Low-speed working vehicle collision early warning system and method for road area traffic perception and intervention

Abstract

The invention belongs to the technical field of traffic intelligent perception, and particularly relates to a road area traffic perception and intervention low-speed operation vehicle collision early warning system and method, and the system comprises a data collection module, a data transmission module, a data processing module, a collision risk judgment module and an early warning information issuing module. Data used by the method are collected data of millimeter wave radar and camera detection equipment installed on a movable device, the method is convenient and rapid to use, real-time radar data is adopted, the method has the characteristics of high detection precision and high detection speed, real-time video data is adopted, the method has the characteristics of light weight and low deployment cost, and the method is suitable for popularization and application. And real-time sensing of vehicle position and size information, weather information and personnel position information can be realized through the edge calculation unit. And the constructor out-of-bound early warning module can judge the position information of the constructors according to the outer contour of the construction operation area and perform early warning prompt on the constructors exceeding the construction operation area, so that the possible interference of the constructors on traffic operation is reduced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent traffic perception, and particularly relates to a low-speed operation vehicle collision warning system and method for road domain traffic perception and intervention. BACKGROUND

[0002] With the popularity of motor vehicles and the growth of transportation demand, the influence of highway construction on traffic operation is becoming greater and greater, and the safety threat of running vehicles to construction personnel is also becoming greater. At present, there is a lack of corresponding safety construction intelligent early warning information system for the various illegal out-of-bound behaviors of highway construction personnel, and the safety of construction personnel needs to be protected by using current advanced position sensing technology. In addition, in the process of road maintenance, in addition to the commonly used closed operation, the mobile operation method is also commonly used. Low-speed operation vehicles often travel at a low speed, and part of the traveling vehicles are prone to collision accidents with low-speed operation vehicles in the case of driver distraction driving, especially in the case of low visibility in abnormal weather. Therefore, advanced sensing technology and information early warning methods also need to be used to protect the safety of vehicles traveling around the low-speed operation vehicle. However, the traditional vehicle position information collection relies on road side fixed sensors, and the meteorological sensor relies on a complex device such as a micro-meteorological station that needs to be fixed for detection. However, the low-speed operation vehicle is prone to traffic risks or construction risks in the complex road environment during operation, and the equipment is difficult to deploy in the whole domain and has a high cost. SUMMARY

[0003] In view of the technical problems in the above-mentioned traditional vehicle position information collection, the application provides a low-speed operation vehicle collision warning system and method for road domain traffic perception and intervention.

[0004] In order to solve the above technical problems, the technical scheme adopted by the application is: The low-speed operation vehicle collision warning system for road traffic perception and intervention comprises a data acquisition module, a data transmission module, a data processing module, a collision risk judgment module and a warning information release module. The data acquisition module uses the millimeter wave radar and camera installed on the low-speed operation vehicle to collect the expressway vehicle position, vehicle size, license plate information, construction personnel position and weather information in real time. The data processing module uses the edge computing unit to perform structured processing and matching analysis on the vehicle, personnel track data, vehicle size and license plate data and weather data. The data transmission module uses the optical fiber or wireless signal transmitter to send the edge computing unit data to the cloud. The collision risk judgment module uses the LSTM-CNN model to predict the future vehicle position, uses the neural network model to predict the vehicle lane changing behavior, and uses the vehicle, personnel position data, vehicle speed data, vehicle size data and weather data to construct the current and future short-time multi-dimensional collision risk judgment model. According to the construction work area contour, the vehicle track perception data and weather perception data are used to judge the collision risk of the low-speed operation vehicle and the surrounding vehicles and the collision risk of the construction personnel and the surrounding vehicles. The warning information release module uses the intelligent wearable device to remind the construction personnel of the risk and uses the high-pitched directional horn to remind the vehicle that may collide of the risk.

