Vehicle accident detection and automatic reporting system capable of analyzing accident type and accident severity using ai deep learning algorithm

Abstract

A vehicle accident detection and automatic reporting system is disclosed. The vehicle accident detection and automatic reporting system comprises: a vehicle terminal connected to an OBD connector mounted in a vehicle; and a user terminal owned by an occupant of the vehicle and configured to communicate with the vehicle terminal. The vehicle terminal includes: a driving information collection unit that collects driving data of the vehicle from the OBD connector; an inertial measurement sensor that measures acceleration data of the vehicle terminal; a processor that determines a vehicle collision on the basis of the acceleration data measured by the inertial measurement sensor, calculates a collision time point, and extracts pre-collision data acquired for a preset time period before the collision time point and post-collision data acquired for a preset time period after the collision time point; and a communication module that communicates with the user terminal and transmits the pre-collision data and the post-collision data to the user terminal. The user terminal includes: an accident determination unit that determines whether an accident has occurred by processing the pre-collision data and the post-collision data as input data using a pre-trained AI deep learning algorithm and comparing learning results with pre-stored accident pattern information; and an accident type and severity analysis unit that, using the pre-trained AI deep learning algorithm, divides the pre-collision data and the post-collision data into preset time units on the basis of the collision time point to generate a plurality of segments, compresses each of the plurality of segments to different sizes to generate segment compression signals, extracts feature points from each of the segment compression signals, and compares the extracted feature points with pre-obtained feature points corresponding to accident types and severities to determine an accident type and accident severity.

Description

Vehicle accident detection and automatic reporting system capable of analyzing accident types and severity using AI deep learning algorithms

[0001] The present invention relates to a vehicle accident detection and automatic reporting system, and more specifically, provides a vehicle accident detection and automatic reporting system that can analyze accident types and accident severity using an AI deep learning algorithm, automatically generate an accident occurrence report, and transmit the generated accident occurrence report to an emergency rescue agency.

[0002] Automobiles have become an indispensable and important means of transportation for modern people, and the ownership rate of automobiles is increasing day by day to the point where it is difficult to find a household without a car these days.

[0003] The continuous increase in automobiles within a limited national territory and complex road networks and transportation systems has inevitably led to an increase in the automobile accident rate. As a result, many lives are lost in automobile accidents each year, and the increase in automobiles and the accidents associated with them have emerged as another problem in modern society.

[0004] While there are various methods to reduce casualties from automobile accidents, such as expanding road networks and establishing systematic traffic order, these also have limitations. In the event of a minor accident, drivers currently use their smartphones to call emergency vehicles, repair shops, or the police, whereas in serious accidents where the driver is unable to use their smartphone, other drivers nearby call for an emergency vehicle.

[0005] However, such accident handling is not very effective in secluded areas or on highways, and therefore, when an emergency situation arises that results in casualties, prompt handling of the accident is impossible, often leading to the loss of the driver's precious life. Given that this results in national losses, it is necessary to have a plan for the prompt handling of accidents.

[0006] In addition, when determining the cause of a vehicle accident, it is possible to identify the cause to some extent through witnesses or CCTV footage such as dashcams; however, if the accident was caused by a vehicle malfunction, it is very difficult to determine the cause. If the cause of the malfunction can be identified in the event of an accident caused by a vehicle malfunction, it would be possible to apply this accident cause information to the vehicle in the future to improve quality and provide a safe vehicle.

[0007] Recent vehicles are equipped with E-Call (emergency call) systems for automatic emergency calls in the event of a vehicle accident. An E-Call system refers to a system that transmits accident information data to an accident reporting center upon the occurrence of an accident, establishes a voice channel between the accident victim and the emergency call center to assess the situation, and enables rapid follow-up actions. In other words, the E-Call system sends vehicle information, such as location, speed, and direction, to nearby PSAPs (Public Answering Points) when an accident occurs, allowing for swift and efficient emergency response.

[0008] The present invention provides an accident severity analysis and automatic vehicle accident reporting system that, when a vehicle accident occurs, notifies a vehicle accident notification server from a user terminal of the fact of the accident, and the vehicle accident notification server notifies an accident response agency of the fact of the accident so that a rapid response to the accident can be carried out.

[0009] In addition, the present invention provides an accident severity analysis and automatic vehicle accident reporting system in which, when an accident occurs, the fact of the accident is automatically notified to a vehicle accident notification server after a preset waiting time for data transmission has elapsed, thereby enabling accident response.

[0010] In addition, the present invention provides an accident severity analysis and automatic vehicle accident reporting system capable of automatically generating a vehicle accident report when a vehicle accident occurs and transmitting the generated accident report to an emergency rescue agency.

[0011] In addition, the present invention provides an accident severity analysis and automatic vehicle accident reporting system capable of determining the type and severity of an accident by using a pre-trained AI deep learning algorithm to extract feature points from pre-vehicle collision data and post-vehicle collision data based on the time of collision, and comparing these with feature points of accident type and severity obtained in advance.

[0012] In addition, the present invention provides an accident severity analysis and automatic vehicle accident reporting system that performs question-and-answer with a passenger based on LLM through a user terminal on a vehicle accident notification server and identifies the severity of the passenger's accident based on the question-and-answer results.

[0013] In addition, the present invention provides a vehicle accident detection and automatic reporting system capable of analyzing vehicle accident types and accident severity using sound data around the vehicle.

[0014] In addition, the present invention provides a vehicle accident detection and automatic reporting system capable of rapidly and efficiently analyzing vehicle accident types and accident severity.

[0015] In addition, the present invention provides a vehicle accident detection and automatic reporting system capable of determining a light collision or a collision with a soft object.

[0016] A vehicle accident detection and automatic reporting system according to an embodiment of the present invention may include: a vehicle terminal connected to an OBD connector mounted on a vehicle; and a user terminal owned by a user riding in the vehicle, which is capable of communicating with the vehicle terminal, wherein the vehicle terminal comprises: a driving information collection unit that collects driving data of the vehicle from the OBD connector; an inertial measurement sensor that measures acceleration data of the vehicle terminal; a processor that determines a vehicle collision using acceleration data measured by the inertial measurement sensor, calculates a collision time, and extracts vehicle collision pre-collision data acquired during a preset time before the collision time and vehicle collision post-collision data acquired during a preset time after the collision time; and a communication module capable of communicating with the vehicle terminal and transmitting the vehicle collision pre-collision data and the vehicle collision post-collision data to the user terminal.

[0017] In addition, the vehicle terminal further includes a GPS module that receives location data of the vehicle terminal, and the communication module can transmit the location data at the time of the collision to the user terminal.

[0018] In addition, the inertial measurement sensor can switch the processor to an operating mode when acceleration data indicating a vehicle collision is detected.

