Early warning system and method for intrusion of highway curve construction area

Abstract

The invention relates to a highway curve construction area intrusion early warning system and method, and belongs to the technical field of highway early warning. The system is divided into a sensing layer, a decision-making layer, an execution layer and a guarantee layer. The sensing layer is composed of a multi-level thundersight all-in-one machine; the decision-making layer is edge computing nodes deployed in a distributed manner; and the execution layer comprises an intelligent roadside warning device and wearable terminal equipment. In the sensing layer, all sensing data monitored by the multi-level thundersight all-in-one machine is transmitted to an edge computing node through a time sensitive switch, the edge computing node carries out space-time alignment on a point cloud signal monitored by the thundersight all-in-one machine and a visual recognition result through a multi-source data fusion algorithm, and the data after space-time alignment is input into a CTRA model, so that the CTRA model is constructed. And the CTRA model judges the intrusion risk of the vehicle according to the received data and outputs a corresponding decision instruction data packet, and the decision instruction data packet is transmitted to the execution layer. And the execution layer executes corresponding sound-light-vibration multi-dimensional early warning according to the corresponding instruction in the decision instruction data packet.

Description

Technical Field

[0001] The present invention belongs to the technical field of highway early warning, and in particular relates to an early warning system and method for intrusion into a curve construction area of ​​a highway. Background Art

[0002] Construction zones on expressway curves are a high-incidence area for traffic accidents. The core risk stems from the dual uncertainties of driver behavior and complex road conditions. Currently, most construction zone intrusion accidents are related to unintentional driver behavior, specifically vehicle failure, distracted driving, fatigue, and external interference. Research shows that these factors contribute to over 60% of construction zone intrusion accidents, with distracted driving leading the charge, primarily involving mobile phone use and operating onboard devices. These factors can reduce drivers' ability to discern construction signs by over 40%. Furthermore, while vehicle failures (such as tire blowouts and brake failures), while a relatively low factor (23%), often have more serious consequences. Due to the centrifugal force of curves, vehicles entering construction zones at average speeds of up to 80 km / h. Fatigue and external interference, such as inclement weather, significantly delay driver reaction times. On nighttime curves, unintentional vehicle deviations due to fatigue can reach distances two to three times greater than normal.

[0003] Existing highway early warning systems have significant technical limitations. The effectiveness of traditional static warning systems (such as cones and warning signs) is limited by their fixed deployment and single visual warning mode. Under conditions of limited visibility on curves, their effective warning range is often less than 200 meters. More critically, these passive protection measures lack real-time awareness of driver status and vehicle anomalies, resulting in a significant lag in response to sudden risks. Even in intelligently retrofitted systems, single-sensor monitoring solutions struggle to cope with the complex environment of curves. For example, pure vision systems experience a drop in detection accuracy of over 50% in backlight, rain, or fog, while millimeter-wave radars have a 15%-20% blind spot on the inside of curves. These technical limitations result in a system recognition rate of less than 60% for driver distraction and fatigue, failing to meet safety requirements in construction zones.

[0004] Current construction warning technology systems struggle to effectively address these risks. Traditional passive protection methods using cones and warning lights have significant flaws: static deployment lacks dynamic adaptability, preventing the warning range from adjusting based on real-time traffic conditions; information perception is limited, and the limitations of relying on visual warnings are particularly pronounced in rainy and foggy conditions; human-machine interaction is inefficient, insufficiently stimulating driver attention. Furthermore, the geometric characteristics of curves exacerbate the risk. Against this backdrop, a dynamic protection system with multi-source perception, intelligent decision-making, and proactive warning capabilities is urgently needed. Summary of the Invention

[0005] The present invention provides a highway curve construction area intrusion warning system and method, so as to solve the technical limitation problem of insufficient warning effect of the warning system in actual highway operation in the prior art.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A highway curve construction zone intrusion warning system includes a perception layer, a decision layer, and an execution layer. The perception layer includes several multi-level integrated radar and vision machines, the decision layer is a distributed edge computing node, and the execution layer is a plurality of warning devices. The multi-level integrated radar and vision machines and the warning devices are distributed and installed upstream of the construction area, and the edge computing nodes are deployed inside the anti-collision buffer vehicle within the anti-collision buffer zone of the construction area. The perception layer, decision layer, execution layer, and security layer form a backbone communication link through an industrial optical fiber ring network. The multi-level integrated radar and vision machines, edge computing nodes, and warning devices use a wireless network as a redundant channel.

[0007] The perception layer includes several multi-level radar and vision integrated machines, including long-range radar and vision integrated machines, medium-range radar and vision integrated machines, short-range radar and vision integrated machines, and roadside millimeter-wave radars; among them, the long-range radar and vision integrated machine is equipped with a high-performance millimeter-wave radar, adopts an antenna array design, and is equipped with an infrared thermal imager; the medium-range radar and vision integrated machine integrates millimeter-wave radar and a camera; the short-range radar and vision integrated machine (construction area boundary) uses a combination of ultra-wideband radar and a camera.

