Digital road driving track data index calculation and evaluation method

Abstract

The invention belongs to the technical field of intelligent traffic, and particularly relates to a digital road driving track data index calculation and evaluation method and system.The digital road driving track data index calculation and evaluation method comprises the steps that traffic vehicle data are collected through various roadside sensing devices, the collected sensing data are integrated and processed to form the traffic vehicle data, and the traffic vehicle data are analyzed; the method comprises the following steps: acquiring data of a traffic vehicle, preparing off-line network data, then applying different data processing technologies to further analyze and process the acquired data to meet data processing requirements of different functions, and then processing related data through modules with different functions such as traffic vehicle fusion, global UID matching, driving track prediction and the like. According to the method, road vehicles can be tracked and positioned, the driving track condition of each road in an urban digital road area can be mastered, and the method has a basic supporting effect on improving the urban intelligent degree and realizing urban digital area management.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and in particular relates to a method and system for calculating and evaluating digital road driving trajectory data indicators. Background Art

[0002] The development of intelligent transportation is an important part of the future urban transportation system. It relies on advanced information technology, big data analysis, artificial intelligence and other scientific and technological means, and aims to achieve intelligent transportation management, personalized services and efficient operations. Summary of the invention

[0003] The present invention aims to provide a method and system for calculating and evaluating digital road driving trajectory data indicators, so as to play a basic supporting role for urban intelligent transportation and improve the intelligence level of urban transportation through the method for calculating and evaluating digital road driving trajectory data indicators.

[0004] To achieve the above objectives, the present invention adopts the following technical solutions:

[0005] A method for calculating and evaluating a digital road driving trajectory data index is provided, comprising:

[0006] Utilize various roadside sensing devices including cameras, lidar and millimeter-wave radar to collect multi-source sensor data indicators;

[0007] Integrate and process the multi-source sensor data indicators to form traffic vehicle data, and also prepare off-line network data;

[0008] Further analyzing and processing the traffic vehicle data and off-road network data to meet the data usage requirements of driving trajectory assessment;

[0009] Based on the previously analyzed and processed data, the road driving status is evaluated and driving trajectory information is generated;

[0010] The driving trajectory information is visualized and an information interface is set for use in developing related applications.

[0011] Preferably, the further parsing and processing of the traffic vehicle data and off-line network data includes:

[0012] The traffic vehicle data integrated with multi-source sensor information is analyzed and processed through the Kalman filter to obtain the data of driving status and driving trajectory estimation;

[0013] Identify the data of the same traffic vehicle collected by different sensors in the traffic vehicle data by UID re-identification;

[0014] Based on the traffic vehicle data, the driving trajectory data of the traffic vehicles is predicted for the blind spot road section which cannot be perceived by the sensor through a bidirectional PID prediction method or a hidden Markov chain statistical model.

[0015] Preferably, the evaluating of the driving state based on the previously analyzed and processed data and generating driving trajectory information comprises:

[0016] Based on the Kalman filter, the traffic vehicle data is integrated by fusing the driving status and driving trajectory estimation data obtained;

[0017] Based on the data of the same traffic vehicle identified by UID re-identification method, global UID matching of vehicle driving data is performed;

[0018] Based on the traffic vehicle trajectory data predicted by the bidirectional PID, the vehicle's trajectory in the perception blind spot section is predicted.

[0019] Preferably, the visual display of the driving trajectory information includes a visual display of a historical trajectory and a visual display of a predicted trajectory.

[0020] Preferably, the collection of multi-source sensor data indicators, the integration of the traffic vehicle data, the analysis and processing of the traffic vehicle data and off-line network data, the evaluation of road driving conditions and the generation of driving trajectory information are all assisted by external AI algorithms and cloud computing resources.

[0021] The present invention also provides a digital road driving trajectory data index calculation and evaluation system, comprising:

[0022] The perception layer uses various roadside perception devices including cameras, lidar and millimeter-wave radar, combined with external computing resources, to collect multi-source sensor data indicators;

[0023] A data layer, which integrates and processes the multi-source sensor data indicators to form traffic vehicle data, and also prepares off-line network data;

[0024] A technical layer, which further analyzes and processes the traffic vehicle data and off-road network data to meet the data usage requirements of driving trajectory evaluation;

[0025] A functional layer, which evaluates the road driving status and generates driving trajectory information based on the previously analyzed and processed data;

[0026] The application layer visualizes the driving trajectory information and sets an information interface for use in developing related applications.

