Human-vehicle collision warning system and method in a connected environment based on pedestrian intention recognition

Abstract

The networked environment human-vehicle collision warning system and method based on pedestrian intention recognition belongs to the field of intelligent transportation technology. The present invention uses the Internet of Vehicles cloud platform as a data interaction platform to collect real-time pedestrian walking parameters and motor vehicle operating parameters from pedestrians' mobile phones and motor vehicles. After identifying the direction of pedestrian movement and predicting the direction of vehicle travel, it predicts whether a traffic accident will occur, thereby pre-alarming vehicles and pedestrians at the same time, thereby reducing the occurrence of traffic accidents involving collisions between pedestrians and vehicles. For situations where there is no alarm and a traffic accident occurs, a new accident model can be established and saved, thereby improving the entire system. The present invention realizes the effective coordination of people, vehicles and roads, can comprehensively protect the safety of pedestrians in traffic, and can also help vulnerable groups in traffic such as the blind, providing them with the possibility of walking freely on the road. It has huge development potential.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and in particular relates to a pedestrian-vehicle collision warning system and method in a networked environment based on pedestrian intention recognition. Background Art

[0002] At present, the automobile industry is developing rapidly, and the number of motor vehicles is increasing rapidly. While cars bring convenience to people, the frequent occurrence of road traffic accidents also threatens the lives of pedestrians. As a vulnerable group among traffic participants, improving the safety of road pedestrians has become a research hotspot in the field of active safety of automobiles. At present, the research on pedestrian safety protection at home and abroad is mainly concentrated in the stage of traffic analysis and discussion, and more is to protect pedestrian safety from the perspective of traffic regulations. At present, the research on pedestrian collision avoidance system is affected by the subjective intention of pedestrians, and the walking trajectory of pedestrians has strong uncertainty, which makes the traditional active collision avoidance system have certain limitations and is not conducive to its widespread promotion and use. The current pedestrian detection technology based on machine vision is the most widely studied pedestrian detection technology, but its application still has great limitations. With the continuous development of vehicle-road collaboration, precise positioning, and real-time communication technology, it has become possible to design a more reasonable and complete intelligent safety warning system for pedestrian travel based on the networked environment. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a connected environment human-vehicle collision warning system and method based on pedestrian intention recognition to solve the technical problem that the current traditional active collision avoidance systems for pedestrians and pedestrian detection technologies based on machine vision have certain limitations and are not conducive to widespread promotion and use.

[0004] The connected environment human-vehicle collision warning system based on pedestrian intention recognition includes a data acquisition module, a data processing module, an information judgment module, a learning module and a warning execution module.

[0005] The data collection module includes a pedestrian data collection module and a moving motor vehicle data collection module; the pedestrian data collection module is installed in the pedestrian's smart phone and connected to the controller of the smart phone, the pedestrian data collection module obtains the pedestrian's real-time geographic location information, walking speed information and walking trajectory information and uploads the collected information to the Internet of Vehicles cloud platform, and the pedestrian data collection module obtains the warning instructions sent by the motor vehicle to the Internet of Vehicles cloud platform from the Internet of Vehicles cloud platform;

[0006] The running motor vehicle data acquisition module, data processing module, information judgment module and learning module are all installed on the motor vehicle. The running motor vehicle data acquisition module is used to obtain information parameters of the running motor vehicle and continuous images of the surrounding environment of the motor vehicle and upload the obtained information to the vehicle information receiving module of the Internet of Vehicles cloud platform in real time; the information parameters of the running motor vehicle include the coordinate position of the motor vehicle, the driving section, the driving speed, the acceleration, the adhesion coefficient, the vehicle mass, the fastest braking time of the brake, the shortest braking distance, the steering wheel rotation angle and the steering wheel rotation angular velocity information;

[0007] The Internet of Vehicles cloud platform is used to update the information parameters of the moving motor vehicle and the environmental image around the moving motor vehicle in real time, and input the real-time updated information parameters of the moving motor vehicle and the environmental image information into the data processing module;

[0008] The data processing module is connected to the data acquisition module, the information judgment module, the learning module and the warning execution module respectively, and the data processing module is used to identify the direction of pedestrian movement and predict the direction of motor vehicle travel, match the motor vehicle operating conditions with the accident model, and perform calculations according to the calculation method of the safe distance between people and vehicles and the actual distance under the specific model; the motor vehicle operating conditions include the type of motor vehicle driving section, the motor vehicle driving direction, the pedestrian walking direction and the relative position of people and vehicles;

[0009] The information judgment module is used to judge whether the motor vehicle operating condition matches the accident model, whether the unmatched model needs to save new data or delete data, and whether the actual distance exceeds the safety distance threshold;

[0010] The learning module receives the data retention application sent by the information judgment module, saves the corresponding accident model feature information and establishes the corresponding model and stores it in the model database. The learning module receives the data deletion application and deletes the relevant information parameters.

[0011] The warning execution module is used to execute the warning function, and the warning execution module includes the vibration and voice reminder of the pedestrian's smartphone, and the voice alarm of the motor vehicle.

