A method and system for expressway ramp accident early warning and guidance based on multi-sensor fusion

Abstract

The application discloses a kind of expressway ramp accident early warning and guiding method and system based on multi-sensor fusion, by dividing ramp into multiple functional areas and deploying infrared sensor, pressure sensor, sound sensor, camera and various warning devices;For each vehicle entering ramp, start timer to monitor the duration of passage, while collecting abnormal sound simultaneously;When the duration of passage of any vehicle exceeds threshold or abnormal sound is detected, determine accident or stagnation state;Immediately start ramp warning light and guardrail light strip to prompt deceleration and avoidance, while driving main line LED signboard to display no entry and early lane change information, guide main line vehicle to detour, and automatically push warning to traffic management platform.The application uses per-vehicle independent timing, two-way counting and acoustic triggering triple determination logic, realizes accurate identification, hierarchical guidance and coordinated reporting of ramp accident, effectively reduces the risk of secondary accident.

Description

Technical Field

[0001] This invention relates to the field of highway traffic condition analysis and intelligent traffic control technology, and in particular to a method and system for early warning and guidance of highway ramp accidents based on multi-sensor fusion. Background Technology

[0002] Accidents are frequent in the connecting ramp area between toll stations and the main road due to abrupt changes in road alignment indicators, steep drops in speed gradients, and high complexity of driving behavior. Among exit ramp accidents, single-vehicle collisions with guardrails and multi-vehicle rear-end collisions caused by speeding, illegal lane changes, illegal parking, and fatigued driving are particularly prominent.

[0003] Currently, accident warning methods mainly rely on warning triangles, which have drawbacks such as delayed response, limited warning range, and poor effectiveness at night and in low visibility conditions, easily leading to serious secondary accidents. Several technical solutions have been proposed for traffic safety in ramp areas. For example, Chinese patent CN118116213A discloses an automatic vehicle congestion diversion system for highway ramps, which uses a vehicle input / output acquisition unit to achieve bidirectional counting and issues congestion warnings based on the number of vehicles in the ramp. Chinese patent CN210000305U discloses an automatic vehicle stagnation alarm system, which uses inductive loops and time relays to trigger an alarm when a vehicle continuously occupies the ramp for a set time. In addition, some toll station-related patents, such as CN202422158467.0 and CN202321760765.6, mainly focus on toll management informatization, ramp congestion distance detection, or remote control of barriers, rather than rapid detection and tiered guidance in accident situations.

[0004] However, existing technologies still have the following drawbacks: First, the application scenarios are vague. Existing technologies related to ramps or toll stations either target service area ramps or general interchange ramps, or focus on the management and congestion detection at the toll station site. There is a lack of accident early warning solutions specifically for the "ramp connecting the toll station and the main road." Toll station ramps have unique characteristics such as high traffic volume, mixed vehicle types (including a large number of large trucks), a steep drop in vehicle speed from over 100 km / h on the main road to 40 km / h or even stopping to retrieve cards on the ramp, frequent vehicle stops and starts, and a high incidence of illegal and delayed incidents. These characteristics require higher requirements for the timeliness, accuracy, and guidance strategies of accident early warning.

[0005] Secondly, the judgment logic is simplistic, resulting in a high false alarm rate. Existing counting-based solutions can only determine the number of vehicles in the ramp, but cannot identify the specific vehicle's dwell time; single-time-based solutions (such as inductive loop detectors + time relays) treat the ramp as a whole, with the timer constantly resetting or accumulating as multiple vehicles pass through continuously, failing to distinguish between normal passage of multiple vehicles and abnormal dwell time of a single vehicle; pure video analysis solutions are significantly affected by rain, fog, and insufficient lighting at night, and their algorithms are complex and costly. None of the above solutions solve the problem of accurately determining the dwell time of a single vehicle in continuous traffic scenarios.