[0019] Additionally, the vehicle terminal further includes a microphone for receiving ambient sound signals within the vehicle, and the vehicle collision pre-data may include ambient sound data received by the microphone for a preset time prior to the collision point, and the vehicle collision post-data may include vehicle collision post-data received by the microphone for a preset time after the collision point.

[0020] Additionally, the user terminal may include: an accident determination unit that uses a pre-learned AI deep learning algorithm to learn the vehicle collision data and the vehicle collision data as input data, and determines whether an accident has occurred by comparing the learning result with pre-stored accident pattern information; and an accident type and severity analysis unit that uses a pre-learned AI deep learning algorithm to divide the vehicle collision data and the vehicle collision data into pre-set time units based on the collision time point to generate a plurality of segments, compresses each of the segments to different sizes to generate segment compression signals, extracts feature points from each of the segment compression signals, and determines the accident type and severity by comparing the feature points with pre-secured feature points of accident type and severity.

[0021] In addition, the accident type and severity analysis unit can generate segments such that the number of segments generated from the data after the vehicle collision is greater than the number of segments generated from the data before the vehicle collision.

[0022] In addition, the accident type and severity analysis unit can divide the segments such that the time length of the segments generated from the vehicle collision data and the vehicle collision data becomes shorter as it approaches the collision time, and the time length of the segments generated from the vehicle collision data and the vehicle collision data becomes longer as it moves away from the collision time.

[0023] In addition, the segment compression signal includes a driving speed segment compression signal and an acceleration segment compression signal, and the accident type and severity analysis unit can determine that it is a light collision or a collision with a soft object if the difference between the feature points extracted from the driving speed segment compression signal and the feature points extracted from the acceleration segment compression signal exceeds a threshold value.

[0024] In addition, the above segment compression signal includes a driving speed segment compression signal, an acceleration segment compression signal, and an ambient sound segment compression signal, and the above accident type and severity analysis unit analyzes the accident type and severity using [Equation 1], a vehicle accident detection and automatic reporting system.

[0025] [Formula 1]

[0026] L = αL1 + βL2 + γL3

[0027] Here,

[0028] α is a weight for feature points extracted from the driving speed segment compressed signal, and

[0029] β is a weight for feature points extracted from the acceleration segment compressed signal, and

[0030] γ is a weight for feature points extracted from the ambient sound segment compressed signal, and

[0031] L1 is the difference value between the feature points extracted from the driving speed segment compressed signal and the previously secured feature points of the driving speed, and

[0032] L2 is the difference value between the feature points extracted from the acceleration segment compressed signal and the previously secured acceleration feature points, and

[0033] L3 is the difference between the feature points extracted from the ambient sound segment compressed signal and the previously secured feature points of the ambient sound.

[0034] A vehicle accident detection and automatic reporting method according to an embodiment of the present invention comprises: a vehicle data collection step of collecting driving data of the vehicle from an OBD connector mounted on the vehicle and measuring acceleration data of a vehicle terminal connected to the OBD connector; a vehicle collision determination step of determining a vehicle collision using the acceleration data, calculating a collision time, and extracting vehicle pre-collision data acquired during a preset time before the collision time and vehicle post-collision data acquired during a preset time after the collision time; a vehicle data transmission step of transmitting the vehicle pre-collision data and the vehicle post-collision data to a user terminal of a user riding in the vehicle; and an accident determination step of learning using a pre-trained AI deep learning algorithm with the vehicle pre-collision data and the vehicle post-collision data as input data, and determining whether an accident has occurred by comparing the learning result with pre-stored accident pattern information. The method further includes an accident type and severity analysis step in which, using a previously trained AI deep learning algorithm based on the collision time, the vehicle collision pre-data and vehicle collision post-data are divided into pre-set time units to generate a plurality of segments, each of the segments is compressed to a different size to generate a segment compression signal, feature points are extracted from each of the segment compression signals, and the accident type and accident severity are determined by comparing the feature points with pre-secured accident type and severity feature points.

[0035] In addition, the accident type and severity analysis step may generate segments such that the number of segments generated from the data after the vehicle collision is greater than the number of segments generated from the data before the vehicle collision.

[0036] In addition, the accident type and severity analysis step may divide the segments such that the time length of the segments generated from the vehicle collision data and the vehicle collision data becomes shorter as it approaches the collision time, and the time length of the segments generated from the vehicle collision data and the vehicle collision data becomes longer as it moves away from the collision time.

[0037] Additionally, the segment compression signal includes a driving speed segment compression signal and an acceleration segment compression signal, and the accident type and severity determination step can determine a light collision or a collision with a soft object if the difference between the feature points extracted from the driving speed segment compression signal and the feature points extracted from the acceleration segment compression signal exceeds a threshold value.

[0038] Additionally, the method further includes an accident occurrence report generation step for generating an accident occurrence report; and an accident notification step for transmitting the accident occurrence report to an accident response agency, wherein the accident occurrence report may include information on the type of accident and the severity of the accident.

[0039] In addition, the above segment compression signal includes a driving speed segment compression signal, an acceleration segment compression signal, and an ambient sound segment compression signal, and the above accident type and severity analysis unit can analyze the accident type and severity using [Equation 1].

[0040] [Formula 1]

[0041] L = αL1 + βL2 + γL3

[0042] Here,

[0043] α is a weight for feature points extracted from the driving speed segment compressed signal, and

[0044] β is a weight for feature points extracted from the acceleration segment compressed signal, and

[0045] γ is a weight for feature points extracted from the ambient sound segment compressed signal, and

[0046] L1 is the difference value between the feature points extracted from the driving speed segment compressed signal and the previously secured feature points of the driving speed, and

[0047] L2 is the difference value between the feature points extracted from the acceleration segment compressed signal and the previously secured acceleration feature points, and

[0048] L3 is the difference between the feature points extracted from the ambient sound segment compressed signal and the previously secured feature points of the ambient sound.

[0049] According to the present invention, when a vehicle accident occurs, the fact of the accident is notified from the user terminal to the vehicle accident notification server, and the vehicle accident notification server then notifies the accident response agency of the fact of the accident so that an accident response can be carried out quickly.

[0050] In addition, according to the present invention, when an accident occurs, if the pre-set transmission waiting period elapses while the vehicle is in a data transmission waiting state, the fact of the accident is automatically notified to the vehicle accident notification server so that an accident response can be carried out.

[0051] In addition, according to the present invention, when a vehicle accident occurs, a vehicle accident report is automatically generated and transmitted to an emergency rescue agency, so the emergency rescue agency can quickly identify the type of vehicle accident, the location of the vehicle accident, information on the vehicle involved in the accident, driver information, etc.

[0052] In addition, according to the present invention, a plurality of segments are generated by dividing the vehicle collision data and the vehicle collision data into predetermined time intervals based on the collision time using a previously learned AI deep learning algorithm, and each of the segments is compressed to a different size to generate a segment compression signal, and feature points are extracted from each of the segment compression signals, and the accident type and severity can be determined by comparing the feature points with the previously secured feature points of accident type and severity.