[0012] The pedestrian's smartphone is installed with a WeChat applet and a server domain name is configured to enable the smartphone to connect to the Internet and synchronize information to the Internet of Vehicles cloud platform. The WeChat applet has GPS positioning function, map navigation function and early warning function; the GPS positioning function is used to obtain the pedestrian's real-time geographic location information and transmit the corresponding information to the Internet of Vehicles cloud platform; the map navigation function is used to obtain the pedestrian's walking trajectory and walking direction; the early warning function includes mobile phone vibration and voice reminder.

[0013] The method for warning a vehicle-to-human collision in a connected environment based on pedestrian intention recognition, using the vehicle-to-human collision warning system in a connected environment based on pedestrian intention recognition, is characterized by comprising the following steps, and the following steps are performed in sequence:

[0014] Step S11, in the vehicle networking environment, all the running motor vehicles collect and upload the information parameters of the running motor vehicles to the vehicle information receiving module of the vehicle networking cloud platform in real time through their respective running motor vehicle data collection modules;

[0015] A pedestrian data collection module is installed in the pedestrian's smartphone, and a corresponding WeChat applet is installed on the smartphone and a server domain name is configured so that it can connect to the network and synchronize information to the Internet of Vehicles cloud platform. The WeChat applet is opened, and the WeChat applet has GPS positioning function, map navigation function and early warning function. The GPS positioning function is used to obtain the real-time geographic location information of the pedestrian; the map navigation function is used to obtain the pedestrian's walking trajectory and auxiliary prediction of the pedestrian's walking direction;

[0016] The pedestrian data collection module sends the real-time geographic location information, walking trajectory information and predicted walking direction information of the pedestrian to the Internet of Vehicles cloud platform, and the pedestrian data collection module obtains the warning instructions sent by the motor vehicle to the Internet of Vehicles cloud platform from the Internet of Vehicles cloud platform; the Internet of Vehicles cloud platform sends the information parameters of the pedestrian and the information parameters of the motor vehicle traveling within the specified range of the pedestrian and the continuous images of the surrounding environment of the motor vehicle to the motor vehicle data collection module at fixed time intervals;

[0017] The pedestrian pre-selects the navigation function in the WeChat applet, and the pedestrian data collection module predicts the pedestrian's walking direction based on the route selected by the pedestrian;

[0018] If the pedestrian does not select the navigation function or does not walk according to the navigation route, the pedestrian data collection module constructs the movement trajectory of the pedestrian in the image based on the continuous images of the surrounding environment of the motor vehicle sent by the Internet of Vehicles cloud platform, thereby predicting the pedestrian's walking direction;

[0019] Step S12, the data processing module of the networked motor vehicle receives continuous images of the surrounding environment of the vehicle and information parameters of pedestrians within the specified range of the vehicle at a certain frequency within a certain time interval, and the data processing module extracts features according to the navigation route of the vehicle, the continuous images of the surrounding environment of the vehicle, the information parameters of the moving motor vehicle and the predicted trajectory of the pedestrian at a set time interval to obtain accident feature data and then obtain a corresponding risk value;

[0020] Matching the accident feature data of the current scene with the feature data of each collision accident model stored in the model database of the data processing module to obtain a similarity value;

[0021] If at least one of all the similarity values ​​obtained is greater than or equal to the set matching similarity threshold, the match is determined to be successful, and the collision accident model with the largest similarity value to the current scene is selected as the corresponding traffic accident model. The safe distance and the actual distance between pedestrians and vehicles are calculated based on the distance algorithm in the traffic accident model and the pedestrian information parameters obtained by the Internet of Vehicles cloud platform and the information parameters of motor vehicles traveling around pedestrians, and the calculation results are sent to the information judgment module;

[0022] If all similarity values ​​obtained are less than the set matching similarity threshold, it is determined that the match is unsuccessful and further risk assessment is performed:

[0023] If the danger value of the current scene is greater than or equal to the set danger coefficient threshold, the data processing module selects the accident model with the highest similarity value with the current scene as the corresponding traffic accident model, makes an initial judgment on whether to issue an early warning based on the information parameters of the pedestrians and vehicles in the current scene, and sends the judgment result to the information judgment module;

[0024] If the danger value of the current scene is less than the set danger coefficient threshold, no warning is issued. If a traffic accident occurs in this scene, the information judgment module sends a request to the learning module to retain the data. If no accident occurs, the information judgment module sends a request to the learning module to delete the data.

[0025] Step S13, the information judgment module receives the judgment result of the data processing module, and the judgment result is no warning;

[0026] The judgment result is the two calculation results of the safety distance and the actual distance between the pedestrian and the vehicle. If the safety distance is greater than or equal to the actual distance between the pedestrian and the vehicle, no warning is issued; if the pedestrian is in danger of colliding with the motor vehicles around him, and the safety distance is less than the actual distance between the pedestrian and the vehicle, it is determined that the pedestrian is in danger of colliding with the motor vehicles around him, and the information judgment module issues a warning instruction to the warning execution module;

[0027] If the judgment result of the information judgment module is that there is no warning and the actual distance between the person and the vehicle obtained in real time by the data processing module is less than or equal to 0, the information judgment module determines that an accident has occurred, and the information judgment module sends a signal to save the accident model to the learning module; if the judgment result of the information judgment module is that there is no warning and the actual distance between the person and the vehicle obtained in real time by the data processing module is not less than or equal to 0, the information judgment module determines that there is no traffic accident, and the information judgment module sends a data deletion signal to the learning module;

[0028] Step S14, when the learning module receives the signal from the information judgment module to save the accident model, it saves the characteristic information of the accident model and establishes a corresponding model and stores it in the model database. When the learning module receives the data deletion signal, it deletes the relevant information parameters;

[0029] Step S15, the warning execution module executes corresponding warning actions according to the warning signal sent by the information judgment module, including mobile phone vibration and voice reminder on the pedestrian side, and voice alarm on the motor vehicle side.

