CN115294757A - A lane-level traffic flow and traffic event recognition and release system - Google Patents

Abstract

The invention discloses a system for recognizing and releasing lane-level traffic flow and traffic events, and particularly relates to the technical field of fine recognition of traffic information and events, wherein the system comprises road sensing equipment, a cloud platform and data application; the road sensing equipment comprises a camera, a laser radar, a radar-vision all-in-one machine and a millimeter wave radar, and senses road information and outputs original data or structured data; the cloud platform comprises an edge platform of edge computing equipment and a central platform, wherein the edge platform is located near roadside equipment, the central platform is deployed in a data center server, the edge platform is uniformly and indiscriminately regarded as a whole, the cloud platform has the functions of data preprocessing, artificial intelligence algorithm and data publishing, the data preprocessing is in data connection with the artificial intelligence algorithm, and the artificial intelligence algorithm is in data connection with the data publishing. Different from the traditional traffic incident identification mode, the method has the advantages that the artificial intelligence algorithm is introduced to be applied to flow and traffic incident analysis, so that the accuracy and precision of wide-area flow and traffic incident information are greatly improved, and the application range of traffic incidents is expanded; meanwhile, the data is preprocessed, so that the computational power consumption of an artificial intelligence algorithm is greatly reduced, and the large-scale popularization of the method is possible; and (4) grading release is performed, and data benefits are utilized to the maximum extent.

Description

A Lane-Level Traffic Flow and Traffic Event Recognition and Release System

technical field

The invention relates to the technical field of refined traffic scene recognition, in particular to a lane-level traffic flow and traffic event recognition and release system.

Background technique

As the basis of smart transportation, the technical links of traffic big data mainly include perception, interconnection, analysis, prediction, control, etc.; in terms of perception, perception equipment represented by video and radar has been relatively mature and has been widely promoted. Technologies such as radar and lidar are also being gradually promoted and applied; network infrastructure based on 4G and 5G communications, optical fiber, and V2X networks enables reliable and high-speed interconnection of transportation facilities; through artificial intelligence algorithms such as image recognition and machine learning, combined with The traffic data cloud platform can realize traffic big data analysis and prediction.

Intelligentization and networking have become important trends in the development of the automobile industry, and vehicle-road coordination is its supporting technology; The side RSU and cloud platform realize the diversified release of traffic information and realize various scenarios.

The lane-level traffic flow and traffic event identification and release system based on artificial intelligence algorithms in the prior art has the problem of insufficient accuracy and precision of traffic flow and traffic event information:

1. In the field of commercial applications such as electronic navigation maps, traffic flow and traffic event information are mainly collected by mobile phone positioning information and analyzed to form traffic flow and congestion information. There are many problems in this processing method. First, there is the problem of incomplete data, that is, It is impossible to collect the information of all vehicles on the road, resulting in inaccurate traffic information; secondly, the accuracy rate is poor, and a large number of non-vehicle mobile phone positioning information near the road will also be identified as congestion, thereby misleading information users.

2. In terms of flow and congestion analysis methods based on road sensing equipment, there are mainly two technical paths: one is traditional traffic monitoring, using simple software to process video and radar data, mainly for road-level vehicle counting and other extensive types Function, the result is mainly road-level traffic data, without lane-level and real-time information, which cannot meet the requirements of refined traffic management and various commercial application scenarios; the other is based on vehicle-road coordination infrastructure, mainly for driving assistance, High-level autonomous driving and other directions have high requirements on the quality and processing of video and laser point clouds, and the investment in perception and computing equipment is huge, so it is difficult to promote it on a large scale in the short term to form the benefits of big data.

To this end, we propose a lane-level traffic flow and traffic event recognition and release system to solve the above problems.

Contents of the invention

The object of the present invention is to provide a lane-level traffic flow and traffic event recognition and release system to solve the problems raised in the above-mentioned background technology.

In order to solve the above-mentioned technical problems, the present invention adopts the following technical solutions: a lane-level traffic flow and traffic event recognition and release system, including:

Road perception devices, cloud platforms and data applications.

Preferably, the road sensing device is data-connected to the cloud platform.

The road perception device includes a camera, a Levision all-in-one machine, a laser radar and a millimeter-wave radar.

Preferably, the cloud platform includes data preprocessing, recognition algorithm and data classification, classification release, the data preprocessing includes data quality classification processing, frame extraction and fitting, and the recognition algorithm includes artificial intelligence recognition algorithm and fusion algorithm .

Preferably, the cloud platform is connected with data applications, and the data applications include but not limited to lane-level traffic flow indication, traffic violation monitoring, traffic control violation monitoring, lane recommendation, route recommendation, route planning, etc.