[0027] Preferably, the technical layer includes:

[0028] A Kalman filter is used to analyze and process traffic vehicle data that integrates multi-source sensor information through the Kalman filter to obtain data on driving status and driving trajectory estimation;

[0029] A UID re-identification module, used to identify the data of the same traffic vehicle collected by different sensors in the traffic vehicle data by means of UID re-identification;

[0030] The bidirectional PID prediction module is used to predict the driving trajectory data of traffic vehicles on the blind spot road section that cannot be perceived by the sensor based on the traffic vehicle data through a bidirectional PID prediction method or a hidden Markov chain statistical model.

[0031] Preferably, the functional layer comprises:

[0032] Traffic vehicle fusion module, used to fuse traffic vehicle data based on the driving status and driving trajectory estimation data obtained by Kalman filter fusion;

[0033] A global UID matching module is used to perform global UID matching of vehicle driving data based on the data of the same traffic vehicle identified by the UID re-identification method;

[0034] The vehicle trajectory prediction module is used to predict the vehicle trajectory data based on the bidirectional PID prediction.

[0035] Preferably, the application layer includes a historical trajectory visualization display module for visually displaying the historical trajectory, and a predicted trajectory visualization display module for visually displaying the predicted trajectory.

[0036] Preferably, it also includes a computing resource connection module, and the collection of multi-source sensor data indicators, the integration of traffic vehicle data, the analysis and processing of the traffic vehicle data and off-line network data, the evaluation of road driving status and the generation of driving trajectory information are all assisted by obtaining external AI algorithms and cloud computing resources through the computing resource connection module.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows: the digital road driving trajectory data index calculation and evaluation method uses various roadside sensing devices such as cameras, laser radars and millimeter wave radars to collect traffic vehicle data, integrates and processes the collected sensor data to form traffic vehicle data, and prepares off-line network data, and then applies Kalman filtering, UID re-identification, bidirectional PID prediction and other technologies to further analyze and process the sorted data to meet the data processing requirements of different functions, and then processes the relevant data through modules with different functions such as traffic vehicle fusion, global UID matching, driving trajectory prediction, etc., and visualizes the data processed by all functions, and opens interfaces for subsequent application development. This method focuses on digital roads, uses terminal devices in the area, collects traffic vehicle information through various roadside sensing devices, analyzes the data, realizes vehicle tracking and positioning, and grasps the driving trajectory of each road in the urban digital road area, which has a basic supporting role in improving the degree of urban intelligence and realizing urban digital regional management. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0039] Figure 1 Flow chart of an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators according to the present invention.

[0040] Figure 2 This is a code diagram of the Kalman filter used in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0041] Figure 3 This is an example chart of Kalman filtering of traffic vehicle data in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0042] Figure 4 This is a code diagram of the IoU intersection-over-union recognition algorithm used in an embodiment of a method for calculating and evaluating digital road vehicle trajectory data indicators of the present invention.

[0043] Figure 5 This is a code diagram of a method for calculating and evaluating digital road driving trajectory data indicators using a PID algorithm according to an embodiment of the present invention.

[0044] Figure 6 A diagram of the architecture of a hidden Markov model in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0045] Figure 7This is an example chart of traffic vehicle data fusion in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0046] Figure 8 This is an example diagram of the vehicle trajectory prediction effect in an embodiment of the digital road driving trajectory data index calculation and evaluation method of the present invention.

[0047] Fig. 9 This is a complete trajectory effect diagram of a vehicle with uid=9 in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0048] Fig.10 This is a complete trajectory effect diagram of a vehicle with uid=29 in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0049] Fig.11 This is a complete trajectory effect diagram of a vehicle with uid=786 in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0050] Fig.12 The present invention provides a method for calculating and evaluating digital road driving trajectory data indices in one embodiment, which is a point cloud diagram of vehicle status at a certain intersection.