[0030] The information parameters of the pedestrian in step S12 include the pedestrian coordinate position information, the pedestrian walking speed and the walking trajectory information.

[0031] The information parameters of the moving motor vehicle in step S12 include the coordinate position of the motor vehicle, the driving section, the driving speed, the acceleration, the adhesion coefficient, the vehicle mass, the fastest braking time of the brake, the shortest braking distance, the steering wheel rotation angle and the steering wheel rotation angular velocity.

[0032] The traffic accident model in step S12 includes a linear motion model or a curvilinear motion model. In the linear motion model, the coordinates of the pedestrian (x1, y1) and the coordinates of the point where the motor vehicle is at the shortest distance from the pedestrian (x2, y2) are taken, and the algorithm for calculating the actual distance X between the pedestrian and the vehicle is:

[0033]

[0034] In the curve motion model, the pedestrian coordinates are taken as the pedestrian coordinates (x1, y1), the motor vehicle coordinates are taken as the origin, and the actual distance algorithm between the person and the vehicle is:

[0035]

[0036]

[0037] In the formula, s represents the actual arc distance between the person and the vehicle, and x and y represent the X-axis and Y-axis coordinates of the ellipse equation respectively.

[0038] In the linear motion model, the direction of the motor vehicle is straight forward or reverse, and the initial judgment process is specifically as follows:

[0039] If the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction in the current scene is not within the range of (-180°, -90°), the initial judgment result is no warning;

[0040] In the current scene, the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is within the range of (-180°, -90°). The pedestrian and the motor vehicle are moving towards each other, vertically, or diagonally in between. The actual distance between the pedestrian and the vehicle and the safety distance algorithm in the linear motion model are used to obtain the actual distance X between the pedestrian and the vehicle and the safety distance as the initial judgment result.

[0041] In the curve motion model, the driving direction of the motor vehicle is a curve turn or a U-turn, and the initial judgment process is specifically as follows:

[0042] In the curve motion model, the road surface is an intersection. In the current scene, the motor vehicle maintains the current acceleration and the steering wheel speed is constant. The pedestrian goes straight. The motor vehicle turns and enters the straight road for pedestrians.

[0043] If the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is not within the range of (90°, 180°), the initial judgment result is no warning;

[0044] The vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is within the range of (90°, 180°). The pedestrian and the motor vehicle are moving towards each other, vertically, or diagonally in between. The actual distance between the pedestrian and the vehicle and the safety distance algorithm in the curve motion model are used to obtain the actual distance between the pedestrian and the vehicle and the safety distance as the initial judgment result.

[0045] The specific calculation process of the safety distance D is as follows:

[0046] The force on a motor vehicle when braking on a horizontal road is:

[0047]

[0048]

[0049] Among them, Fz1 is the ground support force on the front wheels of the motor vehicle, Fz2 is the ground support force on the rear wheels of the motor vehicle, G is the weight of the motor vehicle, a is the distance from the center of mass of the motor vehicle to the center line of the front axle, b is the distance from the center of mass of the motor vehicle to the center line of the rear axle, m is the mass of the motor vehicle, and hg is the height of the center of mass of the vehicle. Indicates the braking deceleration of the car.

[0050]

[0051] In the formula, L represents the wheelbase, represents the road adhesion coefficient,

[0052] Therefore, the safety distance D is:

[0053]

[0054] In the formula, δ1 represents the time when the brake takes effect; δ2 is the coordination time between the communication delay and the brake; and δ3 is the safety margin.

[0055] Through the above design scheme, the present invention can bring the following beneficial effects:

[0056] The present invention uses the Internet of Vehicles cloud platform as a data exchange platform to collect real-time pedestrian walking parameters and motor vehicle operating parameters from pedestrians' mobile phones and motor vehicles, and predicts whether a traffic accident will occur after identifying the direction of pedestrian movement and predicting the direction of vehicle travel, thereby giving advance warnings to vehicles and pedestrians at the same time, thereby reducing the occurrence of traffic accidents caused by collisions between pedestrians and vehicles. In the case of a traffic accident that occurs without an alarm, a new accident model can be established and saved, thereby improving the entire system.

[0057] The present invention can comprehensively protect the safety of pedestrians in traffic, and can also help vulnerable groups in traffic such as the blind, providing them with the possibility of walking freely on the road, and has huge development potential.

[0058] The present invention fully realizes the effective coordination of people, vehicles and roads, realizes real-time communication, provides more optimization solutions for decision-making, reduces traffic accidents to a certain extent, improves traffic efficiency, and thus forms a safer and more efficient road traffic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The present invention is further described below with reference to the accompanying drawings and specific embodiments:

[0060] Figure 1 It is a structural block diagram of the networked environment human-vehicle collision warning system based on pedestrian intention recognition in the present invention.