[0006] Third, the guidance strategies are limited and lack hierarchical coordination. Existing accident or congestion warning signals are mostly only output to the audio-visual equipment within toll booths or localized areas of ramps, failing to transmit warning information to the main road section. This results in vehicles approaching at high speeds from behind not being aware of abnormalities ahead, losing valuable deceleration and avoidance buffer windows. This is one of the key contributing factors to the frequent secondary accidents at toll booth ramps.

[0007] Fourth, there is insufficient coordination with traffic management platforms. Most existing solutions are isolated systems that fail to achieve closed-loop linkage between on-site warnings and information reporting by traffic management departments, making it difficult to meet the real-time requirements of smart transportation and intelligent policing. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for early warning and guidance of highway ramp accidents based on multi-sensor fusion and dual logic judgment mechanism, which solves the technical problems of rapid identification, accurate early warning, hierarchical guidance and collaborative reporting of ramp accidents in the prior art.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A method for early warning and guidance of highway ramp accidents based on multi-sensor fusion includes the following steps: Step 1: Road segment division and equipment deployment; The connecting ramps between the highway toll station and the main line are divided into a pre-entry diversion area, a ramp entrance, a ramp alignment guidance section, and a ramp exit. Multiple sensors, warning devices, and cameras are deployed at key locations along the connecting ramps. The sensors include infrared sensors and pressure sensors deployed at the ramp entrance and exit, and multiple acoustic sensors deployed at the linear guide section and guardrail. The warning equipment includes LED signs deployed in the diversion area before the main entrance, warning lights deployed in the linear guidance section, and light strips deployed on the guardrail; The camera is deployed at the entrance of the ramp to collect license plate and vehicle model information; Step 2: Data collection and vehicle labeling; By collecting vehicle traffic data and environmental data through multiple sensors, the characteristics of vehicles entering the ramp are labeled, and a dynamic database containing the entry time and characteristics of each vehicle is established. Step 3: Status monitoring and anomaly detection; An independent timer is started for each vehicle entering the connecting ramp to monitor the travel time of each vehicle within the ramp; at the same time, abnormal sounds within the ramp are collected by a sound-sensitive sensor; when the travel time of any vehicle exceeds a preset threshold, or when the abnormal sound is detected, it is determined that an accident or vehicle congestion has occurred within the ramp. When a vehicle enters the ramp entrance, the entrance infrared sensor triggers an entry signal, and the microcontroller assigns an independent timer instance to the vehicle, recording the entry timestamp and a preset maximum passage time threshold. When a vehicle exits the ramp exit, the exit infrared sensor triggers an exit signal, and the microcontroller terminates the vehicle's timer instance. When a new vehicle enters, only a new timer instance is added for the new vehicle, without resetting the timers of existing vehicles in the ramp. If the timer value of any timer instance exceeds the preset threshold, a vehicle congestion determination is triggered. Step 4: Tiered early warning and guidance; In response to the aforementioned anomaly determination, the warning devices within the ramp are activated to prompt vehicles within the ramp to slow down and give way. At the same time, the mainline LED signs located in the diversion area before the ramp entrance are activated to display warning messages prohibiting entry and early lane changes, guiding vehicles behind the mainline to detour. Step 5: Information reporting; Abnormal status information and vehicle characteristic information are pushed to the remote traffic management platform in real time.

[0010] Furthermore, the detection and determination of the abnormal sound is performed using the decibel threshold method and / or the spectrum matching method; The decibel threshold method collects audio signals in real time. When the sound pressure level exceeds the preset decibel threshold and the duration exceeds the preset duration, it is determined to be an abnormal sound. The spectrum matching method extracts the time-frequency features of the audio signal and compares them with a pre-stored accident collision audio feature library. When the matching degree exceeds a preset similarity threshold, it is determined to be an abnormal sound.

[0011] Furthermore, the status monitoring and anomaly determination also includes: using infrared sensors or pressure sensors deployed at the entrance and exit of the ramp to count vehicles in both directions and to count the number of vehicles in the ramp in real time; and cross-validating the results of the two-way counting with the timeout determination of each vehicle's independent timer to reduce the false alarm rate.