[0051] Fig.13 This is a block diagram of an example of vehicle position in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0052] Fig.14 This is an example diagram of the vehicle trajectory evaluation effect on a certain road section in an embodiment of the digital road driving trajectory data index calculation and evaluation method of the present invention.

[0053] Fig.15 This is a visualization example chart of vehicle driving data in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0054] Fig.16 This is one of the algorithm code diagrams for cyclically reading offline data of sensors by combining an external AI algorithm and cloud computing resources in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0055] Fig.17 This is the second algorithm code diagram for cyclically reading offline data of sensors by combining an external AI algorithm and cloud computing resources in an embodiment of a method for calculating and evaluating digital road driving trajectory data indicators of the present invention.

[0056] Fig.18 Schematic diagram of an embodiment of a digital road driving trajectory data index calculation and evaluation system of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] In one embodiment, a method for calculating and evaluating a digital road driving trajectory data index is provided. Figure 1 As shown, the digital road driving trajectory data index calculation and evaluation method includes the following steps:

[0059] Step S100: Utilize various roadside sensing devices including cameras, laser radars, and millimeter-wave radars to collect multi-source sensing data indicators.

[0060] Step S200: Integrate and process the multi-source sensor data indicators to form traffic vehicle data, and also prepare off-line network data.

[0061] Step S300: further analyze and process the traffic vehicle data and off-road network data to meet the data usage requirements for driving trajectory evaluation.

[0062] Step S400: Evaluate the road driving status based on the previously analyzed and processed data and generate driving trajectory information.

[0063] Step S500: Visually display the driving trajectory information and set an information interface for use in developing related applications.

[0064] This digital road driving trajectory data indicator calculation and evaluation method takes digital roads as the key scenario, utilizes terminal equipment in the area, mainly through various roadside sensing devices such as cameras, lidar and millimeter wave radar to collect traffic vehicle information, and analyzes the data to achieve vehicle tracking and positioning, and grasp the driving trajectory of each road in the urban digital road area. This method plays a fundamental supporting role in improving the level of urban intelligence and realizing urban digital regional management.

[0065] Optionally, the aforementioned further analysis and processing of traffic vehicle data and off-road network data includes: analyzing and processing traffic vehicle data that integrates multi-source sensor information through a Kalman filter to obtain data on driving status and driving trajectory estimation; identifying data of the same traffic vehicle collected by different sensors in traffic vehicle data through UID re-identification; based on traffic vehicle data, predicting traffic vehicle driving trajectory data for blind spots that cannot be perceived by sensors through a two-way PID prediction method.

[0066] Kalman filter is an algorithm that outputs observation data through system input and makes the best estimate of the system state. Since the observation data includes the influence of noise and interference in the system, the optimal estimate can also be regarded as a filtering process. Kalman filter analyzes the detection system by establishing a state equation through sensor data and outputs the detection results. It then predicts and estimates the state of the system based on the input and output data, and continuously updates the probability value of the detection target according to the updated value of the next step, converting the multi-sensor data fusion into an optimized data. The code of Kalman filter is as follows: Figure 2 As shown. Combined Figure 3 As shown, Figure 3 Example diagram for Kalman filtering of traffic vehicle data. Figure 3 The parameters in are longitude, latitude, vehicle trajectory, timestamp (normalized), fitted curve, and data points.

[0067] UID re-identification is based on the timestamp and IoU intersection and union to identify the same traffic vehicle collected by different sensors. Intersection over Union (IoU) is an algorithm that calculates the overlap ratio of different image targets. Intuitively speaking, IoU is the ratio of the intersection and union of two graphic areas. When using IoU for identification, the vehicle activity range at the intersection is set. First, different vehicles are determined based on the timestamp. Then, the targets under different timestamps are determined based on the IoU distance. Finally, the same vehicle is identified as the same UID. The code of the IoU intersection and union identification algorithm is as follows: Figure 4 shown.