[0053] In addition, according to the present invention, when a vehicle accident occurs, an accident notification server performs a question-and-answer session with the occupant based on LLM through a user terminal, identifies the severity of the accident of the occupant based on the result of the question-and-answer session, and transmits accident notification data and an accident occurrence report to an accident response agency according to the severity of the accident.

[0054] Furthermore, according to the present invention, high-frequency acoustic components are extracted from ambient sound data before and after a vehicle collision, and by compressing and comparing feature points, the type and severity of the accident can be rapidly determined with a small amount of data processing. Additionally, since high-frequency acoustic components from which feature points can be effectively extracted are processed, accurate determination of the type and severity of the accident is possible.

[0055] In addition, according to the present invention, a vehicle terminal detects the time of a vehicle collision, transmits pre-collision data and post-collision data for a preset time period based on the time of the vehicle collision to a user terminal, and analyzes them, thereby enabling rapid and efficient analysis of the vehicle accident type and accident severity.

[0056] In addition, according to the present invention, by comparing the change in the actual driving speed of the vehicle with the change in acceleration measured by a vehicle terminal mounted on the vehicle, it is possible to determine whether the collision is a light collision or a collision with a soft object.

[0057] FIG. 1 is a drawing showing a vehicle accident detection and automatic reporting system according to an embodiment of the present invention.

[0058] FIG. 2 is a drawing showing the detailed configuration of a vehicle terminal according to an embodiment of the present invention.

[0059] Figure 3 is a diagram showing the detailed configuration of the sensor module for the terminal of Figure 2.

[0060] FIG. 4 is a diagram showing the detailed configuration of a user terminal according to an embodiment of the present invention.

[0061] FIG. 5 is a diagram showing the detailed configuration of the control unit of FIG. 4 of the present invention.

[0062] FIG. 6 is a diagram showing vehicle collision pre-collision data and vehicle collision post-collision data divided into a plurality of segments based on the collision time according to various embodiments of the present invention.

[0063] FIG. 7 is a drawing showing an accident occurrence report generated according to an embodiment of the present invention.

[0064] FIG. 8 is a drawing showing a vehicle accident notification server according to an embodiment of the present invention.

[0065] FIG. 9 is a drawing showing the detailed configuration of a central processing unit according to an embodiment of the present invention.

[0066] FIG. 10 is a flowchart illustrating a method for detecting a vehicle accident and automatically reporting it using a vehicle terminal and a user terminal according to the embodiments of FIG. 1 to 9.

[0067] A vehicle accident detection and automatic reporting system according to an embodiment of the present invention may include: a vehicle terminal connected to an OBD connector mounted on a vehicle; and a user terminal owned by a user riding in the vehicle, which is capable of communicating with the vehicle terminal, wherein the vehicle terminal comprises: a driving information collection unit that collects driving data of the vehicle from the OBD connector; an inertial measurement sensor that measures acceleration data of the vehicle terminal; a processor that determines a vehicle collision using acceleration data measured by the inertial measurement sensor, calculates a collision time, and extracts vehicle collision pre-collision data acquired during a preset time before the collision time and vehicle collision post-collision data acquired during a preset time after the collision time; and a communication module capable of communicating with the vehicle terminal and transmitting the vehicle collision pre-collision data and the vehicle collision post-collision data to the user terminal.

[0068] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. However, the technical concept of the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided to ensure that the disclosed content is thorough and complete and to ensure that the concept of the present invention is sufficiently conveyed to those skilled in the art.

[0069] In this specification, when a component is described as being on another component, it means that it may be formed directly on the other component or that a third component may be interposed between them. Additionally, in the drawings, the thicknesses of the films and regions are exaggerated for the effective description of the technical content.

[0070] Additionally, although terms such as first, second, third, etc., have been used to describe various components in the various embodiments of this specification, these components should not be limited by such terms. These terms are used merely to distinguish one component from another. Accordingly, what is referred to as the first component in one embodiment may be referred to as the second component in another embodiment. Each embodiment described and illustrated herein also includes its complementary embodiment. Furthermore, in this specification, "and / or" is used to mean including at least one of the components listed before and after it.

[0071] In the specification, singular expressions include plural expressions unless the context clearly indicates otherwise. Furthermore, terms such as "include" or "have" are intended to specify the existence of the features, numbers, steps, components, or combinations thereof described in the specification, and should not be understood as excluding the existence or addition of one or more other features, numbers, steps, components, or combinations thereof. Additionally, in this specification, "connection" is used to include both indirectly connecting multiple components and directly connecting them.

[0072] Furthermore, in describing the present invention below, if it is determined that a detailed description of related known functions or configurations could unnecessarily obscure the essence of the invention, such detailed description will be omitted.

[0073]

[0074] FIG. 1 is a drawing showing an accident severity analysis and automatic vehicle accident reporting system according to an embodiment of the present invention, FIG. 2 is a drawing showing a detailed configuration of a vehicle terminal according to an embodiment of the present invention, FIG. 3 is a drawing showing a detailed configuration of a sensor module for the terminal of FIG. 2, FIG. 4 is a drawing showing a detailed configuration of a user terminal according to an embodiment of the present invention, and FIG. 5 is a drawing showing a detailed configuration of the control unit of FIG. 4 of the present invention.

[0075] Referring to FIGS. 1 to 5, when an accident occurs in a vehicle in which a user is riding, the accident severity analysis and automatic vehicle accident reporting system (10) determines the fact that a vehicle accident has occurred at the user terminal (200) using information measured at the vehicle terminal (100) and the user terminal (200), and transmits the fact that a vehicle accident has occurred to the vehicle accident notification server (300). The vehicle accident notification server (300) provides a vehicle accident notification service that transmits the fact of the vehicle accident to pre-set accident response agencies, such as an insurance company, an emergency rescue agency such as 119, a vehicle towing company, and other emergency contacts set by the user, so that processing in response to the accident can be carried out.

[0076] The accident severity analysis and automatic vehicle accident reporting system (10) includes a vehicle terminal (100), a user terminal (200), and a vehicle accident notification server (300).

[0077] A vehicle terminal (100) is mounted on a vehicle and collects driving information and acceleration information of the vehicle. The vehicle terminal (100) collects the electrical / electronic operating status of the vehicle. Specifically, the vehicle terminal (100) collects the vehicle's driving speed, driving location information, driving distance information, RPM, brake signal information, accelerator pedal signal information, vehicle interior temperature information, and vehicle exterior temperature information. The vehicle terminal (100) is capable of wireless communication with a user terminal (200). According to an embodiment, the vehicle terminal (100) can communicate with the user terminal (200) via Bluetooth communication. According to an embodiment, an OBD (On-Board Diagnostics) can be used for the vehicle terminal (100).