[0012] Furthermore, the tiered warning and guidance includes: when an accident or vehicle congestion is determined, the warning lights at the linear guidance signs of the drive ramp flash or remain constantly lit, and the guardrail light strips are illuminated to prompt vehicles in the ramp to slow down and avoid the obstacle. At the same time, drive the main line LED sign to display the words "Ramp abnormality, no entry, please change lanes in advance"; At night or in low visibility conditions, the brightness of warning lights and guardrail light strips will be automatically increased or the flashing frequency will be increased.

[0013] Furthermore, the abnormal status information includes: the time of the abnormality, the location of the abnormality, the ramp number, the characteristic information of the suspected accident vehicle, and the license plate information.

[0014] The present invention also provides a highway ramp accident early warning and guidance system based on multi-sensor fusion, used to implement the method, including: The road section deployment module is used to divide the connecting ramps between the highway toll station and the main line into multiple functional sections, and to deploy various sensors, cameras and warning devices in each section. The vehicle perception and labeling module collects the characteristics, entry time and passage data of vehicles entering the ramp by deploying infrared sensors, pressure sensors and cameras at the ramp entrance, and establishes and updates a dynamic database of vehicles in the ramp in real time. The dual logic judgment module includes: an independent timing unit for independently starting a timer for each vehicle in the ramp to monitor its travel time; a bidirectional counting unit for counting the number of vehicles in the ramp; an abnormal sound detection unit for identifying abnormal sounds in the ramp; and a comprehensive judgment unit for outputting an accident / delay judgment signal when the travel time of any vehicle exceeds a preset threshold or an abnormal sound is detected. The graded early warning and guidance module, in response to the judgment signal, drives the warning equipment in the ramp and the LED light signs in the diversion area before the main line entrance to implement deceleration guidance in the ramp and lane change prohibition guidance on the main line. The collaborative reporting module is used to push abnormal information to a remote traffic management platform; The independent timing unit creates an independent timer instance for each vehicle entering the ramp. The timer instances run in parallel without interfering with each other. When a new vehicle enters, only a new timer instance is added, and the timing values ​​of existing timers are not affected. When a vehicle leaves, the corresponding timer instance is terminated.

[0015] Furthermore, the road segment deployment module includes: Section division unit: The connecting ramp is divided into the pre-entry diversion area, the entrance section, the linear guidance section, and the exit section; Sensor deployment unit: Infrared sensors and pressure sensors are deployed at the entrance and exit of the ramp, and acoustic sensors are deployed at the linear guide section and guardrail. Warning equipment unit: LED signs are deployed in the diversion area before the main line entrance, warning lights are deployed in the linear guidance section, and light strips are deployed on the guardrail; Image acquisition unit: Cameras are deployed at the entrance of the ramp to collect license plate and vehicle type information.

[0016] Furthermore, the abnormal acoustic monitoring unit includes: Multiple acoustic sensors are installed at the ramp alignment guidance signs or on the guardrails; The audio signal preprocessing module is used to perform wind noise filtering, background noise modeling, and dynamic suppression. The anomaly detection subunit uses at least one of the following methods to analyze the audio signal in real time: decibel thresholding, spectrum matching, or machine learning classification, and outputs an abnormal sound trigger signal.

[0017] Furthermore, the graded warning and guidance module also includes an environment adaptive unit, which automatically increases the brightness and flashing frequency of the warning lights and light strips at night or in low visibility conditions based on ambient light sensors or real-time weather data.

[0018] Furthermore, the collaborative reporting module includes: The incident generation unit is used to automatically generate structured incident information that includes the time, location, ramp number, and vehicle characteristics of the anomaly. The data push unit pushes the alarm information to the traffic management platform and duty terminals via wired or wireless networks; The log retention unit is used to record the complete process data from anomaly detection to reporting, for subsequent analysis and model optimization.