[0005] The collision risk judgment module comprises a construction personnel out-of-boundary warning module and a construction collision risk judgment module. The construction personnel out-of-boundary warning module judges the position information of the construction personnel according to the construction work area contour and warns the construction personnel who are out of the construction work area. The construction collision risk judgment module judges and warns the possible collision risk between vehicles according to the vehicle position, speed, size on the upstream of the construction area and the position of the construction personnel in combination with the weather condition.

[0006] The construction personnel out-of-boundary warning module marks the contour coordinates of the construction work area in the coordinate system. If the position coordinates of the construction personnel exceed the contour coordinates of the construction work area, the out-of-boundary information warning is given to the construction personnel. The construction collision risk judgment module collects the vehicle track on the upstream of the construction area, uses the LSTM-CNN track prediction model to perform short-time prediction on the relative position of the vehicle to the low-speed operation vehicle, uses the neural network model to perform short-time prediction on the lane changing behavior of the vehicle, and simultaneously constructs the low-speed operation vehicle collision risk model considering the accident severity according to the predicted vehicle track and the low-speed operation vehicle collision probability, the vehicle size and the weather information recognized by the video, to judge the construction collision risk.

[0007] The low-speed operation vehicle collision warning method for road traffic perception and intervention comprises the following steps: Step one, fix the millimeter wave radar and camera through the movable support on the upstream endpoint of the construction area. Step 2: With the camera position as the origin, establish a plane rectangular coordinate system with the road direction as the X-axis. The coordinates and speeds involved in this plane rectangular coordinate system are all relative coordinates and relative speeds. Measure the relative position of the millimeter wave radar with respect to the camera and assign the millimeter wave radar relative coordinates ( ); Step 3: Use millimeter wave radar and camera to collect the coordinates of the construction vehicle at the same position. The relative coordinates of the vehicle centroid collected by the millimeter wave radar are ( ), the relative coordinates of the vehicle centroid collected by the camera are ( ); Step 4: Unify the vehicle positions obtained by the millimeter-wave radar and the camera; Step 5: In a unified coordinate system, collect lane-level position trajectory information and construction worker location information, including vehicle horizontal and vertical coordinates, vehicle speed, and acceleration. Simultaneously, use cameras to collect license plate data, vehicle dimensions, and construction worker horizontal and vertical coordinate data. Match the license plate data with the vehicle trajectory data. The frequency of vehicle and construction worker location collection and transmission should be milliseconds, and the positioning accuracy of vehicle and construction worker locations should be decimeters. Step 6: The data processing module is responsible for data structured processing and analysis on the edge and in the cloud. It uses a distributed processing method to match radar vehicle trajectory information, license plate information, and vehicle size information according to the license plate information and process them into structured data. The input radar vehicle trajectory information is in radar frame data format, and the output structured vehicle trajectory data is vehicle trajectory time series data with vehicle latitude and longitude coordinates, speed information, acceleration information, vehicle weight information, and vehicle type information, which completes the unification of global vehicle IDs and trajectory splicing based on time series. The time series uses UTC time, and the time delay with universal time should be within 100 milliseconds. The vehicle and construction personnel movement trajectory collection frequency is 10Hz; Step 7: Real-time meteorological data is collected using image data obtained by cameras. The image data is used to analyze rain, snow, ice, frost, road icing, and visibility. The collection frequency is 5 minutes.

[0008] The method for unifying the vehicle position obtained by the millimeter wave radar and the camera is as follows: Use the plane rectangular coordinate system established in step 2 to unify the coordinates; the position of a vehicle collected by the millimeter wave radar is ( ), then it is converted to Thus, a unified coordinate system is obtained for vehicles and construction workers, with the millimeter-wave radar as the origin and the road forward direction as the positive direction of the X-axis.