[0068] Bidirectional PID prediction is based on the following assumptions: 1. After passing through an intersection, traffic vehicles will not stop until they reach the next intersection; 2. Traffic vehicles travel along lanes; 3. Traffic vehicles will not exceed the speed limit. Under the premise of meeting the above assumptions, the driving trajectory is predicted and implemented using the bidirectional PID tracking method. Proportional Integral Derivative (PID) control is the most widely used control algorithm in industry and is often used in tracking control, motor control and other fields. On urban roads, traffic vehicles traveling along the road usually use the center line of the lane or the exit point of the intersection as the staged driving target point, which is consistent with the characteristic of the PID controller that the control quantity converges to a certain target value; and the speed control of traffic vehicles itself is a typical feedback control process, which can also be simulated by the PID controller, and the driving trajectory is predicted through bidirectional PID tracking. The PID algorithm code is as follows Figure 5 shown.

[0069] The hidden Markov chain statistical model can also be used to predict the vehicle trajectory in the area without sensor coverage based on the perception data of roadside sensors:

[0070] The Hidden Markov Model (HMM) statistical model is used to describe a Markov process with hidden unknown parameters. HMM can process time series data and sequence data by describing the probabilistic relationship between hidden states and observed states. This model has a wide range of applications in speech recognition, bioinformatics, natural language processing and other fields. In terms of trajectory data modeling, HMM can be used to analyze and predict the movement trajectory of moving objects (such as vehicles, animals, etc.). By combining the observed trajectory data with the model, the behavior pattern of the object in different states can be inferred. For example, in traffic flow analysis, HMM can be used to identify and predict the driving state of the vehicle, such as acceleration, deceleration, parking, etc.

[0071] HMM is one of the most popular trajectory prediction methods based on machine learning. HMM is also a policy-based method using Markov chain. Markov chain refers to a process containing finite events. The state of the system at time t+1 is only related to the previous time t, and the state transition probability is independent of time. The mathematical expression is:

[0072] P(S n+1 =s|S1=s1,S2=s2…,S n =s n )

[0073] =P(S n+1 =s|S n =s n ).

[0074] In real life, we can only observe the unique state exposed on the surface, but there is no intuitive representation of its hidden state. Therefore, it is necessary to establish a Markov process with hidden states and obtain the basic state of the event through the observable state set related to the hidden state probability. This is the so-called hidden Markov model. Fig.18 , HMM is represented by (S, O, A, B, π) as follows:

[0075] S={S1,S2,…S N}represents the hidden states se-quence.

[0076] O={O1,O2,…O M}represents the observation se-quence.

[0077] ·A represents the transition probability matrix between hiddenstates.

[0078] ·B is the output matrix, representing the transition proba-bility ofhidden states to output states.

[0079] ·πis the initial probability matrix,representing the initial probability distribution in hidden states.

[0080] in:

[0081] S={S1,S2,...S N} represents the hidden state sequence.

[0082] O={O1,O2,…O M} represents the observation sequence.

[0083] A represents the transition probability matrix between hidden states.

[0084] B is the output matrix, which represents the transition probability from hidden state to output state.

[0085] π is the initial probability matrix representing the initial probability distribution in the hidden state.

[0086] When HMM is used for trajectory prediction, the historical states of traffic participants are represented by the observation sequence O, and HMM solves the most likely future observation sequence.

[0087] The steps for predicting vehicle trajectories in areas without sensor coverage based on the perception data of roadside sensors using a hidden Markov chain statistical model are as follows:

[0088] (1) Problem definition: Clarify the goal of trajectory prediction, including the prediction time range, prediction accuracy requirements, etc.

[0089] (2) Data collection: Collect relevant trajectory data, which may include timestamp, location coordinates, speed, direction, etc.

[0090] (3) Data preprocessing:

[0091] Clean the data to remove outliers and noise.

[0092] Normalization is performed to ensure that the data are on the same scale.

[0093] Handle missing values, possibly by interpolating or using a model to estimate the missing data.

[0094] (4) Feature extraction:

[0095] Extract useful features such as position, velocity, acceleration, etc. from raw trajectory data.

[0096] Feature engineering, including feature selection and feature construction, may be required.

[0097] (5) State definition: In HMM, state refers to various possible situations of the system. In trajectory prediction, the state can be the position, speed, and direction of the vehicle.

[0098] (6) Model initialization:

[0099] Define the state transition probability matrix, which is the core part of the model and describes the transition probability between states.

