing time and directed distance information, measured along the approach lane with the stop line as the zero point. This avoids directly judging stopping, starting, and crossing events on the original planar trajectory, enabling comparison of sparse trajectory data of different vehicles and different cycles under the same coordinate system. This provides a stable basis for delay calculation under small sample conditions. Furthermore, this application analyzes the geometric feature parameters of the spatiotemporal trajectory information to infer the signal interval information of the target intersection. Without relying on external data such as signal timing or roadside detection equipment, it can still stably and accurately analyze the delay index information of the approach lane. The delay index information of each approach lane is summarized to obtain the total delay index information of the target intersection, which is convenient for large-scale deployment in actual traffic networks. Moreover, in this embodiment, the delay index information is divided into parking delay information, queuing delay information, fixed delay information, and operational delay information, which can evaluate the operational quality of the target intersection from different perspectives and provide a more comprehensive basis for optimizing the signal timing of the intersection.

[0127] This embodiment also provides an intersection delay analysis device based on floating car trajectories with a small sample rate, such as... Figure 11 As shown, the analysis device includes: Stop line acquisition module 210 is used to acquire the stop line of at least one approach lane of the target intersection; The directed distance acquisition module 220 is used to determine the directed distance information of multiple sampling points in the trajectory data based on the stop line and the trajectory data of multiple vehicles passing through the entrance lane; the directed distance information includes the distance information and direction information from the sampling point to the stop line of the entrance lane; The spatiotemporal trajectory construction module 230 is used to construct the spatiotemporal trajectory information corresponding to the vehicle based on the directed distance information, and analyze the geometric feature parameters of the spatiotemporal trajectory information; The spatiotemporal trajectory overlay module 240 is used to overlay multiple spatiotemporal trajectory information onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map. The signal interval acquisition module 250 is used to obtain traffic wave information of the multi-vehicle spatiotemporal trajectory map based on the geometric feature parameters of each spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, and to obtain the signal interval information of the approach lane based on the traffic wave information. The delay index analysis module 260 is used to determine the delay index information of the approach lane based on the geometric feature parameters and the signal interval information. The delay index information includes one or more of the following: parking delay information, queuing delay information, fixed delay information, and operation delay information. The total delay index analysis module 270 is used to obtain the total delay index information of the target intersection based on the delay index information of each of the approach lanes. The total delay index information includes one or more of the following: total parking delay information, total queuing delay information, total fixed delay information, and total operational delay information.

[0128] In one embodiment of the present invention, the stop line acquisition module 210 of the device includes: The historical trajectory processing unit is used to preprocess the historical trajectory data of multiple vehicles passing through the target intersection within at least one historical sampling period to obtain preprocessed historical trajectory data. The low-speed projection point calculation unit is used to project multiple sampling points with instantaneous speeds lower than a first preset threshold from each preprocessed historical trajectory data to the center line of the corresponding entrance lane of the target intersection, forming a low-speed projection point set. The density estimation unit is used to perform density estimation on the set of low-velocity projection points to obtain the spatial distribution density function corresponding to the set of low-velocity projection points. The candidate peak point calculation unit is used to extract the maximum points in the spatial distribution density function to form a set of candidate peak points; The stop line calculation unit is used to determine the stop line of the approach lane corresponding to the target intersection based on the candidate peak point set.

[0129] In one embodiment of the present invention, the stop line acquisition module 210 further includes: The trajectory data acquisition unit is used to acquire trajectory data of multiple vehicles passing through the entrance lane within at least one new sampling period after the historical sampling period. The stop line estimation unit is used to generate stop line estimation information based on the trajectory data within the new sampling period; The stop line calibration unit is used to calibrate the stop line corresponding to the inlet lane based on the stop line estimation information.