[0019] Compared with the prior art, the present invention has the following beneficial effects: (1) A triple judgment mechanism of “independent dynamic timing for each vehicle + bidirectional counting + abnormal sound” is adopted. The independent timer solves the problem of not being able to identify a single vehicle stuck in continuous traffic flow; bidirectional counting provides inventory verification; and abnormal sound serves as an immediate trigger condition. The triple logic cross-verification greatly reduces false alarms and missed alarms caused by environmental interference such as rain, fog, strong light, and wind noise.

[0020] (2) Different guidance strategies are adopted for vehicles in the ramp and vehicles behind the main line respectively. Warning lights and light strips are activated in the ramp to prompt deceleration and avoidance, and LED signs on the main line display "Ramp abnormal, no entry, please change lanes in advance", which moves the warning to the main line and provides sufficient deceleration and lane change buffer time for the vehicles behind the highway, significantly reducing the risk of secondary accidents.

[0021] (3) It is specifically designed for the traffic characteristics of the connecting ramps between toll stations and the main line. It uses low-cost general-purpose sensors such as infrared, pressure-sensitive, and acoustic sensors and single-chip microcomputer control. It does not require large-scale modification of the ramps and is not affected by environmental conditions such as rain, fog, and nighttime lighting. It can be quickly deployed and applied.

[0022] (4) Realize the closed-loop linkage between on-site warning and information reporting by traffic management departments, and automatically push structured police information such as abnormal time, location and vehicle characteristics, which meets the needs of smart transportation and intelligent policing construction. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating Embodiment 1 of the present invention.

[0024] Figure 2 This is a structural block diagram of Embodiment 2 of the present invention. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0026] Example 1

[0027] like Figure 1 As shown in the figure, this embodiment provides a method for early warning and guidance of highway ramp accidents based on multi-sensor fusion, applied to the exit ramp between a highway toll station and the main line. The specific steps are as follows: (1) Road segment division and equipment layout; The exit ramp is divided into a pre-entry diversion area (approximately 80m from the entrance), the ramp entrance, the ramp alignment guidance section (approximately 200m in length, including curves), and the ramp exit (connecting to the toll plaza).

[0028] An LED sign is installed in the diversion area before the ramp entrance; infrared sensors and pressure sensors are installed at the ramp entrance and exit respectively; a warning light and an acoustic sensor are installed every 30m on the guardrail of the linear guidance section; and a high-definition camera is installed at the ramp entrance to capture images of the front of the vehicle.

[0029] (2) Data collection and vehicle labeling; When a vehicle enters the ramp entrance, infrared sensors collect vehicle speed and transit time, pressure sensors collect the number of axles and mass characteristics, and cameras collect license plate and vehicle type information. All data is uploaded to a microcontroller, which assigns a unique ID to the vehicle, records the entry timestamp, vehicle type, license plate, and mass, and stores this information in a dynamic database. Upon exiting, the exit sensor is triggered, updating the vehicle's exit marker and marking it as departed from the database.

[0030] (3) Status monitoring and anomaly detection; The microcontroller creates an independent timer instance (software timer) for each entering vehicle, setting a default passage time threshold of 3 minutes (calculated based on ramp length and speed limit; for example, a 600m ramp with a 40km / h speed limit has a normal passage time of approximately 54 seconds, with 3 minutes providing a safety margin). The timer increments every second, and if any timer value exceeds 180 seconds, the vehicle is considered to be stuck. Simultaneously, a sound sensor continuously collects audio data using a decibel threshold method: a sound pressure level exceeding 100dB for more than 0.5 seconds is considered an abnormal sound. Meeting either condition triggers a warning system.

[0031] In this embodiment, the microcontroller maintains a linked list of vehicle timers. Each node contains the following fields: vehicle ID, entry time, current timer value (seconds), threshold (seconds), and status (driving / stalled). Two pairs of infrared sensors (one pair per lane) are installed at the entrance of the two-lane ramp. When a vehicle triggers the entry signal, the microcontroller adds a new node to the linked list, incrementing the timer value by 1 every second starting from 0. Simultaneously, the system timer interrupt service function periodically traverses the linked list, incrementing the timer value of all nodes by 1. When the timer value of any node is greater than or equal to the threshold, the node's status is marked as "stalled," and a warning is triggered. When the exit infrared sensor detects a vehicle exiting, it matches the oldest vehicle according to the exit detection order, or finds and deletes the corresponding node using pressure-sensitive sensor feature matching.