[0008] A time-sensitive switch is also set up in the perception layer. The time-sensitive switch is deployed at the same site as the radar-visual integrated machine. The target perception data monitored by the radar-visual integrated machine is exchanged through the time-sensitive switch.

[0009] The decision-making layer is also equipped with a data storage unit, which is used to record complete warning data, form an event log, and update the log in real time for subsequent data analysis by staff.

[0010] The execution layer consists of several warning devices, including intelligent lifting cones, directional sound wave transmitters and LED variable information boards. The intelligent lifting cones are hydraulically driven, and their columns are integrated with LED lamp beads. The intelligent lifting cones are arranged in an oblique line in the transition zone upstream of the construction area. The directional sound wave transmitters are composed of phased arrays. The LED variable information board uses an LED panel, and the built-in ambient light sensor can sense the intensity of the surrounding light. The directional sound wave transmitters and LED variable information boards are both arranged upstream of the construction area.

[0011] The execution layer also includes a construction worker positioning bracelet, which is worn by construction area workers. The construction worker positioning bracelet integrates a linear motor and UWB positioning module, and has UWB positioning and early warning vibration prompt functions.

[0012] A method for early warning of intrusion into a highway curve construction area, the early warning method comprising: The perception layer uses multi-level radar and vision integrated devices to monitor the target perception data of intruders; The decision layer receives target perception data, determines the target's intrusion risk through the edge computing node, and generates corresponding decision instruction data packets; The execution layer receives the decision instruction data packet and controls the warning device to issue corresponding warnings according to the decision instruction data packet.

[0013] The data monitored by the multi-level radar-vision integrated machine includes different monitoring data detected by radar and cameras. The different monitoring data detected by radar and camera are synchronized in time and space through the time-space collaborative detection mechanism to achieve multi-source data fusion, and then form target perception data. The target perception data is transmitted to the decision-making layer through the time-sensitive switch.

[0014] All radar and vision integrated devices in the multi-level radar and vision integrated device support long-distance posture adjustment. Its working principle is to synchronize the data monitored by the radar module and the vision module in time and space through a spatiotemporal collaborative detection mechanism, realize multi-source data fusion, ensure that the long-distance posture adjustment action matches the target status in real time, and build a high-precision, real-time perception capability of the long-distance target posture, thereby supporting the terminal to dynamically adjust itself or the system strategy according to the target status.

[0015] The intrusion risk is divided into three levels: Level I (S<0.3) triggers a preventive warning, Level II (0.3≤S<0.7) activates an audible and visual alarm, and Level III (S≥0.7) performs an emergency warning.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The highway curve construction area intrusion warning system and method of the present invention have significant advantages over the existing technology. The system adopts a perception network composed of intelligent multi-level radar and visual integrated devices, and realizes all-weather and all-round vehicle monitoring through radar and visual fusion technology, which effectively solves the problem of low efficiency of traditional manual deployment. The innovative dynamic risk assessment model combined with the multi-source data fusion algorithm greatly improves the accuracy of the warning, especially when dealing with complex scenarios such as large vehicle obstruction and severe weather. The system has built a three-dimensional protection system of "roadside equipment-wearable terminal", realizing full-process safety protection from prevention to emergency response. Modular design and redundant configuration ensure the stability and reliability of the system, while flexible deployment schemes enable it to quickly adapt to the needs of different construction scenarios. The overall solution significantly improves the safety protection level of the construction area and provides an intelligent solution for highway maintenance operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1: Structure diagram of the highway curve construction area intrusion warning system; Figure 2 : Perception layer composition structure diagram; Figure 3 : Decision-making layer composition structure diagram; Figure 4 : Execution layer composition structure diagram; Figure 5 : Structure diagram of security layer; Figure 6 : Flowchart of intelligent early warning method for highway curve construction area. DETAILED DESCRIPTION