[0061] Figure 2 It is a flowchart of the method for warning human-vehicle collision in a networked environment based on pedestrian intention recognition in the present invention.

[0062] Figure 3 It is a flowchart of pedestrian trajectory collection in the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical scheme of the present invention is clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the present invention is not limited to the following embodiments, and the specific implementation method can be determined according to the technical scheme of the present invention and actual conditions.

[0064] The connected environment human-vehicle collision warning system based on pedestrian intention recognition includes a data acquisition module, a data processing module, an information judgment module, a learning module and a warning execution module.

[0065] The data collection module includes a pedestrian data collection module and a moving motor vehicle data collection module; the pedestrian data collection module is installed in the pedestrian's smart phone and connected to the controller of the smart phone, the pedestrian data collection module obtains the pedestrian's real-time geographic location information, walking speed information and walking trajectory information and uploads the collected information to the Internet of Vehicles cloud platform, and the pedestrian data collection module obtains the warning instructions sent by the motor vehicle to the Internet of Vehicles cloud platform from the Internet of Vehicles cloud platform;

[0066] The running motor vehicle data acquisition module, data processing module, information judgment module and learning module are all installed on the motor vehicle. The running motor vehicle data acquisition module is used to obtain information parameters of the running motor vehicle and continuous images of the surrounding environment of the motor vehicle and upload the obtained information to the vehicle information receiving module of the Internet of Vehicles cloud platform in real time; the information parameters of the running motor vehicle include the coordinate position of the motor vehicle, the driving section, the driving speed, the acceleration, the adhesion coefficient, the vehicle mass, the fastest braking time of the brake, the shortest braking distance, the steering wheel rotation angle and the steering wheel rotation angular velocity information;

[0067] The Internet of Vehicles cloud platform is used to update the information parameters of the moving motor vehicle and the environmental image around the moving motor vehicle in real time, and input the real-time updated information parameters of the moving motor vehicle and the environmental image information into the data processing module;

[0068] The data processing module is connected to the data acquisition module, the information judgment module, the learning module and the warning execution module respectively, and the data processing module is used to identify the direction of pedestrian movement and predict the direction of motor vehicle travel, match the motor vehicle operating conditions with the accident model, and perform calculations according to the calculation method of the safe distance between people and vehicles and the actual distance under the specific model; the motor vehicle operating conditions include the type of motor vehicle driving section, the motor vehicle driving direction, the pedestrian walking direction and the relative position of people and vehicles;

[0069] The types of the motor vehicle driving road section include schools, intersections, crossroads, markets and communities, but are not limited to the above types.

[0070] The information judgment module is used to judge whether the motor vehicle operating condition matches the accident model, whether the unmatched model needs to save new data or delete data, and whether the actual distance exceeds the safety distance threshold;

[0071] The learning module receives the data retention application sent by the information judgment module, saves the corresponding accident model feature information and establishes the corresponding model and stores it in the model database. The learning module receives the data deletion application and deletes the relevant information parameters.

[0072] The warning execution module is used to execute the warning function, and the warning execution module includes the vibration and voice reminder of the pedestrian's smartphone, and the voice alarm of the motor vehicle.

[0073] The pedestrian's smartphone is installed with a WeChat applet and a server domain name is configured to enable the smartphone to connect to the Internet and synchronize information to the Internet of Vehicles cloud platform. The WeChat applet has GPS positioning function, map navigation function and early warning function; the GPS positioning function is used to obtain the pedestrian's real-time geographic location information and transmit the corresponding information to the Internet of Vehicles cloud platform; the map navigation function is used to obtain the pedestrian's walking trajectory and walking direction; the early warning function includes mobile phone vibration and voice reminder.

[0074] The connected environment human-vehicle collision warning method based on pedestrian intention recognition, using the connected environment human-vehicle collision warning system based on pedestrian intention recognition, includes the following steps, and the following steps are performed in sequence:

[0075] Step S11, in the vehicle networking environment, all the running motor vehicles collect and upload the information parameters of the running motor vehicles to the vehicle information receiving module of the vehicle networking cloud platform in real time through their respective running motor vehicle data collection modules;

[0076] A pedestrian data collection module is installed in the pedestrian's smartphone, and a corresponding WeChat applet is installed on the smartphone and a server domain name is configured so that it can connect to the network and synchronize information to the Internet of Vehicles cloud platform. The WeChat applet is opened, and the WeChat applet has GPS positioning function, map navigation function and early warning function. The GPS positioning function is used to obtain the real-time geographic location information of the pedestrian; the map navigation function is used to obtain the pedestrian's walking trajectory and auxiliary prediction of the pedestrian's walking direction;

[0077] The pedestrian data collection module sends the real-time geographic location information, walking trajectory information and predicted walking direction information of the pedestrian to the Internet of Vehicles cloud platform, and the pedestrian data collection module obtains the warning instructions sent by the motor vehicle to the Internet of Vehicles cloud platform from the Internet of Vehicles cloud platform; the Internet of Vehicles cloud platform sends the information parameters of the pedestrian and the information parameters of the motor vehicle traveling within the specified range of the pedestrian and the continuous images of the surrounding environment of the motor vehicle to the motor vehicle data collection module at fixed time intervals;