[0009] The following steps are also included: Step 8: Use the vehicle's real-time and historical trajectory data as input, and the predicted vehicle trajectory as the output of the LSTM-CNN. The trajectory prediction model uses a training window of 5 seconds and a prediction window of 2 seconds. The predicted vehicle lane-changing behavior is used as the output of the neural network model. Lane-changing behavior is a discrete output model. By comparing it with the actual vehicle trajectory data and adjusting the network parameters in reverse order, the vehicle trajectory behavior is finally predicted. Step 9: Based on the established LSTM-CNN trajectory prediction model and the vehicle lane-changing behavior model based on the neural network model, the vehicle's future short-term position information is obtained, and the real-time and future short-term relative position and relative speed of the vehicle and the low-speed working vehicle are determined, and the real-time PET and future short-term PET between the two are calculated; Step 10: Comprehensively construct a low-speed vehicle collision risk index based on video recognition of vehicle size and meteorological information, taking into account the severity of the accident. , calculate the real-time collision risk index of vehicles running in the same lane as the low-speed working vehicle .

[0010] The real-time PET calculation method is: is the collision time of vehicle i relative to the low-speed vehicle at time t. Since the vehicle position collected by the data is the relative position of the vehicle head, is the position of the front of vehicle i relative to the millimeter-wave radar on the low-speed operation vehicle at time t, It is the distance from the millimeter-wave radar position on the low-speed operation vehicle to the rear of the low-speed operation vehicle. is the instantaneous speed of car i at time t; The future short-term PET is calculated as follows: It's iCar The time is relative to the collision time of the low-speed working vehicle. It is predicted based on the LSTM-CNN model At time i, the position of the front of the vehicle relative to the millimeter-wave radar on the low-speed working vehicle, i.e., the origin, It is the distance from the millimeter-wave radar position on the low-speed operation vehicle to the rear of the low-speed operation vehicle. It is predicted based on the LSTM-CNN model The instantaneous speed of car i at time i.

[0011] The calculation is to calculate the real-time collision risk index of vehicles running in the same lane as the low-speed working vehicle The method is: * * is the instantaneous speed of car i at time t, is the speed limit of the road section at time t, is the collision time of vehicle i relative to the low-speed vehicle at time t, so is the vehicle size coefficient of car i, is the weather coefficient at time t; in, The values ​​of are as follows: in, The values ​​of are as follows: Computing the Future Real-time collision risk indicator for vehicles that may be operating in the same lane as a low-speed operation vehicle The calculation is as follows: * * * It is t The instantaneous speed of car i at time i, It is t The speed limit value of the road section at that moment, It is i car in t The time is relative to the collision time of the low-speed working vehicle, so is the vehicle size coefficient of car i, It is t Weather coefficient at the moment, is the number of surrounding vehicles t calculated based on the neural network model The probability of driving in the same lane as the low-speed operation vehicle at any given moment.

[0012] The warning information release module is based on the low-speed vehicle collision risk index Make a judgment and the early warning prompt method is: .

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The data used in the method of the present invention is collected by millimeter-wave radar and camera detection equipment installed on a mobile device. It is easy and quick to use. The real-time radar data used has the characteristics of high detection accuracy and fast detection speed. The real-time video data used is lightweight and has low deployment cost. The edge computing unit can realize real-time perception of vehicle location and size information, meteorological information and personnel location information.

[0014] 2. In the present invention, the construction worker out-of-bounds warning module can judge the construction worker's location information based on the outer contour of the construction work area, and issue early warning prompts to construction workers who have exceeded the construction work area, thereby reducing the possible interference of construction workers on traffic operations.

[0015] 3. The construction collision risk judgment module of the present invention judges the probability of possible collision between vehicles based on the relative position and speed of upstream vehicles in the construction area and low-speed work vehicles, and uses the vehicle lane change prediction model to judge the collision risk of vehicles in adjacent lanes. At the same time, combined with the vehicle size and meteorological information obtained by the camera, it constructs a comprehensive construction collision risk that comprehensively considers the collision probability and collision severity to make judgments and early warnings, thereby improving the accuracy, effectiveness and a priori nature of real-time and future short-term collision judgments between vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can, without inventive effort, derive other implementation drawings based on the provided drawings.