[0018] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0019] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0020] This embodiment proposes a highway curve construction area intrusion warning system, and its specific implementation method is as follows: This system adopts a modular layered architecture design, and consists of four core hardware subsystems to form a complete early warning and protection system. Figure 1 As shown in the figure, a closed-loop function is achieved through a four-layer architecture of "perception-decision-execution-guarantee". It is divided into the perception layer, decision layer, execution layer and guarantee layer. The overall function of the system covers four core modules: real-time tracking of vehicle trajectories, intelligent risk assessment, graded warning triggering and system status monitoring. It can automatically adapt to curves with different curvature radii, and build a three-line safety defense of "long-range warning, medium-range interception and close-range protection" for temporary construction areas. Figure 2As shown, the perception layer is a three-dimensional perception network responsible for all-weather vehicle trajectory monitoring. This three-dimensional perception network consists of multiple layers of integrated radar and vision systems, including long-range, medium-range, and short-range radar and vision systems, as well as roadside millimeter-wave radars. The long-range radar and vision systems are located 3-5 km upstream of the construction zone and are equipped with high-performance millimeter-wave radars using an antenna array design and an infrared thermal imager. The medium-range radar and vision systems are located 1-3 km upstream of the construction zone and integrate millimeter-wave radars and cameras. Short-range radar and vision systems are deployed every 200 meters at the construction zone boundary. These systems (at the construction zone boundary) use a combination of ultra-wideband radar and cameras. The multi-layered radar and vision systems support remote attitude adjustment. The radar and vision systems consist of radar and vision modules. The roadside millimeter-wave radars are installed on the outer guardrails of curves, with each radar deployed every 200 meters. A time-sensitive switch is also set up in the perception layer. The time-sensitive switch is deployed at the same site as the radar-visual integrated machine. The target perception data monitored by the radar-visual integrated machine is exchanged through the time-sensitive switch.

[0021] like Figure 3 As shown, the decision layer consists of distributed edge computing nodes, responsible for data fusion and risk assessment. The decision layer consists of edge computing nodes and data storage units. These nodes are deployed inside a crash buffer vehicle within the construction area's crash buffer zone. The edge computing nodes are high-performance computing platforms that run a containerized data processing platform and manage algorithm services. Radar data is rapidly processed using graphics cards, automatically identifying vehicle locations and motion trajectories. Video analysis uses the YOLOv6 model to rapidly calculate the locations and velocities of all detected vehicles.

[0022] like Figure 4As shown, the execution layer includes several warning devices, including intelligent roadside warning devices and wearable terminal devices, enabling multimodal warning output. Specifically, the execution layer includes intelligent lifting cones, directional sonic emitters, and LED variable message boards. The intelligent lifting cones are hydraulically driven, with integrated LED lamps in their columns. They are arranged diagonally in the transition zone upstream of the construction area, ranging from 200 to 500 meters from the construction area, with a gradual spacing of 5-20 meters. The directional sonic emitters consist of a phased array and are located 300 meters upstream of the construction area. The LED variable message board uses an LED panel with a built-in ambient light sensor that detects ambient light levels, ensuring the electronic message board (LED display) remains clearly visible in all weather conditions (daytime, nighttime, and fog), without glare or dimming. The LED variable message board is deployed 1 km upstream of the construction area. The intelligent lifting cones, directional sonic emitters, and LED variable message board all feature cast aluminum housings and are equipped with supercapacitor energy storage modules to support low-temperature startup. After receiving the decision-making instructions from the decision-making layer, the execution layer initiates a multi-level response process. Intelligent lifting cones automatically adjust their height and flashing frequency based on the risk level. LED variable information boards display dynamic graphic warnings. Directional acoustic wave transmitters use parametric array technology to generate audible sound beams, using beamforming to prevent noise pollution. The wearable terminal is a construction worker positioning bracelet, worn by each worker in the construction area. This bracelet integrates a linear motor and a UWB positioning module, providing UWB (Ultra-Wideband) positioning and early warning vibration alerts. The bracelet issues a vibration warning based on the risk level of the decision-making instruction. Upon sensing vibration, workers evacuate the construction area. The UWB positioning module in the bracelet provides real-time feedback on the worker's location to the local FPGA monitoring platform, ensuring their safety. In another preferred embodiment of the present invention, the portable warning device carried by the construction workers may also be a wearable warning epaulette, which projects an aperture warning based on the risk level in the decision-making instruction, that is, projects a visible warning area aperture on the ground around the wearer (construction worker), and the aperture flashes continuously, and the color of the aperture may change with the risk level.

[0023] like Figure 5 As shown, the support layer is provided by highly reliable power supply equipment and a communication network. The power supply equipment includes intelligent power distribution cabinets and a solar energy charging system. The intelligent power distribution cabinets contain lithium iron phosphate battery packs. The communication network utilizes an industrial fiber optic ring network and communication base stations. The power supply subsystem implements dynamic load management. The intelligent power distribution cabinets and lithium iron phosphate battery packs work in conjunction with the solar energy charging system to ensure continuous operation even in rainy days, prioritizing power supply to critical equipment.

[0024] An industrial fiber-optic ring network forms the backbone communication link between the perception, decision-making, execution, and support layers. Wireless networks are used as redundant channels, creating a reliable dual-network hot standby connection solution. The backbone communication link is Ethernet-based, using industrial switches to build a gigabit fiber-optic ring network to ensure real-time data transmission. The wireless network supports both multicast and broadcast transmission modes. In actual deployment, a "three vertical and one horizontal" topology is employed to achieve three-dimensional coverage in the vertical, horizontal, and vertical directions. Specifically, in the vertical direction, radar-based perception units are deployed every 500 meters; in the horizontal direction, equipment chains are formed along the construction zone boundary; and in the vertical direction, roadside equipment (8 meters high) and ground equipment (1.5 meters high) are deployed.