[0078] The pedestrian pre-selects the navigation function in the WeChat applet, and the pedestrian data collection module predicts the pedestrian's walking direction based on the route selected by the pedestrian;

[0079] If the pedestrian does not select the navigation function or does not walk according to the navigation route, the pedestrian data collection module constructs the movement trajectory of the pedestrian in the image based on the continuous images of the surrounding environment of the motor vehicle sent by the Internet of Vehicles cloud platform, thereby predicting the pedestrian's walking direction;

[0080] Step S12, the data processing module of the networked motor vehicle receives continuous images of the surrounding environment of the vehicle and information parameters of pedestrians within the specified range of the vehicle at a certain frequency within a certain time interval, and the data processing module extracts features according to the navigation route of the vehicle, the continuous images of the surrounding environment of the vehicle, the information parameters of the moving motor vehicle and the predicted trajectory of the pedestrian at a set time interval to obtain accident feature data and then obtain a corresponding risk value;

[0081] Matching the accident feature data of the current scene with the feature data of each collision accident model stored in the model database of the data processing module to obtain a similarity value;

[0082] If at least one of all the similarity values ​​obtained is greater than or equal to the set matching similarity threshold, the match is determined to be successful, and the collision accident model with the largest similarity value to the current scene is selected as the corresponding traffic accident model. The safe distance and the actual distance between pedestrians and vehicles are calculated based on the distance algorithm in the traffic accident model and the pedestrian information parameters obtained by the Internet of Vehicles cloud platform and the information parameters of motor vehicles traveling around pedestrians, and the calculation results are sent to the information judgment module;

[0083] If all similarity values ​​obtained are less than the set matching similarity threshold, it is determined that the match is unsuccessful and further risk assessment is performed:

[0084] If the danger value of the current scene is greater than or equal to the set danger coefficient threshold, the data processing module selects the accident model with the highest similarity value with the current scene as the corresponding traffic accident model, makes an initial judgment on whether to issue an early warning based on the information parameters of the pedestrians and vehicles in the current scene, and sends the judgment result to the information judgment module;

[0085] If the danger value of the current scene is less than the set danger coefficient threshold, no warning is issued. If a traffic accident occurs in this scene, the information judgment module sends a request to the learning module to retain the data. If no accident occurs, the information judgment module sends a request to the learning module to delete the data.

[0086] The pedestrian information parameters include pedestrian coordinate position information, pedestrian walking speed and walking trajectory information.

[0087] The information parameters of the moving motor vehicle include the coordinate position of the motor vehicle, the driving section, the driving speed, the acceleration, the adhesion coefficient, the vehicle mass, the fastest braking time of the brake, the shortest braking distance, the steering wheel rotation angle and the steering wheel rotation angular velocity information.

[0088] Step S13, the information judgment module receives the judgment result of the data processing module, and the judgment result is no warning;

[0089] The judgment result is the two calculation results of the safety distance and the actual distance between the pedestrian and the vehicle. If the safety distance is greater than or equal to the actual distance between the pedestrian and the vehicle, no warning is issued; if the pedestrian is in danger of colliding with the motor vehicles around him, and the safety distance is less than the actual distance between the pedestrian and the vehicle, it is determined that the pedestrian is in danger of colliding with the motor vehicles around him, and the information judgment module sends a warning instruction to the warning execution module;

[0090] If the judgment result of the information judgment module is that there is no warning and the actual distance between the person and the vehicle obtained in real time by the data processing module is less than or equal to 0, the information judgment module determines that an accident has occurred, and the information judgment module sends a signal to save the accident model to the learning module; if the judgment result of the information judgment module is that there is no warning and the actual distance between the person and the vehicle obtained in real time by the data processing module is not less than or equal to 0, the information judgment module determines that there is no traffic accident, and the information judgment module sends a data deletion signal to the learning module;

[0091] Step S14, when the learning module receives the signal from the information judgment module to save the accident model, it saves the characteristic information of the accident model and establishes a corresponding model and stores it in the model database. When the learning module receives the data deletion signal, it deletes the relevant information parameters.

[0092] Step S15, the warning execution module executes the corresponding warning action according to the warning signal sent by the information judgment module, including mobile phone vibration and voice reminder on the pedestrian side, such as voice reminder "Please pay attention to the vehicle coming three meters behind" and voice alarm on the motor vehicle side, such as "Please pay attention to the pedestrian three meters ahead". The alarm on the motor vehicle side can be executed directly, while the mobile phone alarm on the pedestrian side needs to be executed after the signal is forwarded through the Internet of Vehicles cloud platform.

[0093] The traffic accident model in step S12 includes a linear motion model or a curvilinear motion model. In the linear motion model, the coordinates of the pedestrian (x1, y1) and the coordinates of the point where the motor vehicle is at the shortest distance from the pedestrian (x2, y2) are taken, and the algorithm for calculating the actual distance X between the pedestrian and the vehicle is:

[0094]

[0095] In the curve motion model, the pedestrian coordinates are taken as the pedestrian coordinates (x1, y1), the motor vehicle coordinates are taken as the origin, and the actual distance algorithm between the person and the vehicle is:

[0096]

[0097]

[0098] In the formula, s represents the actual arc distance between the person and the vehicle, and x and y represent the X-axis and Y-axis coordinates of the ellipse equation respectively.