[0100] Define the observation probability matrix, which is the probability of observing different trajectory features in a given state.

[0101] Set the initial state probability distribution.

[0102] (7) Model training:

[0103] Use the EM algorithm (Expectation Maximization algorithm) or other optimization algorithms to train the model parameters until convergence.

[0104] Several iterations may be required to optimize the model parameters.

[0105] (8) Model evaluation:

[0106] The validation set is used to evaluate the performance of the model. Common evaluation indicators include accuracy, recall, and F1 score.

[0107] Adjust model parameters or structure to improve the model's predictive performance.

[0108] (9) Trajectory prediction:

[0109] The trained HMM model is used to predict the most likely trajectory path through the Viterbi algorithm or other decoding algorithms.

[0110] It may be necessary to consider the integration of real-time data for online prediction.

[0111] (10) Post-processing:

[0112] The prediction results are smoothed to reduce the prediction error.

[0113] It may be necessary to combine other information, such as map data, traffic rules, etc., to optimize the prediction results.

[0114] (11) Application Integration:

[0115] Integrate the HMM model into practical applications, such as autonomous driving systems, traffic monitoring systems, etc.

[0116] Ensure that the model can run stably in the actual environment and provide real-time predictions.

[0117] (12) Continuous Optimization:

[0118] The model is continuously adjusted and optimized based on feedback from actual applications.

[0119] It may be necessary to consider extending the model to accommodate more complex prediction scenarios.

[0120] Throughout the process, multiple iterations and adjustments may be required to ensure that the model can accurately predict trajectory data. In addition, with the development of deep learning technology, it is possible to consider combining HMM with other models (such as CNN, RNN, LSTM) to improve the accuracy and robustness of prediction.

[0121] Optionally, the aforementioned evaluation of the driving state and generation of driving trajectory information based on the previously analyzed and processed data includes: fusing the driving state and driving trajectory estimation data obtained based on the Kalman filter, fusing the traffic vehicle data, and combining Figure 7 As shown, Figure 7This is an example chart for traffic vehicle data fusion, where the parameters of the table are: timestamp, target ID, latitude, longitude, device, and uid. Then, based on the data of the same traffic vehicle identified by the UID re-identification method, global UID matching of vehicle driving data is performed.

[0122] Based on the traffic trajectory data predicted by the bidirectional PID, the vehicle's trajectory is predicted in the blind spot section. Figure 8-10 , Figure 8 This is the vehicle trajectory prediction effect diagram. Figure 8 The parameters are: latitude, longitude, destination, and vehicle position trajectory. Fig. 9 This is the complete trajectory effect diagram of the vehicle with uid=9; Fig.10 This is the complete trajectory effect diagram of the vehicle with uid=29; Fig.11 This is the complete trajectory effect diagram of the vehicle with uid=786.

[0123] Optionally, the aforementioned visual display of driving trajectory information includes visual display of historical trajectories and visual display of predicted trajectories. Figure 12-14 , Fig.12 A point cloud diagram of the vehicle status at a certain intersection; Fig.13 is a block diagram of an example of vehicle position; Fig.14 This is a diagram showing the evaluation effect of vehicle trajectories on a road section. Fig.14 The parameters are: DataPoints, Scatter Plot of Latitude and Longitudewith Timestamps, timestamp, latitude, longitude. Fig.15 This is a sample visualization chart of vehicle driving data. Fig.15 The parameters are: timestamp, box, category, yaw, latitude, longitude, velocityX, and velocityY.

[0124] Optionally, the aforementioned multi-source sensor data index collection, traffic vehicle data integration, traffic vehicle data and off-line network data analysis and processing, road driving status evaluation and driving trajectory information generation are all combined with external AI algorithms and cloud computing resources for auxiliary processing. Among them, the algorithm code for cyclically reading the offline data of the sensor by combining external AI algorithms and cloud computing resources is as follows: Fig.16 and Fig.17 shown.