[0025] In this embodiment, the hardware is deployed upstream from the construction area at the following distances: the long-range radar and visual integrated device is located 3 km upstream from the construction area, the medium-range radar and visual integrated device is located 2 km upstream from the construction area, the LED variable information board is located 1 km upstream from the construction area, the edge computing node is located 500 meters upstream from the construction area, the directional acoustic wave transmitter is located 300 meters upstream from the construction area, and the intelligent cone is located 200 meters upstream from the construction area. During operation, the hardware works together through a four-layer architecture consisting of perception, decision-making, execution, and assurance, forming a complete early warning closed loop. The perception layer equipment constitutes an all-weather monitoring network: the millimeter-wave radars in the long-range and medium-range integrated radar vision machines cooperate with the high-definition cameras on the integrated radar vision machines to conduct early detection and trajectory tracking of vehicles coming from the upstream direction of the construction area, and detect the positioning information and trajectory data of intruding vehicles; the short-range integrated radar vision machine at the boundary of the construction area measures the target characteristics and motion status, and the roadside millimeter-wave radar at the guardrail on the outside of the curve uses radar waves to monitor obscured targets, filling the line of sight of the visual module, and realizing detail capture and blind spot supplementation. All perception data monitored by multi-level radar and vision integrated devices in the perception layer are transmitted to edge computing nodes through time-sensitive switches. These edge computing nodes deployed in the anti-collision buffer car use a multi-source data fusion algorithm to spatially and temporally align the point cloud signals monitored by the radar module with the visual recognition results monitored by the vision module. The vehicle-related data after spatial and temporal alignment is input into the CTRA (Constant Turn Rate and Acceleration) model. The CTRA model predicts the trajectory of the obscured vehicle based on the received vehicle-related data, determines the vehicle's intrusion risk, and outputs the corresponding decision-making instruction data packet. The decision-making instruction data packet is sent to the execution layer through the industrial fiber optic ring network and 5G-V2X dual channels. The execution layer executes the corresponding warnings based on the corresponding instructions in the decision-making instruction data packet. Specifically, the LED variable information board displays the corresponding dynamic warning graphics according to the decision-making instructions; the directional sound wave transmitter activates the beamforming alarm, emitting a directional alarm to the driver of the intruding vehicle; and the intelligent lifting cone adjusts its height and flashing frequency according to the risk level. The height of the intelligent lifting cone for medium risk levels can be adjusted from 0.8 to 1.5 meters, and the flashing frequency can be adjusted from 1 to 5Hz. Simultaneously, the construction worker's positioning bracelet uses UWB positioning to track the location in real time and trigger graded vibration alerts based on the risk level.

[0026] In another preferred embodiment of this embodiment, all radar and visual systems in the multi-layered radar and visual system support long-range attitude adjustment. This system operates through a spatiotemporal collaborative detection mechanism to achieve seamless coverage. The radar module in the radar and visual system transmits a frequency-modulated continuous wave (FCMCC) signal. When this signal encounters a target, it generates a reflected echo. The frequency of the echo signal undergoes a Doppler shift due to the target's radial motion. After receiving the echo signal, the radar module uses the Doppler effect to analyze the target's radial velocity. Simultaneously, the radar module generates a point cloud containing a large number of data points. Using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm, the radar module groups densely distributed vehicles into clusters based on their spatial position and velocity similarity. This achieves point cloud target separation, avoids detection confusion caused by densely packed targets during beam adjustment, ensures accurate tracking in long-range multi-target scenarios, and provides clear targets for attitude adjustment. The vision module uses a modified YOLOv5 algorithm to extract the vehicle's visual features in real time and output the target's position in the image coordinate system. By calibrating parameters, the visual positioning results can be converted into three-dimensional coordinates in the radar coordinate system and fused with the radar point cloud data to form a more accurate target position mapping, achieving high-precision spatial positioning of the target. Through a spatiotemporal collaborative detection mechanism, the data monitored by the radar module and the visual module are synchronized in time and space, achieving multi-source data fusion, ensuring that long-range posture adjustment actions match the target state in real time, and establishing high-precision, real-time perception capabilities for long-range target posture, thereby enabling the terminal to dynamically adjust itself or system strategies based on the target state.

[0027] Dynamic beam control technology has been developed to address the special working conditions of curves. The vision module assists in determining the current road curvature by monitoring the curvature of the lane lines and the vehicle's posture, providing redundant input for the calculation of the radar beam deflection angle and improving the robustness of dynamic adjustment. The radar beam direction can automatically deflect within a certain angle according to the road curvature, effectively eliminating the inner blind spot.