[0078] The user terminal (200) is a terminal owned by a user who is in the vehicle, and is located inside the vehicle while the vehicle is in motion. The user terminal (200) may be carried by the user while boarding the vehicle or may be mounted inside the vehicle. The user terminal (200) may be various types of terminals or electronic equipment capable of wired or wireless data communication, such as a mobile phone, smartphone, or tablet PC. The user terminal (200) may be connected via wireless communication with the vehicle terminal (100) and the vehicle accident notification server (200).

[0079] The vehicle accident notification server (300) is connected to the user terminal (200) and the servers of the accident response agency via a network, and has a connection structure that enables information exchange between each node.

[0080] A vehicle terminal (100) includes a housing (110), a vehicle data collection unit (120), a sensor module for the terminal (130), a memory (140), a processor (150), a communication module (160), and an auxiliary battery (170).

[0081] The housing (110) is provided in a predetermined shape and has a terminal capable of electrical connection with an OBD connector mounted on a vehicle. The vehicle data collection unit (120), the sensor module for the terminal (130), the memory (140), the processor (150), the communication module (160), and the auxiliary battery (170) are provided within the housing (110).

[0082] The vehicle data collection unit (120) collects driving data of the vehicle. The vehicle data collection unit (120) collects the electrical / electronic operating status of the vehicle. The vehicle data collection unit (120) collects the vehicle's driving speed, driving location data, driving distance data, RPM, brake signal data, accelerator pedal signal data, vehicle interior temperature data, and vehicle exterior temperature data. The collected data is stored in memory (140).

[0083] The sensor module (130) for the terminal detects vehicle movement data and vehicle location data. The sensor module (130) for the terminal includes an inertial measurement sensor (IMU, 131) and a GPS module (133).

[0084] The inertial measurement sensor (131) may include a multi-axis accelerometer, for example, a two-axis or three-axis accelerometer. The inertial measurement sensor (131) may further include a gyroscope, which can be used to determine the direction of the vehicle terminal and / or the relative direction to the accelerometer data. This allows the collision detection function of the vehicle terminal (100) to be corrected.

[0085] The vehicle terminal (100) can measure changes in motion along the x-axis, y-axis, and z-axis of the vehicle. The vehicle terminal (100) can calibrate the vehicle terminal (100) by evaluating the direction of the vehicle terminal (100) with respect to the direction system of the vehicle equipped with the vehicle terminal (100). To determine whether a collision event has occurred, the inertial measurement sensor (131) can continuously collect and monitor changes in acceleration indicating a vehicle collision. The inertial measurement sensor (131) may include means for processing acceleration data. When acceleration data indicating a vehicle collision is detected by the inertial measurement sensor (131), the inertial measurement sensor (131) can wake up the processor (150) and switch to an operating mode.

[0086] Additionally, the sensor module (130) for the terminal may further include a microphone (133). The microphone (133) can receive ambient sounds (audible and / or inaudible frequency ranges) inside the vehicle.

[0087] The GPS module (132) receives radio signals from orbiting GPS satellites in the network and processes them through an integrated antenna. The GPS module (132) receives signals from at least three satellites in the GPS network and can determine accurate location data and movement data of the vehicle terminal (100) through an integrated microprocessor. The location data and movement data are then provided to the processor (150). The GPS module (132) can be configured to generate location data at desired intervals. According to an embodiment, the GPS module (132) can generate location data of the vehicle terminal (100) every 10 seconds, every 30 seconds, or every minute.

[0088] The memory (140) stores data collected from the vehicle data collection unit (120), data measured from the terminal sensor module (130), and various software. The memory (140) may be a computer-readable storage medium provided to the processor (150). For example, the memory (140) may be a flash memory. The memory (140) also acts as a buffer memory, allowing the processor (150) to continuously read and overwrite non-collision acceleration related data.

[0089] The processor (150) executes an algorithm stored in memory (140) and runs software. Additionally, the processor (150) may be any type of computer, controller, microcontroller, circuit, chipset, microprocessor, processor system, or computer system capable of loading and executing other types of computer programs.

[0090] According to an embodiment, the processor (150) determines a vehicle collision by analyzing data measured by an inertial measurement sensor (131) through the execution of an accident detection algorithm. The processor (150) determines that a collision has occurred if acceleration data in at least one axis of the data measured by the inertial measurement sensor (131) exceeds an acceleration threshold, or if the value measured by the gyroscope exceeds a threshold.

[0091] When the processor (150) determines that a vehicle collision has occurred, it calculates the collision time and calculates the vehicle location data at the collision time from the location data received by the GPS module (132). Here, the collision time is the time when the measurement value measured by the inertial measurement sensor (131) or gyroscope begins to change due to the vehicle collision.

[0092] When the processor (150) determines that a vehicle collision has occurred, it extracts vehicle collision pre-data acquired during a preset time before the collision point and vehicle collision post-data acquired during a preset time after the collision point.

[0093] The vehicle collision pre-data includes driving data of the vehicle before the collision collected by the vehicle data collection unit (120), acceleration data of the vehicle terminal (100) before the collision measured by the inertial measurement sensor (131), and ambient sound data before the collision measured by the microphone (133).

[0094] The vehicle collision post-data includes driving data of the vehicle after the collision time collected by the vehicle data collection unit (120), acceleration data of the vehicle terminal (100) after the collision time measured by the inertial measurement sensor (131), and ambient sound data after the collision time measured by the microphone (133).

[0095] The communication module (160) is wirelessly connected to the user terminal (200) and transmits data stored in the memory (140) and / or data processed by the processor (150) to the user terminal (200). According to an embodiment, the communication module (160) transmits vehicle collision pre-collision data, vehicle collision post-collision data, and vehicle location information at the time of collision output from the processor (150) to the user terminal (200). The communication module (160) can communicate with the user terminal (200) via Bluetooth communication.

[0096] The auxiliary battery (170) supplies power to the terminal sensor module (130), memory (140), processor (150), and communication module (160) when the power supply from the vehicle is cut off. The auxiliary battery (170) receives power from the vehicle and charges it while the vehicle terminal (100) is mounted in the vehicle. When the vehicle is operating normally, power is supplied from the vehicle, and as a result, the terminal sensor module (130), memory (140), processor (150), and communication module (160) operate. However, when the power supply from the vehicle is cut off due to a vehicle collision, power is supplied from the auxiliary battery (170) to the terminal sensor module (130), memory (140), processor (150), and communication module (160).

[0097] The user terminal (200) includes a communication unit (210), a memory (220), a control unit (230), and a user interface (240).

[0098] The communication unit (210) communicates wirelessly with the vehicle terminal (100) and can communicate wirelessly with the vehicle accident notification server (300). The communication unit (210) receives vehicle collision pre-collision data, vehicle collision post-collision data, and vehicle location information at the time of collision from the communication module (160) of the vehicle terminal (100). Then, the communication unit (210) transmits vehicle accident data and an accident occurrence report to the vehicle accident notification server (300).