[0017] The structures, proportions, sizes, etc. illustrated in this specification are intended solely to complement the contents disclosed herein and to facilitate understanding and reading by persons skilled in the art. They are not intended to limit the conditions under which the present invention may be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportions, or adjustments in sizes, without affecting the efficacy and objectives of the present invention, shall remain within the scope of the technical contents disclosed herein.

[0018] Figure 1 It is a schematic diagram of the principle flow of the present invention; Figure 2 This is a schematic diagram of the traffic operation risk range upstream of the construction area of ​​the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of this application, not all the embodiments. These descriptions are only to further illustrate the features and advantages of the present invention, rather than to limit the claims of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0020] The specific embodiments of the present application are described in further detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0021] A low-speed working vehicle collision risk early warning method and system based on road area traffic perception and driving intervention, comprising a data acquisition module, a data transmission module, a data processing module, a collision risk judgment module, and an early warning information publishing module.

[0022] Preferably, the data acquisition module uses the millimeter wave radar and camera installed on the low-speed working vehicle to collect real-time highway vehicle position, vehicle size, license plate information, construction personnel position, and weather information.

[0023] Preferably, the data processing module uses an edge computing unit to perform structured processing and matching analysis on vehicle and personnel trajectory data, vehicle size and license plate data, and weather data.

[0024] Preferably, the data transmission module uses optical fiber or a wireless signal transmitter to send edge computing unit data to the cloud.

[0025] Preferably, the collision risk judgment module uses an LSTM-CNN model to predict future vehicle positions, uses a neural network model to predict vehicle lane-changing behavior, and uses vehicle and personnel position data, vehicle speed data, vehicle size data, and weather data to construct a current and future short-time multi-dimensional collision risk research and judgment model. The collision risk of the low-speed working vehicle and the surrounding vehicles, as well as the collision risk of the construction personnel and the surrounding vehicles, can be judged according to the working zone outline, using vehicle trajectory perception data and weather perception data.

[0026] Preferably, the early warning information publishing module uses intelligent wearable devices to remind construction personnel of risks and uses a high-pitched directional horn to remind vehicles that may collide of risks.

[0027] Specifically, the data acquisition module and data processing module workflow of the present embodiment includes the following steps: Step one: fix the millimeter wave radar and camera on the upstream endpoint of the construction area through a movable support; Step two: establish a plane rectangular coordinate system with the camera position as the origin and the road forward direction as the X-axis. The coordinates and velocities involved in this plane rectangular coordinate system are all relative coordinates and relative velocities. Measure the relative position of the millimeter wave radar relative to the camera and assign the relative coordinates of the millimeter wave radar as ); Step three: use the millimeter wave radar and camera to collect the coordinates of the construction vehicle at the same position. The vehicle centroid relative coordinates collected by the millimeter wave radar are , and the vehicle centroid relative coordinates collected by the camera are ); Step 4: Unify the vehicle positions obtained by the millimeter-wave radar and the camera using the plane rectangular coordinate system established in step 2; the position of a vehicle collected by the millimeter-wave radar is ( ), then it is converted to Thus, a unified coordinate system is obtained for vehicles and construction workers, with the millimeter-wave radar as the origin and the road forward direction as the positive direction of the X-axis.

[0028] Step 5: In a unified coordinate system, collect highway lane-level position trajectory information and construction personnel location information, including vehicle horizontal and vertical coordinates, vehicle speed, and acceleration. At the same time, use cameras to collect license plate data, vehicle size data, and construction personnel horizontal and vertical coordinate data, and match the license plate data with vehicle trajectory data. The frequency of vehicle position and construction personnel position collection and transmission should be milliseconds, and the positioning accuracy of vehicle position and construction personnel position should be decimeters.

