Lane toll collection system

Abstract

The invention provides a lane toll collection system, which relates to the technical field of traffic toll collection and comprises a global sensing layer, an edge intelligent layer, a cloud intelligent layer and a lane execution layer. The global sensing layer collects data based on the lane execution layer through a multi-modal sensing matrix, and uploads the data to the edge intelligent layer after the data is preprocessed by an edge computing node; the edge intelligent layer completes vehicle identification and rate calculation based on a distributed AI engine, and issues a decision instruction to the lane execution layer; wherein the cloud intelligent layer provides global data storage, model training and cross-domain cooperation functions, and communicates with the edge intelligent layer in real time through a 5G / optical fiber network; the multi-modal sensing matrix collects data based on the lane execution layer, the data are preprocessed by the edge computing node and then uploaded to the edge intelligent layer, the edge intelligent layer completes vehicle identification and rate calculation and issues a decision instruction to the lane execution layer, the architecture realizes rapid processing and efficient decision of the data, and the overall efficiency of the toll collection system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic toll collection, and particularly to a lane toll collection system. Background Art

[0002] With the progress of technology, the types and quantities of transportation vehicles have been increasing continuously. The traditional lane toll collection system has been difficult to meet the growing demands. In the traditional toll collection mode, manual operations are often required. This mode not only has low efficiency, is prone to human errors, but also requires a large amount of manpower and material resources, increasing the operation cost. Therefore, it has become very urgent to develop a lane toll collection system that can automatically identify vehicles and complete toll collection. Most of the existing lane toll collection systems rely on RFID (Radio Frequency Identification) technology or license plate recognition technology, but these systems still have some deficiencies. The popularization and use of RFID toll collection systems are restricted in some areas because vehicles must be installed with special RFID tags, while license plate recognition systems rely on complex cameras and image processing technologies, which also increases the complexity and maintenance cost of the systems. In addition, the recognition accuracy of these systems may also be affected under adverse weather conditions and light changes. Therefore, we propose a lane toll collection system. Summary of the Invention

[0003] The purpose of the present invention is to solve the deficiencies existing in the prior art, improve the toll collection efficiency, reduce human errors, and lower the operation cost.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions: A lane toll collection system includes a global perception layer, an edge intelligence layer, a cloud intelligence layer, and a lane execution layer; The global perception layer collects data based on the lane execution layer through a multi-modal perception matrix, and after preprocessing by an edge computing node, uploads it to the edge intelligence layer; The edge intelligence layer completes vehicle recognition and rate calculation based on a distributed AI engine, and issues decision instructions to the lane execution layer; Among them, the cloud intelligence layer provides global data storage, model training, and cross-domain collaboration functions, and communicates with the edge intelligence layer in real time through a 5G / fiber optic network.

[0005] Preferably, the multi-modal perception matrix includes visual perception, laser perception, and millimeter wave perception.

[0006] Preferably, the visual perception adopts a dynamic vision camera and an event camera; Among them, the dynamic vision camera adopts a global shutter sensor, supports 120dB HDR, adapts to strong light / backlight, and the event camera is based on the principle of biological vision and only records pixels with brightness changes; the power consumption is reduced to 10mW The laser perception adopts a 4D imaging radar and a solid-state lidar; Among them, the 4D imaging radar outputs a point cloud density ≥ 1000 points / ㎡, supports 100-meter distance detection, the horizontal field of view of the solid-state lidar is 120°, the vertical field of view is 30°, and the frame rate is 20Hz, which is used for lane-level modeling; The millimeter-wave perception adopts an ultra-wideband radar with a bandwidth of 1GHz, a ranging accuracy of ±2cm, and supports multi-target tracking.

[0007] Preferably, the multi-modal perception matrix performs feature fusion, maps visual features (color, texture), laser features (contour, reflectivity), and millimeter-wave features (speed, acceleration) to a unified feature space, and the recognition accuracy rate ≥ 99.9%; and through spatio-temporal correlation analysis, it complements some occluded license plate information.

[0008] Preferably, the distributed AI engine includes a lightweight neural network and a federated learning framework. The lightweight neural network adopts a MobileNetV4+Transformer hybrid architecture, the model compression rate ≥ 90%, and the inference speed < 50ms; The federated learning framework supports cross-toll station data collaborative training.

[0009] Preferably, the decision instruction is implemented through an adaptive decision module, and the adaptive decision module automatically adjusts the weights of the perception devices according to the weather, time period, and vehicle type.