[0099] In the linear motion model, the direction of the motor vehicle is straight forward or reverse, and the initial judgment process is specifically as follows:

[0100] If the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction in the current scene is not within the range of (-180°, -90°), the initial judgment result is no warning;

[0101] In the current scene, the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is within the range of (-180°, -90°). The pedestrian and the motor vehicle are moving towards each other, vertically, or diagonally in between. The actual distance between the pedestrian and the vehicle and the safety distance algorithm in the linear motion model are used to obtain the actual distance X between the pedestrian and the vehicle and the safety distance as the initial judgment result.

[0102] In the curve motion model, the driving direction of the motor vehicle is a curve turn or a U-turn, and the initial judgment process is specifically as follows:

[0103] In the curve motion model, the road surface is an intersection. In the current scene, the motor vehicle maintains the current acceleration and the steering wheel speed is constant. The pedestrian goes straight. The motor vehicle turns and enters the straight road for pedestrians.

[0104] If the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is not within the range of (90°, 180°), the initial judgment result is no warning;

[0105] The vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is within the range of (90°, 180°). The pedestrian and the motor vehicle are moving towards each other, vertically, or diagonally in between. The actual distance between the pedestrian and the vehicle and the safety distance algorithm in the curve motion model are used to obtain the actual distance between the pedestrian and the vehicle and the safety distance as the initial judgment result.

[0106] The specific calculation process of the safety distance D is as follows:

[0107] The force on a motor vehicle when braking on a horizontal road is:

[0108]

[0109]

[0110] Among them, Fz1 is the ground support force on the front wheels of the motor vehicle, Fz2 is the ground support force on the rear wheels of the motor vehicle, G is the weight of the motor vehicle, a is the distance from the center of mass of the motor vehicle to the center line of the front axle, b is the distance from the center of mass of the motor vehicle to the center line of the rear axle, m is the mass of the motor vehicle, and hg is the height of the center of mass of the vehicle. Indicates the braking deceleration of the car.

[0111]

[0112] In the formula, L represents the wheelbase, represents the road adhesion coefficient,

[0113] Therefore, the safety distance D is:

[0114]

[0115] In the formula, δ1 represents the time for the brake to take effect, usually δ1 = 0.2s; δ2 is the coordination time between the communication delay and the brake, usually δ2 = 0.24s; δ3 is the safety margin, which is affected by the road section and vehicle speed.

[0116] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A connected environment human-vehicle collision warning system based on pedestrian intention recognition, which is characterized by: It includes data collection module, data processing module, information judgment module, learning module and early warning execution module. The data collection module includes a pedestrian data collection module and a moving motor vehicle data collection module; the pedestrian data collection module is installed in the pedestrian's smart phone and connected to the controller of the smart phone, the pedestrian data collection module obtains the pedestrian's real-time geographic location information, walking speed information and walking trajectory information and uploads the collected information to the Internet of Vehicles cloud platform, and the pedestrian data collection module obtains the warning instructions sent by the motor vehicle to the Internet of Vehicles cloud platform from the Internet of Vehicles cloud platform; The running motor vehicle data acquisition module, data processing module, information judgment module and learning module are all installed on the motor vehicle. The running motor vehicle data acquisition module is used to obtain information parameters of the running motor vehicle and continuous images of the surrounding environment of the motor vehicle and upload the obtained information to the vehicle information receiving module of the Internet of Vehicles cloud platform in real time; the information parameters of the running motor vehicle include the coordinate position of the motor vehicle, the driving section, the driving speed, the acceleration, the adhesion coefficient, the vehicle mass, the fastest braking time of the brake, the shortest braking distance, the steering wheel rotation angle and the steering wheel rotation angular velocity information; The Internet of Vehicles cloud platform is used to update the information parameters of the moving motor vehicle and the environmental image around the moving motor vehicle in real time, and input the real-time updated information parameters of the moving motor vehicle and the environmental image information into the data processing module; The data processing module is connected to the data acquisition module, the information judgment module, the learning module and the warning execution module respectively, and the data processing module is used to identify the direction of pedestrian movement and predict the direction of motor vehicle travel, match the motor vehicle operating conditions with the collision accident model, and perform calculations according to the calculation method of the safe distance between people and vehicles and the actual distance under the specific model; the motor vehicle operating conditions include the type of motor vehicle driving section, the motor vehicle driving direction, the pedestrian walking direction and the relative position of people and vehicles; The data processing module is connected to the motor vehicle, and receives continuous images of the surrounding environment of the vehicle and information parameters of pedestrians within a specified range of the vehicle at a certain frequency within a certain time interval. The data processing module extracts features according to the navigation route of the vehicle, the continuous images of the surrounding environment of the vehicle, the information parameters of the moving motor vehicle and the predicted trajectory of the pedestrian at a set time interval to obtain accident feature data and then obtain a corresponding risk value; The characteristic data of each collision accident model stored in the model database of the data processing module is matched with the accident characteristic data of the current scene to obtain a similarity value; if at least one of all the similarity values ​​obtained is greater than or equal to the set matching similarity threshold, it is determined that the match is successful, and the collision accident model with the largest similarity value of the current scene is used as the corresponding traffic accident model; The information judgment module is used to judge whether the motor vehicle operating condition matches the collision accident model, whether the unmatched model needs to save new data or delete data, and whether the actual distance exceeds the safety distance threshold; The learning module receives a data retention request from the information judgment module and saves the corresponding collision accident model feature information and establishes a corresponding model and stores it in the model database. The learning module receives a data deletion request and deletes the relevant information parameters. The warning execution module is used to execute the warning function, and the warning execution module includes the vibration and voice reminder of the pedestrian's smartphone, and the voice alarm of the motor vehicle.