[0029] Step 6: The data processing module is responsible for data structured processing and analysis at the edge and in the cloud. It uses distributed processing methods to match radar vehicle trajectory information, license plate information, and vehicle size information according to the license plate information and process them into structured data. The input radar vehicle trajectory information is in radar frame data format, and the output structured vehicle trajectory data is vehicle trajectory time series data with vehicle latitude and longitude coordinates, speed information, acceleration information, vehicle weight information, and vehicle type information, which completes the unification of global vehicle IDs and trajectory splicing based on time series. The time series uses UTC time, and the time delay with universal time should be within 100 milliseconds. The vehicle and construction personnel motion trajectory collection frequency is 10Hz.

[0030] Step 7: Real-time meteorological data is collected using image data obtained by cameras. The image data is used to analyze rain, snow, ice, frost, road icing, and visibility. The collection frequency is 5 minutes.

[0031] Specifically, the collision risk judgment module of this embodiment includes a construction worker out-of-bounds warning module and a construction collision risk judgment module; The construction worker out-of-bounds warning module determines the construction worker's location information based on the outer contour of the construction work area and issues warnings to construction workers who have gone beyond the construction work area; The construction collision risk assessment module assesses and issues early warnings of potential collisions between vehicles based on the location, speed, size, and location of upstream vehicles and construction personnel in the construction area, combined with meteorological conditions. Specifically, the construction worker out-of-bounds warning module marks the coordinates of the outer contour of the construction work area in the coordinate system. If the construction worker's position coordinates exceed the outer contour coordinates of the construction work area, an out-of-bounds information warning prompt will be issued to the construction worker. The construction collision risk judgment module collects the vehicle trajectories upstream of the construction area, uses the LSTM-CNN trajectory prediction model to make short-term predictions of the vehicle's position relative to the low-speed work vehicle, and uses a neural network model to make short-term predictions of the vehicle's lane changing behavior. At the same time, based on the predicted vehicle trajectory and the probability of collision with the low-speed work vehicle, video recognition of vehicle size and meteorological information, a low-speed work vehicle collision risk model is comprehensively constructed, taking into account the severity of the accident, to judge the construction collision risk. The specific steps are as follows: Step 1: The real-time and historical vehicle trajectory data are used as input, and the predicted vehicle trajectory is used as the output of the LSTM-CNN. The trajectory prediction model uses a training window of 5 seconds and a prediction window of 2 seconds. The predicted vehicle lane-changing behavior is used as the output of the neural network model. Lane-changing behavior is a discrete output model. By comparing it with the actual vehicle trajectory data and adjusting the network parameters in reverse order, the vehicle trajectory behavior is finally predicted. Step 2: Based on the established LSTM-CNN trajectory prediction model and the vehicle lane-changing behavior model based on the neural network model, the vehicle's future short-term position information is obtained. The real-time and future short-term relative position and relative speed between the vehicle and the low-speed working vehicle are determined, and the real-time PET and future short-term PET between the two are calculated. The real-time PET is calculated as follows: is the collision time of vehicle i relative to the low-speed vehicle at time t. Since the vehicle position collected by the data is the relative position of the vehicle head, is the position of the front of vehicle i relative to the millimeter-wave radar on the low-speed operation vehicle at time t, It is the distance from the millimeter-wave radar position on the low-speed operation vehicle to the rear of the low-speed operation vehicle. is the instantaneous speed of car i at time t.

[0032] The future short-term PET is calculated as follows: It's iCar The time is relative to the collision time of the low-speed working vehicle. It is predicted based on the LSTM-CNN model At time i, the position of the front of the vehicle relative to the millimeter-wave radar on the low-speed working vehicle, i.e., the origin, It is the distance from the millimeter-wave radar position on the low-speed operation vehicle to the rear of the low-speed operation vehicle. It is predicted based on the LSTM-CNN model The instantaneous speed of car at time i.