[0028] At ordinary highway construction sites, before deploying the intrusion warning system in the highway curve construction area, it is necessary to configure the construction personnel and the site in advance. The details are as follows: Construction personnel and work areas are aligned. A "three zones, five posts" deployment system is implemented within the 800-meter radius right-hand curve construction area. Eight construction personnel are deployed in the core work area, including two equipment operators holding special operations certificates, who are stationed on the inside of the curve, responsible for pavement milling operations. A distance of at least 15 meters is maintained between personnel. Four auxiliary workers are assigned to the outer emergency lane (3.5 meters wide) for material transfer and other construction work, with an activity radius of less than 5 meters. Two safety officers are stationed at the upstream and downstream transition zones, each equipped with a tracking wristband. Their real-time location is updated on a monitoring platform. All construction personnel wear reflective clothing and tracking wristbands. Wearable warning shoulder patches can also be added to display a red alert circle in low visibility conditions.

[0029] The 300-meter-long construction zone is divided into three levels of protection: the upstream transition zone (0-150 meters), the core operation zone (150-300 meters), and the downstream termination zone (above 300 meters). Sixty intelligent lifting cones equipped with built-in tilt sensors are arranged in a 38-degree gradient, with spacing decreasing from 20 meters to 5 meters. Two anti-collision buffer vehicles are parked in the core operation zone, the first at 150 meters and the second at 300 meters. Folding guide screens are installed at the rear of the anti-collision buffer vehicles. Twenty sets of solar-powered road studs are deployed in the downstream termination zone to guide traffic back to normal lanes. The curb strip at the superelevated section of the curve is temporarily widened by 0.5 meters and separated by removable steel guardrails, with guardrail posts spaced 3 meters apart and covered with reflective film.

[0030] After the construction personnel and on-site operations are configured, the highway curve construction area intrusion warning system will be deployed at ordinary highway construction sites. The highway curve construction area intelligent warning system realizes all-weather safety protection of the construction area based on multi-source perception and collaborative decision-making mechanism. The vehicle driving trajectory data is collected in real time through the multi-level radar and vision integrated machine deployed around the construction area. The multi-level radar and vision integrated machine consists of multiple millimeter-wave radars, ultra-wideband radars and high-definition cameras. The millimeter-wave radar and ultra-wideband radar detect the target distance and target speed through frequency modulation continuous wave technology. The target distance r is calculated by the formula:

[0031] Where c is the speed of light (3×10 8 m / s), Δt is the time difference between the transmitted and received signals, and B=4GHz bandwidth.

[0032] The target velocity measurement is based on the Doppler effect. The target radial velocity measurement v is calculated as follows:

[0033] Where λ is the wavelength (3.9 mm) and fd is the Doppler shift frequency. The radar uses this frequency shift to determine the target's velocity.

[0034] Millimeter wave radar uses the penetration characteristics of frequency modulated continuous waves to analyze the phase difference of the echo signal. And Doppler frequency shift fd, obtain the radial velocity v and rough position of the obscured vehicle. The phase difference of the echo signal The phase difference is the time difference between the wave emitted by the radar and the wave reflected back. Through this time difference, the radar can calculate the distance between the vehicle and the device. The relationship between the target distance R is ,in, It is the frequency modulated continuous wave wavelength emitted by the radar, which can penetrate the gaps in the truck cargo box to detect targets behind.

[0035] The camera analyzes local visible features based on the YOLOv5 model, such as the outline of the headlights of the blocked vehicle, the edge of the tire and other fragment information, and converts the image coordinates into Mapping to the world coordinate system When the radar and vision simultaneously detect similar motion features, the system determines it as a valid target. The coordinate conversion calculation method is as follows:

[0036] Among them, F is the perspective transformation function, which converts the image coordinates of a point in the 2D picture taken by the camera into the coordinates in the real 3D world.

[0037]

[0038] Where a is the focal length, a core parameter of the camera. A larger focal length results in more accurate ranging. H is the typical vehicle height (1.5m), and h is the pixel height of the detection frame. Spatiotemporal synchronization utilizes a precision clock protocol with a clock synchronization accuracy of ±100ns. The calculation method is as follows:

[0039] in, To synchronize time; is the local time; is a fixed offset; For network delay.

[0040] The different perception data detected by radar and camera are integrated into a three-level processing architecture. First, time and space registration is performed, and the coordinate transformation matrix is ​​used to transform the data. The data from different sensors are unified into the world coordinate system, where R is a 3×3 rotation matrix and t is a 3×1 translation vector. The improved joint probabilistic data association algorithm is used to associate the vector data with the target. The association cost function is:

[0041] in is the spatial distance between target i and measurement j, is the velocity difference between target i and measurement j. For occluded targets, a constant turning rate and acceleration model is used for trajectory prediction. The state equation contains parameters such as position, velocity, heading angle, and turning rate. The data fusion weight w is dynamically adjusted according to the sensor confidence level. Its calculation formula is as follows:

[0042] Where c is the confidence level and σ is the standard deviation of the measurement noise. The final output is the weighted optimal estimate of the target state.