[0130] In one embodiment of the present invention, the directed distance acquisition module 220 of the device includes: The matching calculation unit is used to calculate the matching degree score from the sampling point to each approach lane of the target intersection based on the topological relationship information of the target intersection, and to obtain the first approach lane with the highest matching degree score corresponding to the sampling point. The distance information calculation unit is used to determine the distance information of the sampling point as the distance from the projection point of the sampling point on the center line of the first inlet lane to the stop line corresponding to the first inlet lane along the center line; The direction information calculation unit is used to determine the direction information corresponding to the sampling point based on the upstream and downstream positional relationship of the sampling point relative to the stop line of the first inlet channel.

[0131] In one embodiment of the present invention, the spatiotemporal trajectory construction module 230 of the device includes: The trip association module is used to associate sampling points belonging to the same vehicle and the same approach lane as the same trip based on the topological relationship information of the target intersection. The spatiotemporal trajectory construction unit is used to construct the spatiotemporal trajectory information based on the time information of sampling points belonging to the same journey and the directed distance information. The spatiotemporal trajectory information contains multiple trajectory points, each of which corresponds one-to-one with a sampling point. Each trajectory point is composed of the time information and directed distance information of the corresponding sampling point. The geometric feature analysis unit is used to analyze the geometric feature parameters of each trajectory point in the spatiotemporal trajectory information. The geometric feature parameters include multi-scale slope information, curvature information, and slope consistency information of the trajectory point. The multi-scale slope information includes multiple local slope information of the trajectory point, and the local slope information is the slope between the trajectory point and another trajectory point.

[0132] In one embodiment of the present invention, the signal interval acquisition module 250 of the device includes: A parking trajectory segment calculation unit is used to form a set of parking trajectory segments of the spatiotemporal trajectory information based on the geometric feature parameters of the spatiotemporal trajectory information; the set of parking trajectory segments includes at least one parking trajectory segment; The traffic wave calculation unit is used to obtain the traffic wave information of the multi-vehicle spatiotemporal trajectory map based on the set of parking trajectory segments of each spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map. The traffic wave information includes a parking wave function and a starting wave function. The signal interval calculation unit is used to obtain the signal interval information corresponding to the approach lane based on the parking wave function and the starting wave function.

[0133] In one embodiment of the present invention, the delay index analysis module 260 in the device includes: An effective parking trajectory segment calculation unit is used to analyze the set of parking trajectory segments based on the signal interval information to determine the effective parking trajectory segments, wherein the effective parking trajectory segments are parking trajectory segments generated due to signal waiting; The line crossing time point calculation unit is used to calculate the line crossing time point of the spatiotemporal trajectory information based on the geometric feature parameters of the spatiotemporal trajectory information; The delay index information calculation unit is used to obtain the delay index information corresponding to the approach lane based on the signal interval information, the effective parking trajectory segment, and the crossing time point.

[0134] In one embodiment of the present invention, the parking trajectory segment calculation unit in the device includes: The feature vector construction subunit is used to construct the feature vector of each trajectory point based on the geometric feature parameters of each trajectory point in the spatiotemporal trajectory information. The feature vector analysis subunit is used to analyze the feature vector using the first classification model to determine the motion state information of the trajectory point; The trajectory segment merging subunit is used to merge the trajectory segments between multiple adjacent trajectory points whose motion state information is in the parking state, and generate a parking trajectory segment. A set of generating sub-units is used to form the set of parking trajectory segments based on the parking trajectory segments.

[0135] In one embodiment of the present invention, the traffic wave calculation unit in the device includes: The first trajectory point acquisition subunit is used to acquire the trajectory points at the start time of the first parking trajectory segment corresponding to the multiple sets of parking trajectory information based on the multiple sets of parking trajectory segments of the multiple spatiotemporal trajectory information, and form a first trajectory point set. The second trajectory point acquisition subunit is used to acquire the trajectory points at the termination time of the last parking trajectory segment corresponding to the multiple sets of parking trajectory segments of the multiple spatiotemporal trajectory information, and form a second trajectory point set. The first parking wave function calculation subunit is used to obtain the parking wave function based on the first set of trajectory points; The first initiation wave function calculation subunit is used to obtain the initiation wave function based on the second trajectory point set.