[0033] Step 3: Based on video recognition of vehicle size and meteorological information, a low-speed vehicle collision risk index is constructed that takes into account the severity of the accident. , calculate the real-time collision risk index of vehicles running in the same lane as the low-speed working vehicle Here’s how: * * is the instantaneous speed of car i at time t, is the speed limit of the road section at time t, is the collision time of vehicle i relative to the low-speed vehicle at time t, so is the vehicle size coefficient of car i, is the weather coefficient at time t.

[0034] in, The values ​​of are as follows: in, The values ​​of are as follows: Computing the Future Real-time collision risk indicator for vehicles that may be operating in the same lane as a low-speed operation vehicle The calculation is as follows: * * * It is t The instantaneous speed of car i at time i, It is t The speed limit value of the road section at that moment, It is i car in t The time is relative to the collision time of the low-speed working vehicle, so is the vehicle size coefficient of car i, It is t Weather coefficient at the moment, is the number of surrounding vehicles t calculated based on the neural network model The probability of driving in the same lane as the low-speed operation vehicle at any given moment.

[0035] Preferably, the warning information release module is based on the low-speed vehicle collision risk index Make a judgment and the early warning prompt method is as follows: The application uses millimeter wave radar and camera detection equipment to collect data, is convenient to use, is fast to deploy, uses real-time radar data and image data, has the characteristics of high detection precision, fast detection speed and low cost, and can realize real-time perception of vehicle position information, vehicle size information and weather information through an edge computing unit; a collision risk judgment module can judge the collision risk of a low-speed working vehicle and surrounding vehicles and the collision risk of construction personnel and surrounding vehicles according to the working area contour, using vehicle trajectory perception data and weather perception data; and an early warning information publishing module can use a high-pitched directional horn and wearable devices to give graded early warning prompts to vehicles and construction personnel that may collide based on possible collision risks, realize risk control during the operation of a low-speed working vehicle, and solve the problems that collision risks during the operation of a low-speed working vehicle are difficult to quantify and collision early warning is difficult to pass to construction personnel and related vehicles, propose a low-speed working vehicle collision risk early warning algorithm that fuses weather information and collision severity, solve the problem that different types of vehicle collision risks cannot be quantified, and the application has the characteristics of being replicable, strong robustness and the like.

[0036] The above only describes the preferred embodiments of the application in detail, but the application is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the application, and all the changes shall be included in the protection scope of the application.

Claims

1. A low speed work vehicle collision warning system for road domain traffic awareness and intervention, characterized by: The application relates to a low-speed operation vehicle collision risk early warning system, which comprises a data acquisition module, a data transmission module, a data processing module, a collision risk judgment module and an early warning information publishing module. The data acquisition module utilizes a millimeter wave radar and a camera installed on a low-speed operation vehicle to collect highway vehicle position, vehicle size, license plate information, construction personnel position and meteorological information in real time. The data processing module utilizes an edge computing unit to perform structured processing and matching analysis on vehicle, personnel track data, vehicle size and license plate data and meteorological data. The data transmission module utilizes an optical fiber or a wireless signal transmitter to send the edge computing unit data to the cloud. The collision risk judgment module utilizes an LSTM-CNN model to predict future vehicle positions, utilizes a neural network model to predict vehicle lane changing behavior, utilizes vehicle and personnel position data, vehicle speed data, vehicle size data and meteorological data to construct a current and future short-time multi-dimensional collision risk judgment model, judges the collision risk of a low-speed operation vehicle and surrounding vehicles and the collision risk of construction personnel and surrounding vehicles according to the outline of a construction operation area and the vehicle track sensing data and meteorological sensing data, and the early warning information publishing module utilizes intelligent wearable devices to remind the construction personnel of risks and utilizes a high-pitched directional loudspeaker to remind the vehicles possibly colliding of risks.