[0125] The data calculation in the previous digital road driving trajectory data index calculation evaluation method adopts distributed traffic index calculation. This is mainly because the process of calculating traffic indexes based on trajectory data faces the following problems: (1) Large data volume: Trajectory data includes information such as vehicle location, speed, driving direction and timestamp. Traditional centralized computing systems are difficult to handle such a large amount of data, while distributed computing can process data in parallel through multiple computing nodes, effectively improving data processing capabilities. (2) High real-time requirements: Traffic systems require real-time response. Distributed computing can speed up data processing and achieve real-time or near real-time traffic index calculation. (3) Computational complexity: The calculation of traffic indicators often involves complex algorithms. Distributed computing can decompose complex computing tasks into multiple subtasks and assign them to different computing nodes for parallel execution, thereby improving computing efficiency. (4) System reliability and fault tolerance: Nodes in a distributed computing system can back up each other. When a node fails, other nodes can take over its tasks to ensure the continuity of computing and the stability of the system. (5) Scalability: As the scale of the traffic network expands, computing requirements will also increase. Distributed computing systems can easily expand computing capabilities by adding more computing nodes to adapt to growing computing needs. (5) Load balancing: Distributed computing can use load balancing technology to evenly distribute computing tasks to each node, avoiding overloading of some nodes while other nodes are idle, and improving overall computing efficiency.

[0126] Specifically, the Hadoop distributed system is selected for distributed computing. As a distributed platform that can handle massive amounts of data, Hadoop is an open source software framework. Its reliability, scalability, and distributed computing and storage advantages make it widely used. Hadoop provides distributed computing and distributed storage functions, and can build a stable, reliable, fault-tolerant, and highly available distributed cluster based on cheap machines. For users, it shields the underlying structure of distributed computing, allowing users to efficiently develop distributed applications with simple APIs. The two core parts of Hadoop are HDFS (distributed file system) and the computing framework based on MapReduce.

[0127] Among them, the distributed file storage adopts the HDFS distributed file system. HDFS (Hadoop Distributed File System) is a highly available distributed file system. Its core structure is also a master-slave structure. The master and slave nodes are synchronized through heartbeat detection. Requests are completed with the RPC protocol, and data transmission is completed through the https protocol. It builds a stable distributed file system based on cheap hardware, provides reliable file storage and file reading and writing services, and uses fault-tolerant strategies to ensure data security even in the event of hardware failures.

[0128] This method requires the study of massive trajectory data with a strict and regular data structure. The original data of this method is organized as the trajectory data of each vehicle during the collection time. The size of a single file is small, but the total number of files is large. If it is stored directly, the reading efficiency will be low, thus affecting the subsequent calculation steps. The files can be merged into large files, and the files can be split by the file system's own splitting strategy and stored in HDFS, which can reduce the read and write delays. The operations on the trajectory data in the study are all offline and are not sensitive to the read and write delays. All operations are limited to sequential reading, and there are no operations to modify the original data or read randomly. Therefore, HDFS is suitable for carrying the trajectory data used in this method.

[0129] The distributed computing framework uses the MapReduce computing framework. Hadoop's MapReduce computing framework is one of its core modules. It is based on the idea of ​​"divide and conquer" and takes the construction of a highly fault-tolerant and highly reliable parallel computing cluster on a large number of commercial machines as its starting point. It provides users with efficient batch processing services and makes the operation between users and the underlying layer transparent, so that users only need to focus on upper-level development, greatly improving the development efficiency of parallel applications. The MapReduce process is divided into control flow and data flow, where the control flow describes the process from job startup to operation, and the data flow describes the process of data input, processing, and output.

[0130] The main work in the offline part is data preprocessing and the establishment of prediction models. There is no strict requirement for real-time performance and the delay of Hadoop can be tolerated. In addition, in various tasks, such as data sorting and cleaning, trajectory data map matching, etc., the trajectory data can be split and calculated separately, which conforms to the idea of ​​"divide and conquer". Hadoop's MapReduce parallel computing framework is suitable for the work of processing trajectory data in this method.