[0043] The fused radar monitoring data and camera monitoring data form target perception data. After the target perception data forms a data packet in the time-sensitive switch, it is input into the decision layer through the event-sensitive switch. The data packet format includes fields such as timestamp (32 bits), position coordinates, and velocity vector.

[0044] The decision layer performs risk assessment based on the received perception data. The risk assessment is based on the lane benchmark model and vehicle dynamic parameters. The lane model is fitted using a cubic B-spline curve, and the parameter curve equation C(u) is

[0045] in is the cubic B-spline basis function, For the control point.

[0046] The curvature calculation uses the parametric equation derivation method, and the final curvature for:

[0047] in, is the first derivative of the curve, is the second-order derivative of the curve, is the cross product operation of two-dimensional vectors, is the magnitude (length) of the vector. For the final curvature Typical values ​​are: For a straight line, is 0; for a curve with a radius of 500m, 0.002 m -1 ; For a sharp bend with a radius of 200m, 0.005 m -1.

[0048] The overall risk level of vehicles entering the construction area is defined as risk level S. It is calculated as follows:

[0049] Where, the lateral offset is d (weight 0.4), and the normalized velocity is , The value is usually 0.3, and the normalized acceleration is , The general value is 0.2, and the curvature is The value is usually 0.1. The normalized lateral offset d is calculated by the formula

[0050] in, is the current coordinate of the vehicle under test in the two-dimensional rectangular coordinate system established with the entrance of the construction area as the origin. It is the coordinate point of the lane centerline reference point closest to the vehicle, which is calculated in real time through the lane model. is the angle between the longitudinal axis of the vehicle and the tangent direction of the lane centerline, that is, the current heading angle of the vehicle. is the lane width.

[0051] The risk level S is divided into three levels: Level I (S < 0.3) triggers a preventive warning. At this level, only visual cues are used to draw attention to avoid excessive interference with normal vehicle driving. Specific operations: The LED variable information board displays a yellow prompt "Drive with caution during construction ahead" message, the smart cone slowly flashes yellow, the directional sound wave transmitter is deactivated, the construction worker's wristband does not vibrate, and the warning shoulder badge does not project lasers; Level II (0.3 ≤ S < In a further preferred solution of this embodiment, when communication is congested, the system processes requests in order of transmission priority P value, which can achieve dynamic response adjustment of communication. Transmission priority The calculation is as follows:

[0052] Where S is the risk index and D is the distance (meters). It is a dynamic value accurately calculated by the radar and vision fusion system, representing the shortest vertical distance between the target vehicle and the dangerous boundary of the construction area.

[0053] The risk level S and distance D are converted into a standardized value between 0 and 1. Communication transmission is carried out stably according to the transmission time slot. The allocated transmission time slot The calculation of time slot allocation algorithm is as follows:

[0054] Calculate the allocated transmission time slots in real time based on transmission priority , thereby dynamically allocating the priority of sending decision-making instructions. The higher the risk level, the faster the system "jumps the queue" to grab the communication channel, ensuring that emergency alerts are issued first.

[0055] On curved highways, if a large vehicle (such as a truck or trailer) obstructs a smaller vehicle behind it, it creates a perception blind spot, preventing the system from promptly identifying potential intrusions. This paper addresses this issue by designing a kinematic-based trajectory prediction compensation. This prediction model uses a constant turning rate and acceleration (CTRA) motion model to predict target trajectories. The state update equation of the CTRA prediction model is implemented using Euler integration, and the state update is as follows:

[0056]

[0057]

[0058]

[0059] In the above formulas, the first formula calculates the lateral velocity component by taking the cosine value of the current heading angle ψ(t), thereby realizing the update of the lateral position x; the second formula calculates the longitudinal velocity component by taking the sine value of the current heading angle ψ(t), thereby realizing the update of the longitudinal position y; the third formula obtains the current turning rate Calculate the angle increment and thus the heading angle ; In the fourth formula, by obtaining the current acceleration Calculate the speed change and thus the speed Update value. represents the time step, is the vehicle plane coordinate, is the total speed, is the heading angle, is the turning rate, is the acceleration. The first two formulas represent the prediction of position, and the last two represent the prediction of heading angle and velocity respectively. The motion state is described as a six-tuple The model continuously corrects the prediction results through Kalman filtering, and can immediately trigger an early warning when the closest distance between the predicted trajectory and the construction zone boundary is less than the safety threshold. The detailed implementation steps are as follows: When the radar module detects an unusual cluster of point clouds behind a large vehicle and the vision module identifies a fragmented feature, the system establishes a temporary tracking target. A joint probabilistic data association algorithm is used to calculate the matching confidence between the radar point cloud and the visual features, with the association weights determined by spatial consistency and motion continuity.