2. The low speed work vehicle collision warning system for road- domain traffic awareness and intervention of claim 1, wherein: The collision risk judgment module comprises a construction personnel out-of-bound early warning module and a construction collision risk judgment module. The construction personnel out-of-bound early warning module judges the position information of the construction personnel according to the outline of the construction operation area and early warns and prompts the construction personnel out of the construction operation area. The construction collision risk judgment module judges and early warns the collision risk possibly occurring between vehicles according to the position, speed, size of the vehicles upstream of the construction area and the position of the construction personnel and in combination with the meteorological condition.

3. The low speed work vehicle collision warning system for road way awareness and intervention of claim 1, wherein: The construction personnel out-of-bound early warning module marks the outline coordinates of the construction operation area in a coordinate system, and if the position coordinates of the construction personnel exceed the outline coordinates of the construction operation area, the construction personnel is early warned and prompted of out-of-bound information. The construction collision risk judgment module collects the vehicle track upstream of the construction area, adopts an LSTM-CNN track prediction model to perform short-time prediction on the relative low-speed operation vehicle position of the vehicle, adopts a neural network model to perform short-time prediction on the lane changing behavior of the vehicle, simultaneously constructs a low-speed operation vehicle collision risk model considering the accident severity according to the predicted vehicle track and the low-speed operation vehicle collision probability, the vehicle size and the meteorological information recognized by video, and judges the construction collision risk.

4. The method of collision warning for road domain traffic perception and intervention low speed work vehicle, the method is applied to the road domain traffic perception and intervention low speed work vehicle collision warning system as claimed in any one of claims 1-3, characterized in that, The application further relates to a low-speed operation vehicle collision risk early warning method, which comprises the following steps: Step one: fixing the millimeter wave radar and the camera on the upstream endpoint of the construction area through a movable support; Step two, establish a plane rectangular coordinate system with the camera position as the origin and the road forward direction as the X axis. The coordinates and velocities involved in this plane rectangular coordinate system are relative coordinates and relative velocities. Measure the relative position of the millimeter wave radar relative to the camera, and assign the millimeter wave radar relative coordinates (x, y) ). Step three, collect the coordinates of the construction vehicle at the same position by using millimeter wave radar and camera, wherein the relative coordinates of the vehicle centroid collected by the millimeter wave radar are ( ), and the relative coordinates of the vehicle centroid collected by the camera are ( ); Step four: unifying the vehicle positions acquired by the millimeter wave radar and the camera; Step five, collect highway lane-level position trajectory information and construction personnel position information in a unified coordinate system, including vehicle transverse and longitudinal coordinates, vehicle speed and acceleration, while collecting license plate data, vehicle size data, and construction personnel transverse and longitudinal coordinate data with the camera, and matching the license plate data with the vehicle trajectory data, the frequency of vehicle position and construction personnel position collection and transmission should be millisecond level, and the positioning accuracy of vehicle position and construction personnel position should be decimeter level; Step six, the data processing module is responsible for data structuring processing and analysis at the edge and cloud end, and uses distributed processing method to match and process radar vehicle trajectory information, license plate information and vehicle size information into structured data according to license plate information, the input radar vehicle trajectory information is in radar frame data format, and the output structured vehicle trajectory data is time series based vehicle trajectory time series data with vehicle latitude and longitude coordinates, speed information, acceleration information, vehicle weight information and vehicle type information, the time series uses UTC time, and the time delay with universal time should be within 100 milliseconds, and the vehicle and construction personnel motion trajectory collection frequency is 10Hz; Step seven, real-time weather data is collected by image data collected by the camera, which is used to judge rain, snow, ice, road icing and visibility, and the collection frequency is 5 minutes.

5. The low speed work vehicle road user perception and intervention collision warning method of claim 4, wherein, The method for unifying vehicle positions obtained by millimeter wave radar and camera is: Use the plane rectangular coordinate system established in step 2 to unify the coordinates; the position of a vehicle collected by the millimeter wave radar is ( ), then it is converted to ; so as to obtain the unified coordinate system of the vehicle and the construction personnel with the millimeter wave radar as the origin and the road advancing direction as the X-axis positive direction.