[0099] The memory (240) can store received data, a program for providing a vehicle accident notification service, processed data generated by processing the received data, a program for performing an accident simulation, and various data. The memory (200) may be any non-volatile computer-readable storage medium that stores data and provides it to a processor. For example, the memory (240) may be a flash memory.

[0100] The control unit (230) determines whether an accident has occurred in the vehicle, analyzes the type and severity of the accident, generates an accident occurrence report, and transmits the accident data and the accident occurrence report to the vehicle accident notification server (200).

[0101] The control unit (230) includes an accident judgment unit (231), an accident type and severity analysis unit (232), a report generation unit (233), and an accident notification unit (234).

[0102] The accident determination unit (231) determines whether an accident has occurred using a previously learned AI deep learning algorithm. The accident determination unit (231) learns using data before a vehicle collision and data after a vehicle collision as input data, and determines whether an accident has occurred by comparing the learning results with previously stored accident pattern information. If the accident determination unit (231) determines that an accident has occurred, an accident occurrence confirmation message is displayed on the display of the user terminal (500) for a pre-set time under the control of the accident determination unit (231).

[0103] The accident type and severity analysis unit (232) analyzes the accident type and severity of the vehicle using a pre-trained AI deep learning algorithm when it is determined that an accident has occurred.

[0104] The types of vehicle accidents include whether the accident is a collision between vehicles, a collision between a vehicle and a person, a collision between a vehicle and surrounding facilities, or a rollover accident.

[0105] In addition, vehicle accident types include offset collisions, collisions with hard objects, collisions with soft objects, underliner collisions, frontal collisions, side collisions, local collisions, and frontal collisions.

[0106] Accident severity refers to the severity of the vehicle collision and the severity of injuries sustained by occupants according to the aforementioned accident types.

[0107]

[0108] FIG. 6 is a diagram showing vehicle collision pre-collision data and vehicle collision post-collision data divided into a plurality of segments based on the collision time according to various embodiments of the present invention.

[0109] Referring to (A) of FIG. 6, the accident type and severity analysis unit (232) generates multiple segments (Seg. A1 to Seg. Bn) by dividing the vehicle collision data and the vehicle collision data into pre-set time intervals based on the collision time. According to an embodiment, multiple segments can be generated by dividing the vehicle collision data and the vehicle collision data into equal time intervals.

[0110] Referring to (B), the accident type and severity analysis unit (232) can generate segments with different time lengths. Specifically, the time length of segments generated from data before the vehicle collision (Seg. A1 to Seg. An) becomes shorter as it approaches the time of the collision, and the time length of segments generated from data after the vehicle collision (Seg. B1 to Seg. Bn) becomes shorter as it approaches the time of the collision and longer as it moves away from the time of the collision.

[0111] (C) Referring to the accident type and severity analysis unit (232), the number of segments (Seg. B1 to Seg. Bn) generated from the data after the vehicle collision is greater than the number of segments (Seg. A1 to Seg. Am) generated from the data before the vehicle collision (m <n) 세그먼트를 생성할 수 있다.

[0112] In this way, by generating a relatively large number of segments (Seg. B1 to Seg. Bn) from the data after a vehicle collision and generating a relatively large number of segments (Seg. A1 to Seg. Bn) by shortening the time interval at the time adjacent to the collision point, the accuracy of accident type and severity analysis can be improved.

[0113] The accident type and severity analysis unit (232) can generate multiple driving speed segments by dividing the driving speed data of the vehicle before the collision and the driving speed data of the vehicle after the collision into time intervals. Here, the driving speed data of the vehicle is the actual driving speed of the vehicle collected through the vehicle data collection unit (420).

[0114] The accident type and severity analysis unit (232) can generate multiple acceleration segments by dividing the acceleration data of the vehicle terminal (100) before the vehicle collision and the acceleration data of the vehicle terminal (100) after the vehicle collision into time intervals. Here, the acceleration data of the vehicle terminal (100) is data measured by the inertial measurement sensor (131) and gyroscope of the vehicle terminal (100).

[0115] The accident type and severity analysis unit (232) can generate multiple ambient sound segments by dividing the ambient sound data before the vehicle collision and the ambient sound data after the vehicle collision into time intervals. Here, the ambient sound data is voice data measured by the microphone (133) of the vehicle terminal (100).

[0116] The accident type and severity analysis unit (232) compresses each of the multiple driving speed segments, multiple acceleration segments, and multiple ambient sound segments into multiple different sizes. According to an embodiment, the accident type and severity analysis unit (232) applies a Convolutional Neural Network (CNN) to compress each of the segments into multiple different sizes.

[0117] Specifically, the accident type and severity analysis unit (232) compresses each of the driving speed segments to a different size to generate a driving speed segment compression signal, compresses each of the acceleration segments to a different size to generate an acceleration segment compression signal, and compresses each of the surrounding sound segments to a different size to generate a surrounding sound segment compression signal.

[0118] The convolutional neural network individually analyzes the aforementioned data to train a data recognition model, converts the trained data into a vector form using convolution operations, and generates a feature map by applying multiple filters to the converted vector data. The size of the generated feature map is reduced using pooling, but by calculating the representative values ​​of the data to reduce the size, it minimizes effects such as changes in size, distortion, and warping of the feature map.

[0119] The pooling described above reduces the size of the convolutional neural network by reducing the horizontal and vertical space through the reduction of feature values ​​in the feature map to a single representative value. Pooling is divided into max pooling and average pooling depending on the method of setting the representative value; max pooling sets the maximum value among the feature values ​​in the feature map as the representative value, while average pooling sets the average value of the feature values ​​in the feature map as the representative value.

[0120] The accident type and severity analysis unit (232) can adjust the degree of compression of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal by adjusting the number of pooling operations of the convolutional neural network. Each time pooling is applied 1, 2, 3, or 4 times to each of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal, each of the compressed signals can be reduced to 1 / 2, 1 / 4, 1 / 8, or 1 / 16.

[0121] The accident type and severity analysis unit (232) may vary the number of times pooling is applied for each of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal. As a result, the number of generated driving speed segment compressed signals, acceleration segment compressed signals, and ambient sound segment compressed signals, and the degree of compression may vary. According to an embodiment, the accident type and severity analysis unit (232) may apply pooling to the driving speed segment compressed signal and acceleration segment compressed signal more times than to the pooling of the ambient sound segment compressed signal.

[0122] The accident type and severity analysis unit (232) can extract feature points from each of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal, and determine the accident type and severity by comparing the extracted feature points with the previously obtained feature points of accident type and severity. Here, the previously obtained feature points of accident type and severity are data obtained from various actual accident cases and / or crash test processes, and include information on the accident type, the severity of the collision according to the accident type, and the injury of the occupant.