[0032] When a new vehicle enters, only a new node is added; the timer values ​​of any existing nodes are not modified. Therefore, even if a vehicle has been stuck in the ramp for several minutes, the entry of a new vehicle will not reset the timer of the stuck vehicle, thus avoiding misjudging it as normal.

[0033] In this embodiment, a high-sensitivity microphone (sound sensor) is installed every 30m on the outer guardrail of the long curve of the ramp. The microphone output signal is pre-amplified and band-pass filtered (300Hz~4000Hz, covering the main frequency band of collision sound) before being input into the microcontroller. The microcontroller continuously calculates the RMS value within 20ms, converts it into sound pressure level, and if it is greater than 105dB and the duration is ≥0.5s, an abnormality flag is triggered.

[0034] Alternatively, the audio data within 1 second can be subjected to FFT transformation to extract features such as the spectral centroid, bandwidth, and subband energy ratio. These features can then be compared with the pre-stored templates of "metal collision sound" and "brake screech sound" using cosine similarity. If the similarity is greater than 0.8, the signal is triggered.

[0035] To avoid false alarms such as heavy trucks passing by, an abnormal sound event is only confirmed when the decibel method is triggered and the spectral matching degree is greater than 0.7. Otherwise, it is only recorded as a suspicious event and waits for the timer to expire or other sensors to verify.

[0036] (4) Tiered early warning and guidance; Once an accident or delay is confirmed, the microcontroller drives all warning lights in the linear guidance section within the ramp to flash at a frequency of 2Hz, and the guardrail light strip illuminates red and flashes sequentially along the direction of travel, prompting vehicles within the ramp to immediately slow down and avoid the area. Simultaneously, the mainline LED sign displays "Ramp abnormality, entry prohibited, please change lanes in advance," flashes three times, and then remains constantly lit. If nighttime is detected (based on a light sensor or time determination), the light strip brightness is automatically increased to maximum, and the frequency is increased to 3Hz.

[0037] (5) Information reporting; The microcontroller automatically pushes the police information (time: 2025-03-25, 14:32:17; location: G60 Expressway K45+300 exit ramp; anomaly type: overdue detention + abnormal sound; vehicle characteristics: blue license plate, Anhui A·12345, light truck) to the traffic management command platform and the handheld terminals of nearby police officers via the 4G module.

[0038] Traffic management personnel report the results to the management platform based on the actual handling situation (such as whether it was an accident or a false alarm). The system periodically adjusts the timing threshold (e.g., adjusting the nighttime threshold to 2 minutes and 30 seconds) or adjusts the sensitivity and filtering parameters of the acoustic sensor based on the feedback.

[0039] Example 2

[0040] like Figure 2 As shown, this embodiment provides a highway ramp accident early warning and guidance system based on multi-sensor fusion, including: a road section deployment module, a vehicle perception and labeling module, a dual logic judgment module, a graded early warning and guidance module, and a collaborative reporting module.

[0041] The road segment deployment module is used to divide the connecting ramps between the highway toll station and the main line into multiple functional sections, and deploy various sensors, cameras, and warning devices in each section. Specifically, the road segment deployment module includes a section division unit that divides the connecting ramps into an entrance diversion area, an entrance section, a linear guidance section, and an exit section; a sensor deployment unit that deploys infrared sensors and pressure sensors at the ramp entrances and exits, and acoustic sensors at the linear guidance section and guardrails; a warning device unit that deploys LED signs in the entrance diversion area, warning lights in the linear guidance section, and light strips on the guardrails; and an image acquisition unit that deploys cameras at the ramp entrances to collect license plate and vehicle model information.