[0010] Preferably, the cloud intelligent layer is based on a digital twin platform for real-time traffic simulation, generates a traffic flow heat map based on toll data, and predicts congestion risks; Device health prediction, through the analysis of sensor data streams, warns of device failures 72 hours in advance.

[0011] Preferably, the functions of the cross-domain collaboration center include: a V2X interface for real-time communication with in-vehicle terminals (OBUs) and roadside units (RSUs), supporting vehicle-road collaborative toll collection; A payment federated network that aggregates payment channels including but not limited to ETC, digital RMB, international credit cards, etc.

[0012] Compared with the prior art, the beneficial effects of the present invention are: First, improve the toll collection efficiency: Automated processing: Traditional lane toll collection systems rely on manual operations and are inefficient. The lane toll collection system of the present invention automatically collects data through the multi-modal perception matrix (visual, laser, millimeter-wave perception) of the global perception layer. The edge intelligent layer completes vehicle recognition and rate calculation based on the distributed AI engine, and issues decision instructions to the lane execution layer, realizing full-process automation and greatly improving the toll collection efficiency.

[0013] Quick decision-making: The lightweight neural network in the distributed AI engine adopts the MobileNetV4 + Transformer hybrid architecture, with a model compression rate of ≥90% and an inference speed of <50ms. It can quickly complete vehicle identification and rate calculation, reducing the time vehicles stay at toll gates.

[0014] 2. Reduce human errors: Accurate identification: Existing systems rely on manual operation and are prone to human errors. The system of the present invention uses a multimodal perception matrix for feature fusion, mapping visual features (color, texture), laser features (contour, reflectivity), and millimeter wave features (speed, acceleration) to a unified feature space, with an identification accuracy of ≥99.9%, effectively avoiding possible errors in manual identification.

[0015] Automatically complete information: Through spatiotemporal correlation analysis, partially obscured license plate information can be completed to ensure the accuracy of vehicle identification and reduce charging errors caused by incomplete information.

[0016] 3. Reduce operating costs: Reduce manpower input: Traditional systems require a lot of manual operations, which increases labor costs. The system of the present invention realizes automated charging, does not require a lot of manual intervention, reduces manpower input, and reduces labor costs.

[0017] Reduce maintenance costs: Existing systems are highly complex and have high maintenance costs. The equipment health prediction function of the system of the present invention uses sensor data stream analysis to warn of equipment failures 72 hours in advance, facilitating timely maintenance and repairs, thereby reducing operating losses and maintenance costs caused by equipment failures. At the same time, the cloud-based intelligent layer provides global data storage, model training, and cross-domain collaboration functions, and supports cross-toll station data collaborative training through a federated learning framework, enabling the system to be continuously optimized and improved, reducing additional costs caused by system upgrades and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A system block diagram of a lane toll collection system provided by the present invention. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant content. Several embodiments of the present invention are given. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0021] It should be noted that when an element is referred to as "fixedly provided on" another element, it can be directly on the other element or there may also be an intermediate element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intermediate element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are only for the purpose of illustration.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0023] Embodiment: As Figure 1 shown, the present invention provides a technical solution: a lane toll collection system, including a global perception layer, an edge intelligence layer, a cloud intelligence layer and a lane execution layer. The multi-modal perception matrix collects data based on the lane execution layer, preprocesses it through an edge computing node and uploads it to the edge intelligence layer. The edge intelligence layer completes vehicle identification, rate calculation and issues decision instructions to the lane execution layer. This architecture realizes the rapid processing of data and efficient decision-making, improving the overall efficiency of the toll collection system; The global perception layer collects data based on the lane execution layer through the multi-modal perception matrix, preprocesses it through an edge computing node, and uploads it to the edge intelligence layer; Specifically, the multi-modal perception matrix includes visual perception, laser perception and millimeter-wave perception.