[0123] The accident type and severity analysis unit (232) can determine the accident type and severity by comparing the feature points extracted from the driving speed segment compression signal with the feature points of the driving speed already secured.

[0124] And the accident type and severity analysis unit (232) can determine the accident type and severity by comparing the feature points extracted from the acceleration segment compression signal with the feature points of the acceleration already secured.

[0125] In addition, the accident type and severity analysis unit (232) can determine the accident type and severity by comparing the feature points extracted from the ambient sound segment compressed signal with the feature points of the ambient sound already secured.

[0126] In addition, the accident type and severity analysis unit (232) can determine the type of accident by comparing the feature points extracted from the driving speed segment compressed signal with the feature points extracted from the acceleration segment compressed signal.

[0127] The above driving speed segment compressed signal is data obtained from the actual driving speed of the vehicle, whereas the above acceleration segment compressed signal is acceleration data measured by the vehicle terminal. In the event of a collision of the vehicle, the actual driving speed change of the vehicle and the acceleration data measured by the vehicle terminal (100) may differ. For example, in the case of a light collision or a collision with a soft object, the collision energy is absorbed by the vehicle body, so the actual driving speed change of the vehicle and the acceleration data measured by the vehicle terminal may differ.

[0128] The accident type and severity analysis unit (232) can determine that the accident is a light collision or a collision with a soft object if the difference between the feature points extracted from the driving speed segment compression signal and the feature points extracted from the acceleration segment compression signal exceeds a threshold value.

[0129] In addition, the accident type and severity analysis unit (232) can analyze the accident type and severity using the following [Equation 1].

[0130] [Formula 1]

[0131] L = αL1 + βL2 + γL3

[0132] Here,

[0133] α is a weight for feature points extracted from the driving speed segment compressed signal, and

[0134] β is a weight for feature points extracted from the acceleration segment compressed signal, and

[0135] γ is a weight for feature points extracted from the ambient sound segment compressed signal, and

[0136] L1 is the difference value between the feature points extracted from the driving speed segment compressed signal and the previously secured feature points of the driving speed, and

[0137] L2 is the difference value between the feature points extracted from the acceleration segment compressed signal and the previously secured acceleration feature points, and

[0138] L3 is the difference between the feature points extracted from the ambient sound segment compressed signal and the previously secured feature points of the ambient sound.

[0139]

[0140] According to an embodiment, the weight for feature points extracted from the driving speed segment compressed signal and the weight for feature points extracted from the acceleration segment compressed signal may be greater than the weight for feature points extracted from the ambient sound segment compressed signal.

[0141]

[0142] In addition, according to another embodiment of the present invention, the accident type and severity analysis unit (232) converts the surrounding sound data before the vehicle collision and the surrounding sound data after the vehicle collision, respectively, into acoustic data in the frequency domain. Then, it detects high-frequency acoustic components above a reference frequency in the acoustic data in the frequency domain and divides the high-frequency acoustic components into preset time intervals to generate a plurality of high-frequency acoustic segments.

[0143] The accident type and severity analysis unit (232) applies a Convolutional Neural Network (CNN) to compress each of the multiple high-frequency acoustic segments into multiple different sizes to generate a compressed signal, applies the compressed signal to a convolution operation to convert it into a vector form, and applies multiple filters to the converted vector data to generate a feature map.

[0144] The accident type and severity analysis unit (232) can adjust the degree of compression of the compressed signal by adjusting the number of pooling cycles of the convolutional neural network, extract feature points from each compressed signal, and determine the accident type and severity by comparing the feature points.

[0145] In an embodiment of the present invention, high-frequency acoustic components are extracted from ambient sound data before and after a vehicle collision, and by compressing and comparing feature points, the type and severity of the accident can be rapidly determined with a small amount of data processing. Furthermore, since high-frequency acoustic components from which feature points can be effectively extracted are processed, accurate determination of the type and severity of the accident is possible.

[0146]

[0147] The report generation unit (233) generates an accident occurrence report. The accident occurrence report may include driver information, accident type information, accident occurrence time information, accident occurrence location information, vehicle insurance information, vehicle body movement information, weather information, temperature information, and satellite image information of the accident location. Additionally, vehicle identification information may be further included.

[0148] Driver information is user information stored on a user terminal and may include name, age, gender, height, weight, blood type, driver's license information, driving history information, etc.

[0149] Accident type information refers to an accident type calculated using information measured by a vehicle terminal and a sensor module, and topographical information of the vehicle's driving location. The accident type information (52) indicates whether the vehicle accident is a collision between vehicles, a collision between a vehicle and a person, a collision between a vehicle and surrounding facilities, or a vehicle rollover accident.

[0150] Accident occurrence time information is the time information when a vehicle impact was detected, and includes year / month / day / hour information.

[0151] Accident location information refers to the geographical location where a vehicle accident occurred, and includes GPS information and address information.

[0152] Vehicle body movement information refers to the movement of the vehicle body immediately before and after a vehicle accident, and includes information on the vehicle body's speed, roll, yaw, and acceleration. Vehicle body movement information can be calculated by inputting information measured by vehicle terminals and sensor modules into a pre-stored algorithm. The vehicle body movement information displays yaw, roll, and acceleration separately and can be displayed as numerical values ​​or graphs.

[0153] Weather information refers to the weather information of the area where the vehicle accident occurred.

[0154] Temperature information refers to the temperature of the area where the vehicle accident occurred.

[0155] Satellite photo information of the accident site is provided by displaying the location of the vehicle on the satellite photo of the site where the vehicle accident occurred.

[0156] Vehicle identification information includes information about the vehicle, such as vehicle type, year of manufacture, and license plate number.

[0157] FIG. 7 is a drawing showing an accident occurrence report generated according to an embodiment of the present invention.

[0158] Referring to FIG. 7, the accident report provides the number of times the occupant touched the user terminal before the accident (51), the date and time of the accident (52), the weather and temperature (53), the location of the accident (54), the address of the accident (55), an overview of the accident (56), a summary of the accident (57), and map information (58).

[0159] The accident summary (56) includes the type of dangerous driving judgment, vehicle speed, engine temperature, vehicle RPM, time of accident, vehicle manufacturer, insurance company, vehicle model name, vehicle number, guardian name, guardian contact information, user ID, user name, insurance company contact information, and gear shift mode.

[0160] The accident summary (57) includes data on the vehicle speed before the accident and the vehicle speed after the accident, weather data including weather, precipitation, wind speed and wind direction, and the gear mode at the moment of the accident.

[0161] In addition, the accident report provides a graph of vehicle speed change (61), a graph of vehicle RPM change (62), and a graph of engine temperature change (63) before and after the accident.

[0162] The vehicle speed change graph (61) displays the vehicle speed change during a certain period of time before the accident and the vehicle speed change during a certain period of time after the accident.