2. The connected environment human-vehicle collision warning system based on pedestrian intention recognition according to claim 1 is characterized by: The pedestrian's smartphone is installed with a WeChat applet and a server domain name is configured to enable the smartphone to connect to the Internet and synchronize information to the Internet of Vehicles cloud platform. The WeChat applet has GPS positioning function, map navigation function and early warning function; the GPS positioning function is used to obtain the pedestrian's real-time geographic location information and transmit the corresponding information to the Internet of Vehicles cloud platform; the map navigation function is used to obtain the pedestrian's walking trajectory and walking direction; the early warning function includes mobile phone vibration and voice reminder.

3. A connected environment human-vehicle collision warning method based on pedestrian intention recognition, using the connected environment human-vehicle collision warning system based on pedestrian intention recognition as described in claim 1, characterized in that: The method comprises the following steps, and the following steps are performed in sequence: Step S11, in the vehicle networking environment, all the running motor vehicles collect and upload the information parameters of the running motor vehicles to the vehicle information receiving module of the vehicle networking cloud platform in real time through their respective running motor vehicle data collection modules; A pedestrian data collection module is installed in the pedestrian's smartphone, and a corresponding WeChat applet is installed on the smartphone and a server domain name is configured so that it can connect to the network and synchronize information to the Internet of Vehicles cloud platform. The WeChat applet is opened, and the WeChat applet has GPS positioning function, map navigation function and early warning function. The GPS positioning function is used to obtain the real-time geographic location information of the pedestrian; the map navigation function is used to obtain the pedestrian's walking trajectory and auxiliary prediction of the pedestrian's walking direction; The pedestrian data collection module sends the real-time geographic location information, walking trajectory information and predicted walking direction information of the pedestrian to the Internet of Vehicles cloud platform, and the pedestrian data collection module obtains the warning instructions sent by the motor vehicle to the Internet of Vehicles cloud platform from the Internet of Vehicles cloud platform; the Internet of Vehicles cloud platform sends the information parameters of the pedestrian and the information parameters of the motor vehicle traveling within the specified range of the pedestrian and the continuous images of the surrounding environment of the motor vehicle to the motor vehicle data collection module at fixed time intervals; The pedestrian pre-selects the navigation function in the WeChat applet, and the pedestrian data collection module predicts the pedestrian's walking direction based on the route selected by the pedestrian; If the pedestrian does not select the navigation function or does not walk according to the navigation route, the pedestrian data collection module constructs the movement trajectory of the pedestrian in the image based on the continuous images of the surrounding environment of the motor vehicle sent by the Internet of Vehicles cloud platform, thereby predicting the pedestrian's walking direction; Step S12, the data processing module of the networked motor vehicle receives continuous images of the surrounding environment of the vehicle and information parameters of pedestrians within the specified range of the vehicle at a certain frequency within a certain time interval, and the data processing module extracts features according to the navigation route of the vehicle, the continuous images of the surrounding environment of the vehicle, the information parameters of the moving motor vehicle and the predicted trajectory of the pedestrian at the set time interval to obtain accident feature data and then obtain the corresponding risk value; Matching the accident feature data of the current scene with the feature data of each collision accident model stored in the model database of the data processing module to obtain a similarity value; If at least one of all the similarity values ​​obtained is greater than or equal to the set matching similarity threshold, the match is determined to be successful, and the collision accident model with the largest similarity value to the current scene is selected as the corresponding traffic accident model. The safe distance and the actual distance between pedestrians and vehicles are calculated based on the distance algorithm in the traffic accident model and the pedestrian information parameters obtained by the Internet of Vehicles cloud platform and the information parameters of motor vehicles traveling around pedestrians, and the calculation results are sent to the information judgment module; If all similarity values ​​obtained are less than the set matching similarity threshold, it is determined that the match is unsuccessful and further risk assessment is performed: If the danger value of the current scene is greater than or equal to the set danger coefficient threshold, the data processing module selects the collision accident model with the highest similarity value with the current scene as the corresponding traffic accident model, makes an initial judgment on whether to issue an early warning based on the information parameters of the pedestrians and vehicles in the current scene, and sends the judgment result to the information judgment module; If the danger value of the current scene is less than the set danger coefficient threshold, no warning is issued. If a traffic accident occurs in this scene, the information judgment module sends a request to the learning module to retain the data. If no accident occurs, the information judgment module sends a request to the learning module to delete the data. Step S13, the information judgment module receives the judgment result of the data processing module, and the judgment result is no warning; The judgment result is the two calculation results of the safety distance and the actual distance between the pedestrian and the vehicle. If the safety distance is greater than or equal to the actual distance between the pedestrian and the vehicle, no warning is issued; if the pedestrian is in danger of colliding with the motor vehicles around him, and the safety distance is less than the actual distance between the pedestrian and the vehicle, it is determined that the pedestrian is in danger of colliding with the motor vehicles around him, and the information judgment module issues a warning instruction to the warning execution module; If the judgment result of the information judgment module is that there is no warning and the actual distance between the person and the vehicle obtained in real time by the data processing module is less than or equal to 0, the information judgment module determines that an accident has occurred, and the information judgment module sends a signal to save the collision accident model to the learning module; if the judgment result of the information judgment module is that there is no warning and the actual distance between the person and the vehicle obtained in real time by the data processing module is not less than or equal to 0, the information judgment module determines that there is no traffic accident, and the information judgment module sends a data deletion signal to the learning module; Step S14, when the learning module receives the signal from the information judgment module to save the collision accident model, it saves the characteristic information of the collision accident model and establishes a corresponding model and stores it in the model database. When the learning module receives the data deletion signal, it deletes the relevant information parameters. Step S15, the warning execution module executes corresponding warning actions according to the warning signal sent by the information judgment module, including mobile phone vibration and voice reminder on the pedestrian side, and voice alarm on the motor vehicle side.