A lane-level traffic flow and traffic event identification and release system, the steps are as follows:

Step 1: Data collection: The data sources are mainly traffic cameras, supplemented by millimeter-wave radar, lidar, and Levision all-in-one machines; data types include sensor raw information, generated target information, etc.; directly access the platform through sensing devices , or through the connection and access between platforms, combining the advantages of traditional traffic and intelligent network data processing methods, applying artificial intelligence algorithms to wide-area road sensing device data processing.

Step 2: Data preprocessing: collect, clean and adapt various types of data, process video and lidar data through configurable frame extraction, and optimize data pixels, laser data points, gray scale and frame rate, both Real-time, lane-level high-precision requirements can be achieved, and the computing power consumption of artificial intelligence algorithms can be greatly reduced, making large-scale promotion possible.

Step 3: Artificial intelligence algorithm: through image recognition, the algorithm analyzes the data generated by preprocessing to generate lane-level target information. The identified target information mainly includes the target's position, speed, heading angle, etc., and the position information is passed through the artificial intelligence algorithm. and fusion algorithm to improve accuracy; the speed information comes from the direct output of the millimeter-wave radar, or is calculated based on the speed information calculated from the target position. On this basis, congestion analysis is performed to form lane-level traffic basic information.

Step 4: Data release: According to the data accuracy and real-time requirements, the lane-level traffic basic information is released in different grades and applied to various traffic applications.

Step 5: Pass the lane-level traffic basic information through a specific algorithm to generate lane-level traffic information, congestion information, etc. Further combining it with high-precision maps, recognized road markings, and red and green information, the application can be upgraded to the identification of traffic scenes. Typical applications include vehicle compaction lines and yellow lines, running red lights, and vehicles that are not polite to pedestrians. Including but not limited to the above scenarios.

Compared with prior art, the beneficial effect of the present invention is:

1. The present invention greatly improves the accuracy of wide-area traffic and traffic event information by introducing artificial intelligence algorithms and applying them to traffic and congestion analysis; through data preprocessing, the computing power consumption of artificial intelligence algorithms for lane-level traffic analysis is greatly reduced, Make it possible to promote on a large scale; through the release of files, it is applicable to different scenarios and maximizes the use of data benefits.

2. The present invention is mainly aimed at solving the problem of the accuracy of congestion data, combining the respective advantages of traditional traffic and intelligent network data processing methods, applying artificial intelligence algorithms to data processing of wide-area road sensing equipment; Carry out hierarchical preprocessing to rationalize the algorithm's requirements for computing power, and generate various real-time traffic flow statistics at the lane level. And on this basis, release vehicle congestion information products to the outside world, and derive applications such as lane recommendation, route recommendation, and route planning.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a schematic diagram of the overall structure of the present invention.

Fig. 2 is a schematic diagram of typical data processing in the present invention.

Fig. 3 is an example of partial location scenarios that can be recognized by the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments and drawings are only part of the embodiments of the present invention, not all of them. Examples are not intended to limit the application. . Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

Embodiment: As shown in Fig. 1-2, the present invention provides a kind of identification and release system of lane-level traffic volume and traffic event, comprising:

Road sensing devices, cloud platforms and data applications;

Road perception devices include cameras, all-in-one cameras, lidar and millimeter-wave radar, of which video and lidar are the main perception devices, and at least one of the two is required.

Furthermore, the camera is connected electromechanically to the LeVideo, the LeVe All-in-One is electrically connected to the laser radar, and the laser radar is electrically connected to the millimeter-wave radar.

Further, the road sensing device is connected with cloud platform data.

Furthermore, the cloud platform includes data preprocessing, recognition algorithms, data classification, classification and release, data preprocessing includes data quality classification processing, frame extraction and fitting, and recognition algorithms include artificial intelligence recognition algorithms and fusion algorithms.

Furthermore, the cloud platform is connected with data applications, and the data applications include lane-level traffic flow indication, traffic violation monitoring, and traffic control violation monitoring.

A lane-level traffic flow and traffic event identification and release system, the steps are as follows:

Step 1: Data collection: The data sources are mainly traffic cameras, supplemented by millimeter-wave radar, lidar, and Levision all-in-one machines; data types include sensor raw information, generated target information, etc.; directly access the platform through sensing devices , or through the connection and access between platforms, combining the advantages of traditional traffic and intelligent network data processing methods, applying artificial intelligence algorithms to data processing of wide-area road sensing devices, and applying artificial intelligence algorithms to traffic and congestion Analysis greatly improves the accuracy of wide-area traffic and traffic event information.