[0163] The vehicle RPM change graph (62) displays the vehicle RPM change during a certain period of time before the accident and the vehicle RPM change during a certain period of time after the accident.

[0164] The engine temperature change graph (63) displays the engine temperature change during a certain period of time before the accident and the engine temperature change during a certain period of time after the accident.

[0165] In addition to the information described in Fig. 7, the accident report displays the type of accident, the severity of the accident, and the severity of the injury to the occupant as determined by the accident type and severity analysis unit.

[0166] At least one of the following accident types is indicated: collision between vehicles, collision between a vehicle and a person, collision between a vehicle and surrounding facilities, vehicle rollover accident, offset collision, collision with a hard object, collision with a soft object, underliner collision, frontal collision, side collision, local collision, and frontal collision.

[0167] Accident severity can be expressed as a numerical value or level indicating the extent of vehicle damage caused by the collision.

[0168] The severity of occupant injury can be expressed as a numerical value or level, indicating the extent of injury to the occupant caused by the collision.

[0169]

[0170] When the report generation unit (243) determines that a vehicle accident has occurred, an accident occurrence confirmation message is displayed on the display of the user terminal (200) for a preset time, and an accident occurrence report is generated when the user touches the message, checks the message, or when the preset time has elapsed. The preset time may be set to 20 to 40 seconds. Additionally, the preset time may be provided to be adjusted after the user runs the client program.

[0171] The accident notification unit (234) transmits accident data and an accident occurrence report to the vehicle accident notification server (300) when the user touches a message displayed on the display of the user terminal (200), checks the message, or when a preset time elapses.

[0172]

[0173] FIG. 8 is a drawing showing a vehicle accident notification server according to an embodiment of the present invention.

[0174] Referring to FIG. 8, the vehicle accident notification server (300) includes a communication module (310), a memory module (320), and a central processing unit (330).

[0175] The communication module (310) is provided to enable wired and wireless data communication through a network and can transmit and receive data with the user terminal (200). The communication module (310) receives accident data and an accident occurrence report from the user terminal (200) and can transmit accident notification data to the server, terminal, etc. of a relevant organization. At this time, the accident notification data and the accident occurrence report are data that notifies the fact of an accident to an insurance company, an emergency rescue agency such as 119, a vehicle towing company, or other emergency contacts set by the user.

[0176] The memory module (320) can store received data, a program for providing a vehicle accident notification service, processed data generated by processing the received data, a program for performing an accident simulation, and various data. Here, the memory module (320) is a general term for a non-volatile storage device that continues to maintain stored information even when power is not supplied.

[0177] The central processing unit (330) executes a program stored in the memory module (320) and applies the stored data to the program. The central processing unit (330) can be understood as a processor capable of controlling the communication module (310) and the memory module (320), and capable of reading and processing data received through the communication module (310) or stored in the memory module (320) according to a predetermined program.

[0178] FIG. 9 is a drawing showing the detailed configuration of a central processing unit according to an embodiment of the present invention.

[0179] Referring to FIG. 9, the central processing unit (330) includes an accident occurrence notification unit (331), a pattern information generation unit (332), an accident occurrence classification criteria generation unit (333), a manual information provision unit (334), a simulation unit (335), and an automatic question and answer unit (336).

[0180] The accident occurrence notification unit (331) transmits accident notification data to pre-set accident response agencies. That is, the accident occurrence notification unit (331) is provided with a URL address, phone number, etc., to transmit accident notification data to each of the following: an insurance company, a PM company, an emergency rescue agency such as 119, a vehicle towing company, a school control center, a living lab center, an emergency agency, family / acquaintances, and other emergency contacts set by the user.

[0181] The pattern information generation unit (332) generates a pattern between accident data and the actual occurrence of an accident. That is, the pattern information generation unit (332) defines the accident data as an input factor and the output of the accident type and severity analysis unit (232) as an output factor, and then derives the correlation between the input factor and the output factor to generate pattern information. The pattern information generation unit (332) can be implemented through deep learning based on a deep neural network to derive the correlation between the input factor and the output factor. Additionally, the pattern information generation unit (332) can update existing pattern information with new pattern information. When the pattern information is updated, the pattern information generation unit (332) transmits the updated pattern information to the user terminal (200), and the accident type and severity analysis unit (232) can detect whether an accident has occurred using the updated pattern information.

[0182] The accident occurrence classification standard generation unit (333) generates accident type classification standards. That is, the accident occurrence classification standard generation unit (333) defines accident data as input factors and the output of the accident type and severity analysis unit (232) as output factors, and then derives a correlation between the input factors and the output factors to generate accident type classification standards. Additionally, the accident occurrence classification standard generation unit (333) can update existing accident type classification standards through new accident type classification standards. The accident occurrence classification standard generation unit (333) can be implemented through deep learning based on a deep neural network to derive a correlation between the input factors and the output factors.

[0183] The manual information provision unit (334) enables the accident response manual to be transmitted to the user terminal (200). The response manual may be provided in the form of text information displayed on the user terminal (200) or voice information output through a speaker provided to the user terminal (200). The response manual may include notification information regarding the situation in which accident notification data is notified to a pre-set agency, notification information instructing to photograph the accident scene for post-accident processing, and information instructing to move to a safe place and wait. When accident data is automatically sent from the accident notification unit (234) of the user terminal (200) after the expiration of the waiting period for transmission, the manual information provision unit (334) may transmit a consciousness confirmation message to the user terminal (200) to confirm whether the user is conscious. If a reply to the consciousness confirmation message is not received within a preset period, the manual information provision unit (334) may transmit to the accident notification unit (234) that the user is in a state of unconsciousness, and the accident notification unit (234) may additionally transmit an emergency message to a pre-set agency indicating that it may be an emergency situation.

[0184] The simulation unit (335) performs a simulation using accident data received from the user terminal (200). For example, the simulation unit (335) may be provided to perform a simulation based on the MADYMO (Mathematical Dynamic Models) program. The simulation unit (335) can perform the function of analyzing the cause of a vehicle accident and determining the causal relationship of an injury through a three-dimensional simulation using the received accident data. Subsequently, the simulation results may be provided to an accident handling agency to enable accurate situational judgment. The simulation unit (335) may be provided to perform a simulation for each of the different simulation conditions. That is, the simulation unit (335) includes various data necessary for analyzing the behavior of the occupants of the vehicle, such as data regarding the vehicle type and the structure of the vehicle accordingly, data regarding dummies used in experiments, data regarding safety components, and data regarding vehicle motion characteristics during a collision. Preferably, test conditions for each vehicle type are provided in a database state so that simulations can be performed quickly for various vehicle types.

[0185] The automatic question-and-answer unit (336) performs question-and-answer with the user terminal. The automatic question-and-answer unit (336) generates a query by analyzing an accident report based on a Large Language Model (LLM). The LLM model generates a query by analyzing visual elements such as text, images, tables, and graphs included in the accident report.