[0042] This embodiment uses an STM32F407 microcontroller as the core, and externally connects an E3F-20L infrared sensor, a FlexiForce A401 pressure-sensitive film sensor, a MAX4466 sound sensor, an OV5640 camera (FIFO buffer), an LED warning light (12V DC, with constant current drive), an LED sign (P10 unit board, communicating via RS485), a guardrail light strip (WS2812B addressable LED light strip), a 4G communication module (EC200N), and a BH1750 ambient light sensor.

[0043] The vehicle perception and labeling module collects the characteristics, entry time and traffic data of vehicles entering the ramp by deploying infrared sensors, pressure sensors and cameras at the ramp entrance, and establishes and updates a dynamic database of vehicles in the ramp in real time.

[0044] The dual logic determination module includes an independent timing unit for independently starting a timer for each vehicle in the ramp to monitor its passage time; a bidirectional counting unit for counting the number of vehicles in the ramp; an abnormal sound detection unit for identifying abnormal sounds in the ramp; and a comprehensive determination unit for outputting an accident / delay determination signal when the passage time of any vehicle exceeds a preset threshold or an abnormal sound is detected.

[0045] The independent timing unit creates an independent timer instance for each vehicle entering the ramp. The timer instances run in parallel without interfering with each other. When a new vehicle enters, only a new timer instance is added, and the timing values ​​of existing timers are not affected. When a vehicle leaves, the corresponding timer instance is terminated.

[0046] The abnormal acoustic monitoring unit includes multiple acoustic sensors installed at the ramp alignment guidance signs or guardrails; an audio signal preprocessing module for performing wind noise filtering, background noise modeling, and dynamic suppression; and an abnormality determination subunit that uses the decibel threshold method and / or spectrum matching method to perform real-time analysis of the audio signal and outputs an abnormal sound trigger signal.

[0047] The graded warning and guidance module is used to respond to the judgment signal and drive the warning equipment in the ramp and the LED light signs in the diversion area before the main line entrance to implement deceleration guidance in the ramp and lane change prohibition guidance on the main line.

[0048] The graded early warning and guidance module includes an environment adaptive unit, which automatically increases the brightness and flashing frequency of warning lights and light strips at night or in low visibility conditions based on ambient light sensors or real-time weather data.

[0049] The collaborative reporting module is used to push abnormal information to a remote traffic management platform; specifically, it includes: an alarm generation unit for automatically generating structured alarm information containing abnormal time, location, ramp number and vehicle characteristics; a data push unit for pushing the alarm information to the traffic management platform and duty terminals via wired or wireless network; and a log retention unit for recording the complete process data from anomaly determination to reporting for subsequent analysis and model optimization.

[0050] The workflow involves a road segment deployment module that pre-configures the installation location, orientation, and threshold parameters of each sensor. The vehicle perception and labeling module continuously collects infrared, pressure-sensitive, and camera data, generating vehicle entry and exit events and feature libraries. A dual-logic judgment module runs independent timer lists and acoustic detection algorithms, outputting abnormal trigger signals. A tiered warning and guidance module automatically selects the warning level based on the judgment result, the current time period (day / night), and weather (obtained via a network API): Level 1 (triggered only by timeout, no abnormal sound) only illuminates warning lights within the ramp; Level 2 (triggered by abnormal sound or delay + suspicious sound) simultaneously illuminates the mainline LED signs and pushes an alert; Level 3 (delay + confirmed genuine abnormal sound) initiates high-speed flashing of all light strips and adds voice broadcast (optional). The collaborative reporting module encapsulates the alert information in JSON format and uploads it to the traffic management platform via the MQTT protocol, while simultaneously storing logs on a local SD card.

[0051] This system was deployed at the entrance ramp (from the toll station to the main line) and the corresponding exit ramp (from the main line to the toll station) of the G4211 expressway at kilometer marker 12+300 in a certain province. The exit ramp is a curved downhill section, prone to accidents. Within three months of deployment, a total of 7 real accident / delay events were detected (including two cases of truck breakdowns and delays, one case of a car colliding with a guardrail, and four cases of single-vehicle delays exceeding the time limit), with an average response time of 2.8 seconds and no missed reports; there were 2 false alarms (both caused by low-frequency vibrations from large trucks passing by at night, which were later resolved by adding vibration sensors for verification). After receiving alerts through the platform, the traffic management department's average response time was reduced by 40%, and no secondary accidents caused by ramp accidents occurred.