[0024] More specifically, visual perception uses a dynamic vision camera and an event camera; Among them, the dynamic vision camera uses a global shutter sensor, supports 120dB HDR, adapts to strong light / backlight, and the event camera is based on the principle of biological vision and only records pixels with brightness changes; the power consumption is reduced to 10mW Laser perception uses a 4D imaging radar and a solid-state lidar; Among them, the 4D imaging radar outputs a point cloud density ≥ 1000 points / ㎡, supports 100-meter distance detection, the horizontal field of view of the solid-state lidar is 120°, the vertical field of view is 30°, and the frame rate is 20Hz, which is used for lane-level modeling; Millimeter-wave sensing uses ultra-wideband radar with a bandwidth of 1 GHz, a ranging accuracy of ±2 cm, and supports multi-target tracking; It should be noted that the layout of the sensing devices (location-device type-function description): Lane entrance - Dynamic vision camera + Event camera - Capture license plate, vehicle type, and color features; Above the lane - 4D imaging radar + Solid-state lidar - 3D modeling, vehicle speed measurement, anomaly detection; On both sides of the lane - Ultra-wideband millimeter-wave radar - High-precision ranging, pedestrian intrusion warning; What can be further supplemented is that on the lane ground - Piezoelectric film sensor - Detect the number of vehicle axles and weight (assist in vehicle type classification); During the implementation process, the working process of typical scenarios: Ordinary vehicle passing: The dynamic vision camera captures the license plate, and the event camera tracks the vehicle trajectory; The lidar generates a 3D contour, and the millimeter-wave radar measures the speed; After multi-modal data fusion, vehicle type classification is completed through the edge node.

[0025] Special scenario processing: Example 1: License plate occlusion: Combine the trajectory association of the front and rear frames (spatiotemporal correlation analysis) to complete the information of the occluded part; Example 2: New energy vehicle identification: Assist in judgment through the lidar reflectivity characteristics (such as a smooth roof); Based on the above, the multi-modal perception matrix performs feature fusion, maps visual features (color, texture), lidar features (contour, reflectivity), and millimeter-wave features (speed, acceleration) to a unified feature space, with an identification accuracy rate of ≥99.9%, and through spatiotemporal correlation analysis, complete the information of partially occluded license plates.

[0026] The edge intelligent layer completes vehicle identification and rate calculation based on the distributed AI engine, and issues decision instructions to the lane execution layer; For the distributed AI engine, this AI engine integrates a lightweight neural network and a federated learning framework. The lightweight neural network adopts a MobileNetV4 + Transformer hybrid architecture, with a model compression rate of ≥90% and an inference speed of <50 ms; MobileNetV4: Used for lightweight object detection (such as license plates, vehicle types).

[0027] Transformer: Used for processing time-series data (such as vehicle trajectory prediction).

[0028] Support cross-toll station data collaborative training through the federated learning framework to protect user privacy.

[0029] The decision instructions of the edge intelligence layer are implemented through an adaptive decision-making module, which automatically adjusts the weights of sensing devices according to weather, time period, and vehicle type; Dynamic weight allocation (adaptive decision-making module) uses a rule engine: adjusts the weights of sensing devices based on weather, time period, and vehicle type. Example: Rainy day: lidar weight +20%, visual camera weight -10%; Night: event camera weight +30%, millimeter-wave radar weight +15%.

[0030] Abnormal event detection: Abnormal speed: <10 km / h or >120 km / h.

[0031] Vehicle type mismatch: the difference between the lidar contour and the visual recognition result >20%.

[0032] Based on the above detections, a response mechanism is adopted to automatically trigger high-definition video capture and upload it to the cloud for review.

[0033] Among them, the cloud intelligence layer provides global data storage, model training, and cross-domain collaboration functions, and communicates with the edge intelligence layer in real time through 5G / fiber optic network; The cloud intelligence layer is based on a digital twin platform for real-time traffic simulation, generates a traffic flow heat map based on toll data, and predicts congestion risks; Device health prediction, through sensor data stream analysis, warns of device failures 72 hours in advance.

[0034] The cross-domain collaboration hub function includes: a V2X interface, which communicates with in-vehicle terminals (OBUs) and roadside units (RSUs) in real time to support vehicle-road collaborative toll collection; A payment federation network that aggregates payment channels including but not limited to ETC, digital RMB, international credit cards, etc.