[0131] In one embodiment, a digital road driving trajectory data index calculation and evaluation system is provided, combining Fig.18As shown, the system includes a perception layer, a data layer, a technical layer, a functional layer, and an application layer, wherein the perception layer utilizes various roadside perception devices including cameras, lidars, and millimeter-wave radars, combined with external computing resources, to collect multi-source sensor data indicators; the data layer integrates and processes the multi-source sensor data indicators to form traffic vehicle data, and also prepares off-road network data; the technical layer further analyzes and processes the traffic vehicle data and off-road network data to meet the data usage requirements for driving trajectory evaluation; the functional layer evaluates the road driving status based on the previously analyzed and processed data and generates driving trajectory information; the application layer visualizes the driving trajectory information and sets an information interface for use when developing related applications.

[0132] The digital road driving trajectory data index calculation and evaluation system is mainly divided into five layers with different functions. The system takes digital roads as the key scene and uses terminal equipment in the area, mainly through various roadside sensing devices such as cameras, lidar and millimeter-wave radar to collect traffic vehicle information, and analyze the data to achieve vehicle tracking and positioning, and grasp the driving trajectory of each road in the urban digital road area. This method plays a fundamental supporting role in improving the level of urban intelligence and realizing urban digital regional management.

[0133] Optionally, the aforementioned technical layer further includes a Kalman filter, a UID re-identification module, and a bidirectional PID prediction module, wherein the Kalman filter is used to parse and process the traffic vehicle data that integrates multi-source sensor information through the Kalman filter to obtain the data of driving status and driving trajectory estimation; the UID re-identification module is used to identify the data of the same traffic vehicle collected by different sensors in the traffic vehicle data through the UID re-identification method; the bidirectional PID prediction module is used to predict the driving trajectory data of the traffic vehicle on the blind spot road section that the sensor cannot perceive based on the traffic vehicle data through the bidirectional PID prediction method. The Kalman filter is an algorithm that outputs observation data through system input and makes the best estimate of the system state. Since the observation data includes the influence of noise and interference in the system, the best estimate can also be regarded as a filtering process. The UID re-identification module is based on the Timestamp timestamp and IoU intersection to identify the same traffic vehicle collected by different sensors. The bidirectional PID prediction module is based on the following assumptions: 1. After the traffic vehicle passes the intersection, it will not stop until the next intersection; 2. The traffic vehicle travels along the lane; 3. The traffic vehicle will not exceed the speed limit. Under the premise of meeting the above assumptions, the vehicle trajectory is predicted using a bidirectional PID tracking method.

[0134] Optionally, the aforementioned functional layer includes a traffic vehicle fusion module, a global UID matching module, and a driving trajectory prediction module, wherein the traffic vehicle fusion module is used to fuse traffic vehicle data based on driving status and driving trajectory estimation data obtained by Kalman filter fusion; the global UID matching module is used to perform global UID matching of vehicle driving data based on data of the same traffic vehicle identified by UID re-identification method; the driving trajectory prediction module is used to predict the driving trajectory of traffic vehicles based on the driving trajectory data of bidirectional PID prediction.

[0135] Optionally, the aforementioned application layer includes a historical trajectory visualization display module for visually displaying historical trajectories, and a predicted trajectory visualization display module for visually displaying predicted trajectories. The historical trajectory visualization display module and the predicted trajectory visualization display module can display intersection point cloud data, road driving trajectory information, driving predicted trajectory information, etc.

[0136] Optionally, the system also includes a computing resource connection module, which collects multi-source sensor data indicators, integrates traffic vehicle data, analyzes and processes traffic vehicle data and off-line network data, evaluates road driving status, and generates driving trajectory information. All of these are processed by obtaining external AI algorithms and cloud computing resources through the computing resource connection module. By connecting to external computing resources, cloud computing resources can be fully utilized to achieve lightweight processing of large amounts of data.

[0137] It should be noted that, in this article, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus.

[0138] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for calculating and evaluating digital road driving trajectory data indicators, characterized in that: include: Utilize various roadside sensing devices including cameras, lidar and millimeter-wave radar to collect multi-source sensor data indicators; Integrate and process the multi-source sensor data indicators to form traffic vehicle data, and also prepare off-line network data; Further analyzing and processing the traffic vehicle data and off-road network data to meet the data usage requirements of driving trajectory assessment; Based on the previously analyzed and processed data, the road driving status is evaluated and the driving trajectory information is generated; The driving trajectory information is visualized and an information interface is set for use in developing related applications.