[0136] In one embodiment of the present invention, the traffic wave calculation unit in the device includes: The start-stop point set acquisition subunit is used to acquire the start time trajectory point and end time trajectory point of all parking trajectory segments based on the parking trajectory segment set of multiple spatiotemporal trajectory information, and form multiple start-stop point sets; The third trajectory point acquisition subunit is used to divide the multiple sets of start and stop points into multiple candidate signal periods according to a preset time interval condition, and to form the third trajectory point set of the candidate signal period by the trajectory points of the starting time of each parking trajectory segment in each candidate signal period. The fourth trajectory point acquisition subunit is used to construct the fourth time trajectory point set of the candidate signal period from the trajectory points at the end time of each parking trajectory segment within each candidate signal period. The second parking wave function calculation subunit is used to obtain the parking wave function for each candidate signal period based on the third trajectory point set corresponding to the candidate signal period. The second initiation wave function calculation subunit is used to obtain the initiation wave function for each candidate signal period based on the fourth trajectory point set corresponding to the candidate signal period.

[0137] In one embodiment of the present invention, the signal interval calculation unit in the device includes: The red light start time calculation subunit is used to obtain the red light start time corresponding to the multi-vehicle spatiotemporal trajectory map based on the parking wave function. The green light start time calculation subunit is used to obtain the green light start time corresponding to the multi-vehicle spatiotemporal trajectory map based on the start wave function; The red light information calculation subunit is used to determine the red light interval information and red light duration information corresponding to the entrance lane based on the red light start time and green light start time of the same multi-vehicle spatiotemporal trajectory map. The signal period information calculation subunit is used to determine the signal period information corresponding to the entrance lane based on the red light start time or green light start time of the multi-vehicle spatiotemporal trajectory map of adjacent sampling periods. The green light ratio calculation subunit is used to determine the green light ratio information corresponding to the approach lane based on the signal cycle information and the red light duration information. The signal interval information includes the signal cycle information, the red light interval information, the red light duration information, and the green light ratio information.

[0138] In one embodiment of the present invention, the apparatus further includes a sample enhancement module, the sample enhancement module comprising: The historical data acquisition unit is used to acquire at least one historical trajectory data of the inlet channel within a historical sampling period that is in the same time period as the current sampling period; An alignment unit is used to align the historical trajectory data with the trajectory data in the current sampling period based on the directed distance information of each trajectory point in the historical trajectory data and the directed distance information of each trajectory point in the trajectory data in the current sampling period. The first enhanced data acquisition unit is used to form an enhanced trajectory data set based on the aligned historical trajectory data and the trajectory data; An enhanced traffic wave function calculation unit is used to obtain the enhanced parking wave function and the enhanced start wave function of the enhanced trajectory data set; The enhanced signal interval calculation unit is used to obtain enhanced signal interval information based on the enhanced parking wave function and the enhanced start wave function, and to use the enhanced signal interval information as the signal interval information of the approach lane.

[0139] In one embodiment of the present invention, the sample enhancement module further includes: The second enhanced data acquisition unit is used to acquire multiple enhanced trajectory data sets, wherein the multiple enhanced trajectory data sets are formed by aligning historical trajectory data using different statistical windows with the trajectory data in the current sampling period, and the statistical window includes one or more historical sampling periods; The second enhanced signal interval calculation unit is used to obtain multiple enhanced signal interval information based on multiple enhanced trajectory data sets, and to use the enhanced signal interval information located at the median as the signal interval information of the approach lane.

[0140] In one embodiment of the present invention, the effective parking trajectory segment calculation unit in the device further includes: The overlap calculation subunit is used to obtain the overlap between the time interval information of the parking trajectory segment and the red light interval information. The effective parking trajectory segment calculation subunit is used to determine the parking trajectory segment that meets the preset overlap condition as the effective parking trajectory segment.

