Expressway Incident Sensing System

System Introduction

Expressway operations today face the triple challenges of high traffic volume, high speed, and high risk. Traditional monitoring methods have significant shortcomings in sensing accuracy, response timeliness, and coordinated response: video monitoring is easily affected by weather and lighting, with low recognition rates for subtle abnormal behaviors and frequent false/missed alarms. Incident handling relies on manual detection and reporting, often missing the optimal response window. Sensing, decision-making, and information dissemination are fragmented, lacking efficient coordination. Massive trajectory data cannot be quickly analyzed, and abnormal behavior mining lags behind.

To address these pain points, the Expressway Incident Sensing System uses radar-video fusion technology as its core, integrating millimeter-wave radar, high-definition video, edge computing, and intelligent dome cameras to build an integrated “cloud-edge-end” incident linkage system. The system delivers all-weather, high-precision vehicle trajectory sensing. Radar accurately captures motion information while video synchronously extracts visual features. Spatial-temporal fusion generates continuous trajectories with unique identity. Edge computing enables on-site data fusion and incident assessment, achieving second-level detection and automatic alerts for incidents such as abnormal stop, reverse drive, congestion, and illegal lane change. Dome cameras are triggered for snapshot evidence, and warning information is pushed to variable message signs, broadcasting devices, etc., forming a full closed loop of “sensing – assessment – alert – handling – archiving”, significantly improving incident response efficiency and coordinated handling capability.

System Architecture

The Expressway Incident Sensing System adopts a three-layer architecture: front-end sensing, edge computing, and business application.

1) Front-end Sensing Layer
Millimeter-wave radar and high-definition video are fused. Radar collects vehicle position, speed, heading angle; fixed cameras extract visual features such as license plate and vehicle type; dome cameras handle incident evidence capture, forming an all-weather multi-source sensing network.

2) Edge Computing Layer
Built-in intelligent algorithms perform millisecond-level incident assessment, triggering radar-dome linkage for automatic snapshot evidence capture. Structured data is uploaded to the platform in real time, enabling rapid edge response and coordinated in-depth analysis at the center.

3) Business Application Layer
The platform presents macroscopic network situations via digital twin, supports full-process incident closed-loop handling, and triggers alerts through variable message signs and broadcasting, forming a fast closed loop of “sensing – assessment – evidence – alert – handling – archiving”.

System Functions

1) Traffic Data Sensing

Lane-level trajectory data: Continuous output of real-time position coordinates, instantaneous speed, heading angle, and acceleration for each vehicle, supporting full tracking and path reconstruction.
Multi-section statistical data: Over a statistical period, provides traffic volume, time occupancy, average speed, time headway, space headway, turning ratio, and 85th percentile speed.
Area status detection: At a given moment within a detection area, provides vehicle count, space occupancy, average speed, distribution coefficient, position of first and last vehicles.

2) Traffic Incident Sensing
Based on millimeter-wave radar and video fusion sensing, the system achieves accurate identification and second-level alerts for incidents across entire expressway sections under all weather conditions. Incident types include:

Stop incident: Radar detects vehicle speed approaching zero and exceeding a time threshold, determined as abnormal stop.
High-speed / low-speed incident: Radar speed measurement compared against speed limit thresholds to identify overspeed or underspeed.
Reverse drive incident: Radar calculates heading angle, compares with lane direction to detect reverse driving.
Lane change incident: Radar tracking analyzes lateral displacement to identify illegal lane changes.
Congestion incident: Detects congestion start/end points, length, duration, and congestion level (mild/moderate/severe), updating impact range in real time.
No-entry incident: Radar-video target classification identifies pedestrians or non-motor vehicles entering restricted areas.

Through edge computing and cloud collaboration, the system implements a full closed loop from detection, evidence capture, alerting to archiving, effectively improving expressway safety operations and incident handling efficiency.

3) Incident Linkage Handling

Incident sensing: Radar continuously tracks lane-level trajectories. When a vehicle is determined to be an abnormal stop, incident information (object ID, position, lane, timestamp) is generated and pushed to the edge computing unit.
Linkage evidence capture: Upon receiving the incident, the edge computing unit drives the dome camera to automatically pan, tilt, and zoom based on radar coordinates, capturing high-definition video and snapshots. Radar trajectory and video stream are spatially-temporally fused to generate a structured incident package.
Platform management: The incident package is uploaded to the management platform via dedicated network, automatically categorized and archived, supporting multi-dimensional retrieval and playback, providing a basis for accident attribution and traffic analysis.

System Deployment

Deployment uses pole-mounted installation on existing roadside infrastructure such as gantries, monitoring poles, and light poles – no additional pole construction required. Radar installation height: 8–12 m, with a high downward viewing angle to ensure wide, unobstructed detection coverage, effectively avoiding occlusion by trees and vehicles, and achieving large-area, blind-spot-free coverage of mainlines and ramps.

At each location, a high-definition dome camera and a structured fixed camera are deployed. The fixed camera captures license plate and vehicle type features, while the dome camera handles incident evidence capture and dynamic tracking.

Back to Solutions