2. The method for calculating and evaluating digital road driving trajectory data indicators according to claim 1, characterized in that: The further parsing and processing of the traffic vehicle data and off-line network data includes: The traffic vehicle data integrated with multi-source sensor information is analyzed and processed through the Kalman filter to obtain the data of driving status and driving trajectory estimation; Identify the data of the same traffic vehicle collected by different sensors in the traffic vehicle data by UID re-identification; Based on the traffic vehicle data, the driving trajectory data of the traffic vehicles is predicted for the blind spot road section which cannot be perceived by the sensor by means of a bidirectional PID prediction method or a hidden Markov chain statistical model.

3. The method for calculating and evaluating digital road driving trajectory data indicators according to claim 2, characterized in that: The evaluation of the driving state based on the previously analyzed and processed data and the generation of driving trajectory information include: Based on the Kalman filter, the traffic vehicle data is integrated by fusing the driving status and driving trajectory estimation data obtained; Based on the data of the same traffic vehicle identified by UID re-identification method, global UID matching of vehicle driving data is performed; Based on the traffic vehicle trajectory data predicted by the bidirectional PID, the vehicle's trajectory in the perception blind spot section is predicted.

4. The method for calculating and evaluating digital road driving trajectory data indicators according to claim 1, characterized in that: The visual display of the driving trajectory information includes a visual display of a historical trajectory and a visual display of a predicted trajectory.

5. The method for calculating and evaluating digital road driving trajectory data indicators according to claim 1, characterized in that: The collection of multi-source sensor data indicators, the integration of the traffic vehicle data, the analysis and processing of the traffic vehicle data and off-road network data, the evaluation of road driving conditions and the generation of driving trajectory information are all assisted by external AI algorithms and cloud computing resources.

6. A digital road driving trajectory data index calculation and evaluation system, characterized in that: include: The perception layer uses various roadside perception devices including cameras, lidar and millimeter-wave radar, combined with external computing resources, to collect multi-source sensor data indicators; A data layer, which integrates and processes the multi-source sensor data indicators to form traffic vehicle data, and also prepares off-line network data; A technical layer, which further analyzes and processes the traffic vehicle data and off-road network data to meet the data usage requirements of driving trajectory evaluation; A functional layer, which evaluates the road driving status and generates driving trajectory information based on the previously analyzed and processed data; The application layer visualizes the driving trajectory information and sets an information interface for use in developing related applications.

7. The digital road driving trajectory data index calculation and evaluation system according to claim 6 is characterized in that: The technical layer includes: A Kalman filter is used to analyze and process traffic vehicle data that integrates multi-source sensor information through the Kalman filter to obtain data on driving status and driving trajectory estimation; A UID re-identification module, used to identify the data of the same traffic vehicle collected by different sensors in the traffic vehicle data by means of UID re-identification; The bidirectional PID prediction module is used to predict the driving trajectory data of traffic vehicles on the blind spot road section that cannot be perceived by the sensor based on the traffic vehicle data through a bidirectional PID prediction method or a hidden Markov chain statistical model.

8. The digital road driving trajectory data index calculation and evaluation system according to claim 7 is characterized in that: The functional layer comprises: Traffic vehicle fusion module, used to fuse traffic vehicle data based on the driving status and driving trajectory estimation data obtained by Kalman filter fusion; A global UID matching module is used to perform global UID matching of vehicle driving data based on the data of the same traffic vehicle identified by the UID re-identification method; The vehicle trajectory prediction module is used to predict the vehicle trajectory data based on the bidirectional PID prediction.

9. The digital road driving trajectory data index calculation and evaluation system according to claim 6, characterized in that: The application layer includes a historical trajectory visualization display module for visually displaying the historical trajectory, and a predicted trajectory visualization display module for visually displaying the predicted trajectory.

10. The digital road driving trajectory data index calculation and evaluation system according to claim 6, characterized in that: It also includes a computing resource connection module, which is used to acquire external AI algorithms and cloud computing resources for auxiliary processing through the computing resource connection module for the collection of multi-source sensor data indicators, the integration of traffic vehicle data, the analysis and processing of the traffic vehicle data and off-line network data, the evaluation of road driving conditions and the generation of driving trajectory information.

Images