[0060] If the target is completely blocked, the system starts prediction based on the last valid observation state. Taking into account the geometric constraints of the curve, the final curvature of the road is introduced. As a constraint: For a curve with radius r, calculate the theoretical turning rate: ; Actual turning rate Take the historical mean of the vehicle and The weighted value of; the lateral offset is calculated in real time during the prediction process ,in, is the predicted lane centerline coordinate, is the lane centerline coordinate, when Continues to grow and exceeds the lane width When the number of intrusions reaches 40%, it is determined that there is a tendency to break in.

[0061] When the target reappears in the sensor's field of view, the system uses the Mahalanobis distance to check the consistency between the predicted trajectory and the actual observation. If the deviation exceeds the threshold, the tracking is reset. For predicted targets that are continuously occluded, the warning duration is inversely proportional to the curve radius. The empirical formula for the warning duration is: Where k is a dynamic adjustment coefficient that automatically adjusts based on traffic volume and weather conditions. In rainy weather, k increases by 20%, extending the warning time. r is the curve radius, which is used to avoid excessive warnings based on the warning duration.

[0062] When a vehicle obstructed by a larger vehicle is re-visible by the radar or camera, the system performs an identity verification: it compares the predicted trajectory with the actual observation using the Mahalanobis distance. If the deviation exceeds a safety threshold, the system determines the target has been lost and immediately resets the tracking process. If a match is successful, tracking continues and the prediction model is fine-tuned. For vehicles that are continuously obstructed, the system adopts a dynamic warning duration strategy: smaller curve radius r results in shorter warning time, while larger curve radius r results in longer warning time.

[0063] There are three conditions for the warning to be lifted: the vehicle returns to the normal lane, the warning time expires, or the vehicle leaves the monitoring area. At night, a weak warning will be retained as a buffer, that is, the smart lifting cone will flash slowly. The control parameters of the smart lifting cone are: The flicker frequency f is in Hertz. It is calculated as follows:

[0064] Lifting height h, in meters. The calculation method is:

[0065] The acoustic wave transmitter uses parametric array technology, and the sound intensity , in dB, and its calculation formula is:

[0066] Beamwidth , the unit is degree, and its calculation formula is:

[0067] Among wearable devices, the vibration mode of the construction worker positioning bracelet adopts PWM modulation, with an operating frequency of 15Hz. The duty cycle Dh is calculated as follows:

[0068] The laser projection radius r of the warning shoulder badge is in meters and is calculated as follows:

[0069] LED variable information boards adjust the brightness L through an adaptive algorithm. The unit is nit. The calculation formula is:

[0070] Ensure visibility under all lighting conditions. All execution terminals use hardware-level synchronization to create multi-modal warning effects including sound, light, and vibration.

[0071] The roadside equipment uses FPGA programmable chips, which are equipped with a monitoring module to achieve full link status management. The monitoring module includes: detection rate DR, false alarm rate FAR and average response delay Three major performance indicators. The specific calculation method is as follows: Detection rate DR

[0072] Where TP is the number of correctly identified intrusion events, and FN is the number of missed intrusion events.

[0073] False Alarm Rate (FAR)

[0074] Where FP is the number of false alarms and TN is the number of non-risk states that are correctly ignored.

[0075] Average response delay

[0076]

[0077] in, The response time to detect the early warning trigger, is the timestamp when the target is first identified, and N is the total number of events that occurred during the statistical period.

[0078] In the highway construction zone warning system, the main focus is on parameter optimization of the trajectory prediction algorithm built into the radar-based integrated machine. The parameter optimization uses the stochastic gradient descent method with momentum to calculate the parameter increment. for

[0079] Among them, ▽L is the gradient of the loss function, is the previous parameter increment, η is the learning rate, generally 0.01, and the momentum coefficient α is generally 0.9. The loss function L is defined as

[0080] Throughout this operation, the intelligent power distribution cabinet and solar energy replenishment system ensured continuous power supply. The data storage unit recorded a complete event log and updated it in real time for subsequent analysis. The data storage unit utilizes a ring buffer structure. The storage format is an optimized JSON structure, including fields such as timestamp, event type, and risk level. System health is monitored via a heartbeat mechanism. All monitoring data is stored using blockchain technology, ensuring the reliability of audit traceability.

[0081] Through the above steps and formulas, the system can realize the automation of the entire process from environmental perception, risk assessment to early warning execution, significantly improving the safety protection level of the construction area. At the same time, it has the characteristics of flexible deployment and rapid response, which is particularly suitable for the safety protection needs of temporary construction scenarios on highways.