[0052] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for early warning and guidance of highway ramp accidents based on multi-sensor fusion, characterized in that, Includes the following steps: Step 1: Road segment division and equipment deployment; The connecting ramps between the highway toll station and the main line are divided into a pre-entry diversion area, a ramp entrance, a ramp alignment guidance section, and a ramp exit. Multiple sensors, warning devices, and cameras are deployed at key locations along the connecting ramps. The sensors include infrared sensors and pressure sensors deployed at the ramp entrance and exit, and multiple acoustic sensors deployed at the linear guide section and guardrail. The warning equipment includes LED signs deployed in the diversion area before the main entrance, warning lights deployed in the linear guidance section, and light strips deployed on the guardrail; The camera is deployed at the entrance of the ramp to collect license plate and vehicle model information; Step 2: Data collection and vehicle labeling; By collecting vehicle traffic data and environmental data through multiple sensors, the characteristics of vehicles entering the ramp are labeled, and a dynamic database containing the entry time and characteristics of each vehicle is established. Step 3: Status monitoring and anomaly detection; An independent timer is started for each vehicle entering the connecting ramp to monitor the travel time of each vehicle within the ramp; at the same time, abnormal sounds within the ramp are collected by a sound-sensitive sensor; when the travel time of any vehicle exceeds a preset threshold, or when the abnormal sound is detected, it is determined that an accident or vehicle congestion has occurred within the ramp. When a vehicle enters the ramp entrance, the entrance infrared sensor triggers an entry signal, and the microcontroller assigns an independent timer instance to the vehicle, recording the entry timestamp and a preset maximum passage time threshold. When a vehicle exits the ramp exit, the exit infrared sensor triggers an exit signal, and the microcontroller terminates the vehicle's timer instance. When a new vehicle enters, only a new timer instance is added for the new vehicle, without resetting the timers of existing vehicles in the ramp. If the timer value of any timer instance exceeds the preset threshold, a vehicle congestion determination is triggered. Step 4: Tiered early warning and guidance; In response to the aforementioned anomaly determination, the warning devices within the ramp are activated to prompt vehicles within the ramp to slow down and give way. At the same time, the mainline LED signs located in the diversion area before the ramp entrance are activated to display warning messages prohibiting entry and early lane changes, guiding vehicles behind the mainline to detour. Step 5: Information reporting; Abnormal status information and vehicle characteristic information are pushed to the remote traffic management platform in real time.

2. The method for early warning and guidance of highway ramp accidents based on multi-sensor fusion according to claim 1, characterized in that, The detection and determination of the abnormal sound are performed using the decibel threshold method and / or the spectrum matching method; The decibel threshold method collects audio signals in real time. When the sound pressure level exceeds the preset decibel threshold and the duration exceeds the preset duration, it is determined to be an abnormal sound. The spectrum matching method extracts the time-frequency features of the audio signal and compares them with a pre-stored accident collision audio feature library. When the matching degree exceeds a preset similarity threshold, it is determined to be an abnormal sound.

3. The method for early warning and guidance of highway ramp accidents based on multi-sensor fusion according to claim 1, characterized in that, The status monitoring and anomaly determination also include: using infrared sensors or pressure sensors deployed at the entrance and exit of the ramp to count vehicles in both directions and to count the number of vehicles in the ramp in real time; and cross-validating the results of the two-way counting with the timeout determination of each vehicle's independent timer to reduce the false alarm rate.

4. The method for early warning and guidance of highway ramp accidents based on multi-sensor fusion according to claim 1, characterized in that, The graded early warning and guidance includes: when an accident or vehicle congestion is determined, the warning lights at the linear guidance signs of the drive ramp will flash or remain on, and the guardrail light strips will be illuminated to prompt vehicles in the ramp to slow down and avoid the obstacle. At the same time, drive the main line LED sign to display the words "Ramp abnormality, no entry, please change lanes in advance"; At night or in low visibility conditions, the brightness of warning lights and guardrail light strips will be automatically increased or the flashing frequency will be increased.