Step 2: Data preprocessing: collect, clean and adapt various types of data, process video and lidar data through configurable frame extraction, optimize, extract, and extract data pixels, laser data points, grayscale and frame rate Compression processing. How it is handled depends on the application requirements. For example, for a road with a low speed limit, the traffic flow statistics have relatively low real-time requirements, so more frames can be extracted to save more computing power; For small roads, each frame can be compressed to reduce the number of pixels and data points. The above processing can not only meet the real-time requirements of the scene and the high-precision requirements of the lane level, but also greatly reduce the computing power consumption of artificial intelligence algorithms.

Step 3: Artificial intelligence algorithm: through image recognition, the algorithm analyzes the data generated by preprocessing to generate lane-level target information. Generally, deep neural networks or related algorithms are mainly used for processing. The recognition results mainly include target information such as the target's position, speed, and heading angle. The position information can be fused with various sensor inputs through artificial intelligence algorithms and fusion algorithms to improve accuracy; the speed information can come from the output of the millimeter-wave radar. Or it can be calculated according to the target location information, and on this basis, the congestion analysis is carried out to form the lane-level traffic basic information.

Step 4: Data release: According to the data accuracy and real-time requirements, the lane-level traffic basic information is released in different grades and applied to various traffic applications. Depending on the location of the application, the data may be published to different platforms. The computing device can be used to directly send the application platform, or it can be collected through a specific platform and then released through the platform interface.

Step 5: Pass the lane-level traffic basic information through a specific algorithm to generate lane-level traffic information, congestion information, etc. Further combining it with high-precision maps, recognized road markings, and red and green information, the application can be upgraded to the identification of traffic scenes. Typical applications include vehicle compaction lines and yellow lines, running red lights, and vehicles that are not polite to pedestrians. Including but not limited to the above scenarios.

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims (5)

1. A lane-level traffic flow and traffic event recognition and release system, characterized in that, comprising: road perception equipment, cloud platform and data applications; said cloud platform, data applications can be deployed in the edge computing equipment near the road The edge platform can also be deployed on the central platform of the data center server, and other software deployment carriers that can be used as the data preprocessing, recognition algorithm, and release. The description focuses on logical functions rather than its deployment location, and is collectively referred to as "cloud platform”, “data application”. 2. A lane-level traffic flow and traffic event recognition and publishing system as claimed in claim 1, wherein said road perception equipment includes a camera, a laser radar, a laser-vision integrated machine, and a middle part of a millimeter-wave radar or all kinds. 3. a kind of identification and release system of lane-level traffic flow and traffic event as claimed in claim 1, it is characterized in that, described cloud platform comprises data preprocessing, recognition algorithm and data classification, classification release, and described data preprocessing The processing includes data quality classification processing, frame extraction and fitting, and the identification algorithm includes artificial intelligence identification algorithm and fusion algorithm. 4. a kind of traffic flow of lane level as claimed in claim 1 and the identification and publishing system of traffic event, it is characterized in that, described road sensing device is connected with cloud platform data, and described cloud platform is connected with data application data, so The above data applications include lane-level traffic flow indication, traffic violation monitoring and traffic control violation monitoring. 5. A kind of identification and release system according to any one of claim 1-4, characterized in that the steps are as follows: Step 1: Data collection: The data sources are mainly traffic cameras or lidar, supplemented by millimeter-wave radar and Levision all-in-one machines; data types include sensor raw information, generated target information, etc.; directly access the platform through sensing devices , or access through connecting devices between platforms, combining the respective advantages of traditional traffic and intelligent network data processing methods, applying artificial intelligence algorithms to wide-area road sensing device data processing; Step 2: Data preprocessing: collecting, cleaning and adapting various types of data, including: processing video data through configurable frame extraction, optimizing or compressing video data pixels and gray levels, and extracting laser point clouds Thinning, fitting or compression, etc., form different levels of perception data, thereby generating application scenarios that support different precision and real-time performance; different application scenarios have different requirements for data quality, and the target characteristics that need to be identified for traffic statistics in the present invention are obvious , large volume, relatively low real-time requirements; Step 3: Recognition algorithm: Analyze the data generated by preprocessing through image recognition algorithm to generate lane-level target information. The recognized target information mainly includes the target's position, speed, heading angle, etc.; typically, the position information comes from artificial intelligence Algorithms and fusion algorithms sequentially process the input data, and the speed information comes from the direct output of the millimeter-wave radar, or is calculated based on the speed information calculated from the target position; Step 4: Data release: According to the classification of preprocessing methods, lane-level traffic flow and congestion information are formed, vehicle congestion information products are released externally, and applications such as lane recommendation, route recommendation, and route planning are derived; Step 5: Combining the identified traffic participant information with high-precision maps, identified road markings, and red and green information, the application can be upgraded to the identification of traffic scenes. Typical applications include vehicle compaction lines and yellow lines, Running a red light, vehicles not yielding to pedestrians, etc., including but not limited to the above scenarios.

Images