6. The low speed work vehicle road user perception and intervention collision warning method of claim 4, wherein, Further comprising the following steps: Step eight, taking real-time trajectory data and historical trajectory data of the vehicle as input values, and taking predicted vehicle trajectory as output values of LSTM-CNN, wherein the training time window of the trajectory prediction model is 5s, the prediction time window is 2s, and the lane changing behavior of the predicted vehicle is taken as the output value of the neural network model, the lane changing behavior is a discrete output model, which is compared with the real vehicle trajectory data, and the network parameters are adjusted in reverse direction, finally realizing the prediction of vehicle trajectory behavior; Step nine, based on the established LSTM-CNN trajectory prediction model and the vehicle future short-time position information obtained by the vehicle lane changing behavior model based on the neural network model, the real-time and future short-time relative position and relative speed of the vehicle and the low-speed operation vehicle are judged, and the real-time PET and future short-time PET between them are calculated; Step ten: Based on video recognition of vehicle size and weather information, a low-speed working vehicle collision risk index considering the severity of the accident is constructed , and the real-time collision risk index of the vehicle running in the same lane as the low-speed working vehicle is calculated .

7. The low speed work vehicle road user perception and intervention collision warning method of claim 6, wherein, The real-time PET calculation method is as follows: is the collision time of the i-th vehicle relative to the low-speed working vehicle at time t, because the position of the vehicle collected by the data is the relative position of the vehicle head, is the position of the i-th vehicle head relative to the millimeter wave radar on the low-speed working vehicle, i.e., the origin, at time t, is the distance from the position of the millimeter wave radar on the low-speed working vehicle to the tail of the low-speed working vehicle, is the instantaneous speed of the i-th vehicle at time t; The future short-time PET calculation method is as follows: is the collision time of the i-th vehicle relative to the low-speed working vehicle at the time is the collision time of the i-th vehicle relative to the low-speed working vehicle at the time is the position of the i-th vehicle head relative to the low-speed working vehicle at the time is the position of the i-th vehicle head relative to the low-speed working vehicle at the time is the distance from the position of the i-th vehicle head relative to the low-speed working vehicle at the time is the position of the i-th vehicle head relative to the low-speed working vehicle at the time is the instantaneous speed of the i-th vehicle at the time 8. The low speed work vehicle road way awareness and intervention collision warning method of claim 6, wherein, The computing a real-time collision risk indicator for a vehicle operating in the same lane as the low-speed work vehicle The method is: * * is the instantaneous speed of the i vehicle at time t, is the speed limit value of the road section at time t, is the collision time of the i vehicle relative to the low-speed working vehicle at time t, so is the vehicle size coefficient of the i vehicle, is the weather coefficient at time t; wherein The values of the parameters are as follows: wherein, The value of the above-mentioned parameters is as follows: Computing the future Real-time collision risk indicators for vehicles that can be operating in the same lane as the low speed work vehicle The computation is as follows: * * * is the instantaneous speed of the i-th vehicle at time t is the instantaneous speed of the i-th vehicle at time t is the speed limit value of the road segment at time t is the speed limit value of the road segment at time t is the collision time of the i-th vehicle relative to the low-speed working vehicle at time t, so is the collision time of the i-th vehicle relative to the low-speed working vehicle at time t, so is the vehicle size coefficient of the i-th vehicle, is the weather coefficient at time t is the weather coefficient at time t is the probability of the surrounding vehicle t is the probability of the surrounding vehicle t 9. The low speed work vehicle road user perception and intervention collision warning method of claim 6, wherein, The pre-warning information issuing module judges according to the low-speed working vehicle collision risk index The judgment is made, and the pre-warning prompt mode is: 。

Images