5. The method for early warning and guidance of highway ramp accidents based on multi-sensor fusion according to claim 1, characterized in that, The abnormal status information includes: the time of the abnormality, the location of the abnormality, the ramp number, the characteristic information of the suspected accident vehicle, and the license plate information.

6. A highway ramp accident early warning and guidance system based on multi-sensor fusion, characterized in that, For implementing the method according to any one of claims 1-5, comprising: The road section deployment module is used to divide the connecting ramps between the highway toll station and the main line into multiple functional sections, and to deploy various sensors, cameras and warning devices in each section. The vehicle perception and labeling module collects the characteristics, entry time and passage data of vehicles entering the ramp by deploying infrared sensors, pressure sensors and cameras at the ramp entrance, and establishes and updates a dynamic database of vehicles in the ramp in real time. The dual logic judgment module includes: an independent timing unit for independently starting a timer for each vehicle in the ramp to monitor its travel time; a bidirectional counting unit for counting the number of vehicles in the ramp; an abnormal sound detection unit for identifying abnormal sounds in the ramp; and a comprehensive judgment unit for outputting an accident / delay judgment signal when the travel time of any vehicle exceeds a preset threshold or an abnormal sound is detected. The graded early warning and guidance module, in response to the judgment signal, drives the warning equipment in the ramp and the LED light signs in the diversion area before the main line entrance to implement deceleration guidance in the ramp and lane change prohibition guidance on the main line. The collaborative reporting module is used to push abnormal information to a remote traffic management platform; The independent timing unit creates an independent timer instance for each vehicle entering the ramp. The timer instances run in parallel without interfering with each other. When a new vehicle enters, only a new timer instance is added, and the timing values ​​of existing timers are not affected. When a vehicle leaves, the corresponding timer instance is terminated.

7. The highway ramp accident early warning and guidance system based on multi-sensor fusion according to claim 6, characterized in that, The road segment deployment module includes: Section division unit: The connecting ramp is divided into the pre-entry diversion area, the entrance section, the linear guidance section, and the exit section; Sensor deployment unit: Infrared sensors and pressure sensors are deployed at the entrance and exit of the ramp, and acoustic sensors are deployed at the linear guide section and guardrail. Warning equipment unit: LED signs are deployed in the diversion area before the main line entrance, warning lights are deployed in the linear guidance section, and light strips are deployed on the guardrail; Image acquisition unit: Cameras are deployed at the entrance of the ramp to collect license plate and vehicle type information.

8. The highway ramp accident early warning and guidance system based on multi-sensor fusion according to claim 6, characterized in that, The abnormal acoustic monitoring unit includes: Multiple acoustic sensors are installed at the ramp alignment guidance signs or on the guardrails; The audio signal preprocessing module is used to perform wind noise filtering, background noise modeling, and dynamic suppression. The anomaly detection subunit uses at least one of the following methods to analyze the audio signal in real time: decibel thresholding, spectrum matching, or machine learning classification, and outputs an abnormal sound trigger signal.

9. The highway ramp accident early warning and guidance system based on multi-sensor fusion according to claim 6, characterized in that, The graded early warning and guidance module also includes an environment adaptive unit, which automatically increases the brightness and flashing frequency of warning lights and light strips at night or in low visibility conditions based on ambient light sensors or real-time weather data.

10. The highway ramp accident early warning and guidance system based on multi-sensor fusion according to claim 6, characterized in that, The collaborative reporting module includes: The incident generation unit is used to automatically generate structured incident information that includes the time, location, ramp number, and vehicle characteristics of the anomaly. The data push unit pushes the alarm information to the traffic management platform and duty terminals via wired or wireless networks; The log retention unit is used to record the complete process data from anomaly detection to reporting, for subsequent analysis and model optimization.

Images