[0082] In addition, it should be understood that although this specification describes the embodiments, not every embodiment contains only one independent technical solution. This description is for clarity only. Those skilled in the art should consider the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for the purpose of illustrating the technical concept of the present invention and cannot be used to limit the scope of protection of the present invention. Any changes made based on the technical solution in accordance with the technical concept proposed by the present invention fall within the scope of protection of the claims of the present invention.

Claims

1. A highway curve construction area intrusion warning system, characterized by: It includes a perception layer, a decision layer and an execution layer. The perception layer includes several multi-level radar and vision integrated machines, the decision layer is a distributed edge computing node, and the execution layer is several warning devices. The multi-level radar and vision integrated machines and warning devices are distributed and installed upstream of the construction area. The edge computing nodes are deployed inside the anti-collision buffer vehicle in the anti-collision buffer zone of the construction area. The perception layer, decision layer, execution layer and security layer form a backbone communication link through an industrial optical fiber ring network. The multi-level radar and vision integrated machines, edge computing nodes and warning devices use a wireless network as a redundant channel.

2. A highway curve construction zone intrusion warning system according to claim 1, characterized in that: The perception layer includes several multi-level radar and vision integrated machines, including long-range radar and vision integrated machines, medium-range radar and vision integrated machines, short-range radar and vision integrated machines, and roadside millimeter-wave radars; among them, the long-range radar and vision integrated machines are equipped with high-performance millimeter-wave radars, adopt an antenna array design, and are equipped with infrared thermal imagers; the medium-range radar and vision integrated machines integrate millimeter-wave radars and cameras; and the short-range radar and vision integrated machines (construction area boundaries) use a combination of ultra-wideband radars and cameras.

3. The highway curve construction zone intrusion warning system according to claim 2 is characterized in that: A time-sensitive switch is also provided in the perception layer. The time-sensitive switch is deployed at the same site as the radar-visual integrated machine. The target perception data monitored by the radar-visual integrated machine is exchanged through the time-sensitive switch.

4. The highway curve construction zone intrusion warning system according to claim 1 is characterized in that: The decision layer is also provided with a data storage unit, which is used to record complete warning data, form an event log, and update the log in real time for subsequent data analysis by staff.

5. The highway curve construction zone intrusion warning system according to claim 1 is characterized in that: The execution layer is composed of several warning devices, which include intelligent lifting cones, directional sound wave transmitters and LED variable information boards. The intelligent lifting cones are hydraulically driven, and their columns are integrated with LED lamp beads. The intelligent lifting cones are arranged in an oblique line in the transition zone upstream of the construction area. The directional sound wave transmitter is composed of a phased array. The LED variable information board uses an LED panel, and the built-in ambient light sensor can sense the intensity of the surrounding light. The directional sound wave transmitter and the LED variable information board are both arranged upstream of the construction area.

6. The highway curve construction zone intrusion warning system according to claim 5 is characterized in that: The execution layer also includes a construction personnel positioning bracelet, which is worn by construction area workers. The construction personnel positioning bracelet integrates a linear motor and a UWB positioning module, and has UWB positioning and early warning vibration prompt functions.

7. A method for early warning of intrusion into a highway curve construction area, based on a system for early warning of intrusion into a highway curve construction area according to any one of claims 1 to 6, characterized in that: The early warning methods are: The perception layer uses multi-level radar and vision integrated devices to monitor the target perception data of intruders; The decision layer receives target perception data, determines the target's intrusion risk through the edge computing node, and generates corresponding decision instruction data packets; The execution layer receives the decision instruction data packet and controls the warning device to issue corresponding warnings according to the decision instruction data packet.

8. The method for early warning of intrusion into a highway curve construction area according to claim 7, characterized in that: The data monitored by the multi-level radar and vision integrated machine includes different monitoring data monitored by the radar module and the vision module. The different monitoring data monitored by the radar module and the vision module are synchronized in time and space through the time-space collaborative detection mechanism to achieve multi-source data fusion, and then form target perception data. The target perception data is transmitted to the decision-making layer through the time-sensitive switch.

9. The method for early warning of intrusion into a highway curve construction area according to claim 7, characterized in that: All radar and vision integrated devices in the multi-level radar and vision integrated device support long-distance posture adjustment. Its working principle is to synchronize the data monitored by the radar module and the vision module in time and space through a spatiotemporal collaborative detection mechanism, realize multi-source data fusion, ensure that the long-distance posture adjustment action matches the target status in real time, and build a high-precision, real-time perception capability of the long-distance target posture, thereby supporting the terminal to dynamically adjust itself or the system strategy according to the target status.

10. The method for early warning of intrusion into a highway curve construction area according to claim 7, characterized in that: The intrusion risk is divided into three levels: Level I (S<0.3) triggers a preventive reminder, Level II (0.3≤S<0.7) activates an audible and visual alarm, and Level III (S≥0.7) executes an emergency warning.

Images