[0035] Workflow of the present invention: When using a lane toll collection system, first, a vehicle enters the sensing range of the lane execution layer. The lane execution layer triggers the multi-modal perception matrix to start working. The multi-modal perception matrix is activated to collect visual information of the vehicle (such as color, texture), laser information (such as contour, reflectivity), and millimeter-wave information (such as speed, acceleration). The collected data is first preprocessed at the edge computing node to remove noise data and extract key features. Then, the preprocessed data is uploaded to the edge intelligence layer. After receiving the data, the distributed AI engine in the edge intelligence layer uses a lightweight neural network to identify the vehicle and calculates the toll according to the preset rate rules. At the same time, the adaptive decision-making module automatically adjusts the weights of the sensing devices according to information such as the current weather, time period, and vehicle type to ensure the accuracy of identification and the rationality of decision-making. After completing vehicle identification and toll calculation, the edge intelligence layer issues a decision instruction to the lane execution layer. The lane execution layer controls the actions of the toll collection devices according to the decision instruction, such as raising the railing, displaying toll information, etc. If the vehicle supports vehicle-road collaborative toll collection, it communicates with the on-vehicle terminal (OBU) and roadside unit (RSU) in real time through the V2X interface of the toll collection unit and the cloud intelligence layer to complete the payment process; During the operation of the system, the cloud intelligence layer receives the data of each toll station in real time, performs global data storage and model training, conducts real-time traffic simulation through the digital twin platform, generates a traffic flow heat map, predicts congestion risks, and adjusts traffic management strategies in a timely manner. At the same time, through the analysis of sensor data streams, equipment failures are predicted 72 hours in advance for timely maintenance and repair; Through the above workflow, the lane toll collection system of the present invention realizes efficient, accurate, and intelligent vehicle toll collection and management, effectively solving the problems existing in traditional and existing lane toll collection systems.

[0036] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A lane toll system, characterized in that: It includes a global perception layer, an edge intelligence layer, a cloud intelligence layer, and a lane execution layer; The global perception layer collects data based on the lane execution layer through a multi-modal perception matrix. After preprocessing by edge computing nodes, it is uploaded to the edge intelligence layer; The edge intelligence layer completes vehicle recognition and toll rate calculation based on a distributed AI engine, and issues decision instructions to the lane execution layer; Among them, the cloud intelligence layer provides global data storage, model training, and cross-domain collaboration functions, and communicates with the edge intelligence layer in real time through a 5G / fiber optic network.

2. The lane toll collection system according to claim 1, wherein: The multi-modal perception matrix includes visual perception, lidar perception, and millimeter-wave perception.

3. The lane toll system according to claim 2, wherein: The visual perception uses a dynamic vision camera and an event camera; Among them, the dynamic vision camera uses a global shutter sensor, supports 120dB HDR, and adapts to strong light / backlight. The event camera is based on the principle of biological vision and only records pixels with brightness changes; The lidar perception uses a 4D imaging lidar and a solid-state lidar; Among them, the 4D imaging lidar outputs a point cloud density ≥ 1000 points / ㎡, supports 100-meter distance detection, the solid-state lidar has a horizontal field of view of 120°, a vertical field of view of 30°, and a frame rate of 20Hz, and is used for lane-level modeling; The millimeter-wave perception uses an ultra-wideband radar with a bandwidth of 1GHz, a ranging accuracy of ±2cm, and supports multi-target tracking.

4. The lane toll collection system according to claim 2, characterized in that: The multi-modal perception matrix performs feature fusion, maps visual features (color, texture), lidar features (contour, reflectivity), and millimeter-wave features (speed, acceleration) to a unified feature space, and completes part of the occluded license plate information through spatio-temporal correlation analysis.

5. A lane toll collection system according to claim 1, characterized in that: The distributed AI engine includes a lightweight neural network and a federated learning framework. The lightweight neural network adopts a MobileNetV4+Transformer hybrid architecture, with a model compression rate ≥ 90% and an inference speed < 50ms; Supports cross-toll station data collaborative training through the federated learning framework.

6. The lane toll collection system according to claim 1, wherein: The decision instruction is implemented through an adaptive decision module, which automatically adjusts the weights of perception devices according to weather, time period, and vehicle type.

7. A lane toll collection system according to claim 1, characterized in that: The cloud intelligence layer is based on a digital twin platform for real-time traffic simulation, generates a traffic flow heat map based on toll data, and predicts congestion risks; Equipment health prediction, analyzes the sensor data stream to predict equipment failures 72 hours in advance.

8. A lane toll system according to claim 6, characterized in that: The cross-domain collaboration center function includes: a V2X interface, communicates with in-vehicle terminals (OBUs) and roadside units (RSUs) in real time, and supports vehicle-road collaborative toll collection; A payment federated network aggregates payment channels including but not limited to ETC, digital RMB, international credit cards, etc.

Images