4. The method for warning human-vehicle collision in a connected environment based on pedestrian intention recognition according to claim 3 is characterized in that: The information parameters of the pedestrian in step S12 include the pedestrian coordinate position information, the pedestrian walking speed and the walking trajectory information.

5. The method for warning human-vehicle collision in a connected environment based on pedestrian intention recognition according to claim 3 is characterized by: The information parameters of the moving motor vehicle in step S12 include the coordinate position of the motor vehicle, the driving section, the driving speed, the acceleration, the adhesion coefficient, the vehicle mass, the fastest braking time of the brake, the shortest braking distance, the steering wheel rotation angle and the steering wheel rotation angular velocity.

6. The method for warning human-vehicle collision in a connected environment based on pedestrian intention recognition according to claim 3 is characterized by: The traffic accident model in step S12 includes a linear motion model or a curvilinear motion model. In the linear motion model, the coordinates of the pedestrian (x1, y1) and the coordinates of the point where the motor vehicle is at the shortest distance from the pedestrian (x2, y2) are taken, and the algorithm for calculating the actual distance X between the pedestrian and the vehicle is: In the curve motion model, the pedestrian coordinates are taken as the pedestrian coordinates (x1, y1), the motor vehicle coordinates are taken as the origin, and the actual distance algorithm between the person and the vehicle is: In the formula, s represents the actual arc distance between the person and the vehicle, and x and y represent the X-axis and Y-axis coordinates of the ellipse equation respectively.

7. The method for warning human-vehicle collision in a connected environment based on pedestrian intention recognition according to claim 6 is characterized by: In the linear motion model, the direction of the motor vehicle is straight forward or reverse, and the initial judgment process is specifically as follows: If the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction in the current scene is not within the range of (-180°, -90°), the initial judgment result is no warning; In the current scene, the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is within the range of (-180°, -90°). The pedestrian and the motor vehicle are moving towards each other, vertically, or diagonally in between. The actual distance between the pedestrian and the vehicle and the safety distance algorithm in the linear motion model are used to obtain the actual distance X between the pedestrian and the vehicle and the safety distance as the initial judgment result.

8. The method for warning human-vehicle collision in a connected environment based on pedestrian intention recognition according to claim 6 is characterized by: In the curve motion model, the driving direction of the motor vehicle is a curve turn or a U-turn, and the initial judgment process is specifically as follows: In the curve motion model, the road surface is an intersection. In the current scene, the motor vehicle maintains the current acceleration and the steering wheel speed is constant. The pedestrian goes straight. The motor vehicle turns and enters the straight road for pedestrians. If the vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is not within the range of (90°, 180°), the initial judgment result is no warning; The vector angle between the pedestrian's movement direction and the motor vehicle's movement direction is within the range of (90°, 180°). The pedestrian and the motor vehicle are moving towards each other, vertically, or diagonally in between. The actual distance between the pedestrian and the vehicle and the safety distance algorithm in the curve motion model are used to obtain the actual distance between the pedestrian and the vehicle and the safety distance as the initial judgment result.

9. The method for warning human-vehicle collision in a connected environment based on pedestrian intention recognition according to claim 3 is characterized by: The specific calculation process of the safety distance D is as follows: The force on a motor vehicle when braking on a horizontal road is: Among them, Fz1 is the ground support force on the front wheels of the motor vehicle, Fz2 is the ground support force on the rear wheels of the motor vehicle, G is the weight of the motor vehicle, a is the distance from the center of mass of the motor vehicle to the center line of the front axle, b is the distance from the center of mass of the motor vehicle to the center line of the rear axle, m is the mass of the motor vehicle, and hg is the height of the center of mass of the vehicle. Indicates the braking deceleration of the car. In the formula, L represents the wheelbase, represents the road adhesion coefficient, Therefore, the safety distance D is: In the formula, δ1 represents the time when the brake takes effect; δ2 is the coordination time between the communication delay and the brake; and δ3 is the safety margin.

Images