[0141] In one embodiment of the present invention, the device further includes: The first upstream trajectory point acquisition subunit is used to acquire the first upstream trajectory point of the spatiotemporal trajectory information; the first upstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located upstream of the stop line; The first downstream trajectory point acquisition subunit is used to acquire the first downstream trajectory point of the spatiotemporal trajectory information; the first downstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located downstream of the stop line; The interpolation calculation subunit is used to calculate the crossing time point of the spatiotemporal trajectory information by selecting either a uniform interpolation model or an acceleration / deceleration interpolation model based on the geometric feature parameters of the first upstream trajectory point and the first downstream trajectory point.

[0142] In one embodiment of the present invention, the delay index information calculation unit in the device includes: The parking delay information calculation subunit is used to calculate the single-vehicle parking delay information of the spatiotemporal trajectory information based on the effective parking trajectory segment, and to calculate the parking delay information of the entrance lane based on the single-vehicle parking delay information of each spatiotemporal trajectory information. The queuing delay information calculation subunit is used to calculate the single-vehicle queuing delay information of the spatiotemporal trajectory information based on the effective parking trajectory segment and the crossing time point, and to calculate the queuing delay information of the entrance lane based on the single-vehicle queuing delay information of each spatiotemporal trajectory information. The fixed delay information calculation subunit is used to calculate the single-vehicle fixed delay information of the spatiotemporal trajectory information based on the signal interval information and the effective parking trajectory segment, and to calculate the fixed delay information of the approach lane based on the single-vehicle fixed delay information of each spatiotemporal trajectory information. The operation delay information calculation subunit is used to calculate the vehicle operation delay information of the spatiotemporal trajectory information based on the vehicle queuing delay information and the vehicle fixed delay information, and to calculate the operation delay information of the entrance lane based on the vehicle operation delay information of each of the spatiotemporal trajectory information.

[0143] This embodiment also provides an analysis device, including a memory, a processor, and a computer program stored in the memory and executable on the processor; wherein, when the processor executes the program, it implements the intersection delay analysis method based on small sample rate floating car trajectories as described in any of the above embodiments.

[0144] This embodiment also provides a computer-readable storage medium storing a computer program thereon, wherein when the program is executed by a processor, it implements the steps in the intersection delay analysis method based on small sample rate floating car trajectories as described in any of the above embodiments.

[0145] This embodiment also provides a computer program product, which includes computer instructions that, when executed by a processor, implement the steps in the intersection delay analysis method based on small sample rate floating car trajectories as described in any of the above embodiments.

[0146] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0147] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for intersection delay analysis based on floating car trajectories with a small sample rate, characterized in that, include: Obtain the stop line of at least one approach lane at the target intersection; Based on the stop line and trajectory data of multiple vehicles passing through the approach lane, the directed distance information of multiple sampling points in the trajectory data is determined; the directed distance information includes the distance and direction information from the sampling point to the stop line of the approach lane. Based on the directed distance information, the spatiotemporal trajectory information corresponding to the vehicle is constructed, and the geometric feature parameters of the spatiotemporal trajectory information are analyzed; Multiple spatiotemporal trajectory information is superimposed onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map; Based on the geometric feature parameters of each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory diagram, traffic wave information of the multi-vehicle spatiotemporal trajectory diagram is obtained, and based on the traffic wave information, the signal interval information of the approach lane is obtained; Based on the geometric feature parameters and the signal interval information, the delay index information of the approach lane is determined. The delay index information includes one or more of the following: parking delay information, queuing delay information, fixed delay information, and operational delay information. Based on the delay index information of each of the aforementioned approach lanes, the total delay index information of the target intersection is obtained. The total delay index information includes one or more of the following: total parking delay information, total queuing delay information, total fixed delay information, and total operational delay information.

2. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 1, characterized in that, Obtain the stop lines of at least one approach lane at the target intersection, including: The historical trajectory data of multiple vehicles passing through the target intersection within at least one historical sampling period are preprocessed to obtain preprocessed historical trajectory data. Multiple sampling points in each preprocessed historical trajectory data whose instantaneous speed is lower than the first preset threshold are projected onto the center line of the entrance lane of the corresponding target intersection to form a set of low-speed projection points. Density estimation is performed on the set of low-velocity projection points to obtain the spatial distribution density function corresponding to the set of low-velocity projection points. Extract the maximum points from the spatial distribution density function to form a set of candidate peak points; The stop line of the approach lane corresponding to the target intersection is determined based on the set of candidate peak points.

3. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 2, characterized in that, Also includes: Acquire trajectory data of multiple vehicles passing through the entrance lane within at least one new sampling period following the historical sampling period; Based on the trajectory data within the new sampling period, stop line estimation information is generated; Based on the stop line estimation information, the stop line corresponding to the entrance lane is calibrated.

4. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 1, characterized in that, Based on the stop line and trajectory data of multiple vehicles passing through the approach lane, the directed distance information of multiple sampling points in the trajectory data is determined, including: Based on the topological relationship information of the target intersection, calculate the matching degree score from the sampling point to each approach lane of the target intersection, and obtain the first approach lane with the highest matching degree score corresponding to the sampling point; The distance from the projection point of the sampling point on the center line of the first inlet channel to the corresponding stop line of the first inlet channel is determined as the distance information of the sampling point. Based on the upstream and downstream positional relationship of the sampling point relative to the stop line of the first inlet channel, the directional information corresponding to the sampling point is determined.

5. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 1, characterized in that, Based on the directed distance information, the spatiotemporal trajectory information corresponding to the vehicle is constructed, and the geometric feature parameters of the spatiotemporal trajectory information are analyzed, including: Based on the topological relationship information of the target intersection, sampling points belonging to the same vehicle and the same approach lane are associated as the same trip; Based on the time information of sampling points belonging to the same trip and the directed distance information, the spatiotemporal trajectory information is constructed. The spatiotemporal trajectory information contains multiple trajectory points, each of which corresponds one-to-one with a sampling point. Each trajectory point is composed of the time information and directed distance information of the corresponding sampling point. The geometric feature parameters of each trajectory point in the spatiotemporal trajectory information are analyzed. The geometric feature parameters include the multi-scale slope information, curvature information and slope consistency information of the trajectory point. The multi-scale slope information includes multiple local slope information of the trajectory point. The local slope information is the slope between the trajectory point and another trajectory point.

6. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 1, characterized in that, Based on the geometric feature parameters of each spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory diagram, traffic wave information of the multi-vehicle spatiotemporal trajectory diagram is obtained, and based on the traffic wave information, signal interval information of the approach lane is obtained, including: Based on the geometric feature parameters of the spatiotemporal trajectory information, a set of parking trajectory segments of the spatiotemporal trajectory information is formed; the set of parking trajectory segments includes at least one parking trajectory segment. Based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained, and the traffic wave information includes parking wave function and starting wave function; Based on the parking wave function and the starting wave function, the signal interval information corresponding to the approach lane is obtained.

7. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 6, characterized in that, Based on the geometric feature parameters and the signal interval information, the delay index information of the approach lane is determined, including: The set of parking trajectory segments is analyzed based on the signal interval information to determine the effective parking trajectory segments, which are parking trajectory segments generated due to signal waiting. Based on the geometric feature parameters of the spatiotemporal trajectory information, calculate the line crossing time points of the spatiotemporal trajectory information; Based on the signal interval information, the effective parking trajectory segment, and the crossing time point, the delay index information corresponding to the approach lane is obtained.

8. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 6, characterized in that, Based on the geometric feature parameters of the spatiotemporal trajectory information, a set of parking trajectory segments of the spatiotemporal trajectory information is formed, including: Based on the geometric feature parameters of each trajectory point in the spatiotemporal trajectory information, a feature vector of the trajectory point is constructed; The feature vector is analyzed using a first classification model to determine the motion state information of the trajectory point; Merge the trajectory segments between multiple adjacent trajectory points whose motion state information is in the parking state to generate a parking trajectory segment; Based on the parking trajectory segments, the parking trajectory segment set is formed.

9. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 6, characterized in that, Based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, the traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained, including: Based on the set of parking trajectory segments of multiple spatiotemporal trajectory information, the trajectory points at the start time of the first parking trajectory segment corresponding to the multiple spatiotemporal trajectory information are obtained to form a first trajectory point set; Obtain the trajectory points at the termination time of the last parking trajectory segment corresponding to multiple spatiotemporal trajectory information to form a second trajectory point set; The parking wave function is obtained based on the first set of trajectory points, and the starting wave function is obtained based on the second set of trajectory points.

10. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 6, characterized in that, Based on the set of parking trajectory segments for each of the spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, the traffic wave information of the multi-vehicle spatiotemporal trajectory map is obtained, including: Based on the multiple sets of parking trajectory segments containing the aforementioned spatiotemporal trajectory information, the start time trajectory points and end time trajectory points of all parking trajectory segments are obtained to form multiple sets of start and stop points. The multiple sets of start and stop points are divided into multiple candidate signal periods according to a preset time interval condition. The starting time trajectory points of each parking trajectory segment in each candidate signal period constitute the third trajectory point set of the candidate signal period, and the ending time trajectory points of each parking trajectory segment constitute the fourth time trajectory point set of the candidate signal period. For each candidate signal period, the parking wave function is obtained based on the third trajectory point set corresponding to the candidate signal period, and the starting wave function is obtained based on the fourth trajectory point set corresponding to the candidate signal period.

11. The intersection delay analysis method based on small sample rate floating car trajectories according to claim 9 or 10, characterized in that, Based on the parking wave function and the starting wave function, the signal interval information corresponding to the approach lane is obtained, including: Based on the parking wave function, the red light activation time corresponding to the multi-vehicle spatiotemporal trajectory map is obtained; Based on the start-up wave function, obtain the green light start time corresponding to the multi-vehicle spatiotemporal trajectory map; Based on the red light start time and green light start time of the same multi-vehicle spatiotemporal trajectory map, determine the red light interval information and red light duration information corresponding to the entrance lane; Based on the red light or green light turn-on time of the multi-vehicle spatiotemporal trajectory map of adjacent sampling periods, the signal cycle information corresponding to the entrance lane is determined; Based on the signal cycle information and the red light duration information, determine the green light ratio information corresponding to the approach lane; The signal interval information includes the signal cycle information, the red light interval information, the red light duration information, and the green light ratio information.

12. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 11, characterized in that, The method also includes: Obtain at least one historical trajectory data of the inlet channel within a historical sampling period that is in the same time period as the current sampling period; Based on the directed distance information of each trajectory point in the historical trajectory data and the directed distance information of each trajectory point in the trajectory data within the current sampling period, the historical trajectory data is aligned with the trajectory data within the current sampling period; An enhanced trajectory data set is formed based on the aligned historical trajectory data and the trajectory data; The enhanced parking wave function and enhanced start wave function of the enhanced trajectory data set are obtained, and the enhanced signal interval information is obtained based on the enhanced parking wave function and the enhanced start wave function. The enhanced signal interval information is used as the signal interval information of the approach lane.

13. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 12, characterized in that, The method also includes: Multiple enhanced trajectory data sets are acquired, wherein the multiple enhanced trajectory data sets are formed by aligning historical trajectory data with trajectory data within the current sampling period using different statistical windows, and the statistical window includes one or more historical sampling periods; Based on multiple sets of enhanced trajectory data, multiple enhanced signal interval information is obtained, and the enhanced signal interval information located at the median is used as the signal interval information of the approach lane.

14. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 11, characterized in that, The set of parking trajectory segments is analyzed based on the signal interval information to determine the valid parking trajectory segments, including: Obtain the overlap between the time interval information of the parking trajectory segment and the red light interval information; The parking trajectory segment that meets the preset overlap condition is determined as the valid parking trajectory segment.

15. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 7, characterized in that, Based on the geometric feature parameters of the spatiotemporal trajectory information, the time points at which the spatiotemporal trajectory information crosses the line are calculated, including: The first upstream trajectory point and the first downstream trajectory point are obtained from the spatiotemporal trajectory information; the first upstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located upstream of the stop line, and the first downstream trajectory point is the trajectory point with the smallest absolute value of directed distance information among all trajectory points located downstream of the stop line. Based on the geometric feature parameters of the first upstream trajectory point and the first downstream trajectory point, either a uniform interpolation model or an acceleration / deceleration interpolation model is selected to calculate the time point of the spatiotemporal trajectory information.

16. The intersection delay analysis method based on small sample rate floating car trajectory according to claim 7, characterized in that, Based on the signal interval information, the effective parking trajectory segment, and the crossing time point, the delay index information corresponding to the approach lane is obtained, including: Based on the effective parking trajectory segment, calculate the single-vehicle parking delay information of the spatiotemporal trajectory information, and based on the single-vehicle parking delay information of each spatiotemporal trajectory information, calculate the parking delay information of the entrance lane; Based on the effective parking trajectory segment and the crossing time point, calculate the single-vehicle queuing delay information of the spatiotemporal trajectory information, and calculate the queuing delay information of the entrance lane based on the single-vehicle queuing delay information of each spatiotemporal trajectory information. Based on the signal interval information and the effective parking trajectory segment, calculate the fixed delay information of a single vehicle in the spatiotemporal trajectory information, and calculate the fixed delay information of the approach lane based on the fixed delay information of a single vehicle in each spatiotemporal trajectory information. Based on the single-vehicle queuing delay information and the single-vehicle fixed delay information, the single-vehicle operation delay information of the spatiotemporal trajectory information is calculated, and based on the single-vehicle operation delay information of each of the spatiotemporal trajectory information, the operation delay information of the entrance lane is calculated.

17. A device for analyzing intersection delays based on floating car trajectories with a small sample rate, characterized in that, include: The stop line acquisition module is used to acquire the stop line of at least one approach lane of the target intersection; A directed distance acquisition module is used to determine directed distance information of multiple sampling points in the trajectory data based on the stop line and trajectory data of multiple vehicles passing through the approach lane; the directed distance information includes the distance information and direction information from the sampling point to the stop line of the approach lane; The spatiotemporal trajectory construction module is used to construct the spatiotemporal trajectory information corresponding to the vehicle based on the directed distance information, and to analyze the geometric feature parameters of the spatiotemporal trajectory information; The spatiotemporal trajectory overlay module is used to overlay multiple spatiotemporal trajectory information onto the same coordinate system to form a multi-vehicle spatiotemporal trajectory map; The signal interval acquisition module is used to obtain traffic wave information of the multi-vehicle spatiotemporal trajectory map based on the geometric feature parameters of each spatiotemporal trajectory information in the multi-vehicle spatiotemporal trajectory map, and to obtain the signal interval information of the approach lane based on the traffic wave information. The delay index analysis module is used to determine the delay index information of the approach lane based on the geometric feature parameters and the signal interval information. The delay index information includes one or more of the following: parking delay information, queuing delay information, fixed delay information, and operation delay information. The total delay index analysis module is used to obtain the total delay index information of the target intersection based on the delay index information of each of the approach lanes. The total delay index information includes one or more of the following: total parking delay information, total queuing delay information, total fixed delay information, and total operational delay information.

Images