[0186] According to an embodiment, the automatic question-answering unit (336) identifies the layout of an image, text, and table in an accident report and uses the position coordinates of the layout to cut out the image, text, and table to create image files. Then, the image, text, and table are converted into Markdown format and information is extracted from each converted data. The extracted information is learned based on LLM to generate a query for the user.

[0187] The generated query is transmitted to the user terminal (200) via the communication module (310). The passenger speaks an answer to the voice query output from the user terminal (200), and the passenger's voice signal is received by the user terminal (200). The received voice signal is converted into text by the user terminal (200) and then transmitted to the communication module (210).

[0188] The automatic question-and-answer unit (336) learns and analyzes the user's answer content based on LLM.

[0189] When the automatic question-and-answer unit (336) determines that a re-question is necessary based on the user's answer, it reproduces the question based on the user's answer and information extracted from the accident report.

[0190] The reproduced query is transmitted to the user terminal (200) via the communication module (310).

[0191] Through the process described above, questions and answers are exchanged between the passenger and the vehicle accident notification server (300) via the user terminal (200).

[0192] The automatic question-and-answer unit (336) can analyze the passenger's emotions and whether there is an emergency situation based on the words included in the passenger's answer and the voice volume. Specifically, the automatic question-and-answer unit (336) can determine the emergency situation based on whether the passenger's answer includes words indicating an emergency situation and whether the voice volume is above a reference value in decibels (dB).

[0193] The above accident occurrence notification unit (331) can transmit the results analyzed from the accident occurrence report and the automatic question and answer unit (336) to the accident response agency.

[0194]

[0195] A vehicle accident notification system (10) according to one embodiment of the present invention is provided such that when an accident occurs, the fact of such occurrence is detected through a vehicle terminal (100) and a user terminal (200), the fact of the accident is notified to a vehicle accident notification server (300), and the fact of the accident is notified to an agency for accident processing. Accordingly, the fact of the accident is notified to each agency more quickly than when the user directly notifies each agency of the accident for accident processing, and accident processing can be carried out. In particular, considering that the driver becomes flustered when an accident occurs, the vehicle accident notification system according to one embodiment of the present invention enables significantly faster accident notification and accident processing.

[0196] In addition, the vehicle accident notification system (10) according to one embodiment of the present invention notifies an agency to respond to the accident even if the driver loses consciousness due to the accident, so that prompt notification of the accident and rescue of the driver can be carried out even if the driver loses consciousness due to the accident.

[0197] In addition, a vehicle accident notification system (10) according to one embodiment of the present invention can provide a safety integrated data management base service and a data analysis and linkage service based on AI deep learning. The safety integrated data management base service can perform a safety data linkage collection system, integrated big data construction, and infrastructure system operation and management. The data analysis and linkage system can provide an AI-based risk section analysis system, a GIS safety analysis system, and a data linkage / sharing service with related organizations.

[0198]

[0199] FIG. 10 is a flowchart illustrating a method for detecting a vehicle accident and automatically reporting it using a vehicle terminal and a user terminal according to the embodiments of FIG. 1 to 9.

[0200] Referring to FIG. 10, the vehicle accident detection and automatic reporting method includes a vehicle data collection step (S10), a vehicle collision determination step (S20), a vehicle data transmission step (S30), an accident determination step (S40), an accident type and severity analysis step (S50), an accident occurrence report generation step (S60), and an accident notification step (S70).

[0201] The vehicle data collection step (S10) collects the vehicle's driving data, acceleration data, GPS data, and ambient sound data.

[0202] Driving data of the vehicle is collected from the vehicle through the vehicle data collection unit (120). The driving data of the vehicle includes the vehicle's driving speed, driving location data, driving distance data, RPM, brake signal data, accelerator pedal signal data, vehicle interior temperature data, and vehicle exterior temperature data.

[0203] Acceleration data is collected through an inertial measurement sensor (131) of a vehicle terminal (100). The acceleration data can measure changes in acceleration along the x-axis, y-axis, and z-axis of the vehicle. When acceleration data indicating a vehicle collision is detected, the inertial measurement sensor (131) wakes up the processor (150) and switches it to an operating mode.

[0204] The GPS data is collected through the GPS module (132) for accurate location data and movement data of the vehicle terminal (400).

[0205] The vehicle collision determination step (S20) is performed when the processor (150) is switched to an operating mode by the inertial measurement sensor (131). In the vehicle collision determination step (S20), the processor (150) performs a collision detection algorithm and analyzes the data measured by the inertial measurement sensor (131) to determine a vehicle collision. The processor (150) determines that a collision has occurred if the acceleration data in at least one axis among the data measured by the inertial measurement sensor (131) exceeds an acceleration threshold, or if the data measured by the gyroscope exceeds a threshold.

[0206] When the processor (150) determines that a collision has occurred, it calculates the time of the collision and calculates the vehicle location data at the time of the collision from the location data received by the GPS module (432).

[0207] Additionally, when the processor (150) determines that a vehicle collision has occurred, it extracts vehicle collision pre-data acquired during a preset time before the collision point and vehicle collision post-data acquired during a preset time after the collision point.

[0208] The vehicle collision pre-data includes driving data of the vehicle before the collision collected by the vehicle data collection unit (120), acceleration data of the vehicle terminal (100) before the vehicle collision measured by the inertial measurement sensor (131), and ambient sound data before the vehicle collision measured by the microphone (133).

[0209] The post-vehicle collision data includes driving data of the vehicle after the collision collected by the vehicle data collection unit (120), acceleration data of the vehicle terminal (100) after the vehicle collision measured by the inertial measurement sensor (131), and ambient sound data after the vehicle collision measured by the microphone (133).

[0210] The vehicle data transmission step (S30) transmits vehicle collision pre-collision data, vehicle collision post-collision data, and vehicle location data at the time of collision, which are output from the processor (150), to the user terminal (200).

[0211] In the accident determination step (S40), the control unit (230) of the user terminal (200) determines whether an accident has occurred using a previously learned AI deep learning algorithm. The accident determination step (S40) learns using data before the vehicle collision and data after the vehicle collision as input data, and determines whether an accident has occurred by comparing the learning results with previously stored accident pattern information.

[0212] If it is determined that an accident has occurred, an accident occurrence confirmation message is displayed on the display of the user terminal (500) for a preset time. At the same time, an accident type and severity analysis step is performed.

[0213] In the accident type and severity analysis step (S50), the control unit (230) analyzes the accident type and severity of the vehicle using a pre-trained AI deep learning algorithm.

[0214] Specifically, the control unit (230) generates a plurality of segments (Seg. A1 to Seg. Bn) by dividing the vehicle collision data and the vehicle collision data into preset time intervals based on the collision time.

[0215] Additionally, the control unit (230) can generate segments such that the number of segments (Seg. B1 to Seg. Bn) generated from the data after the vehicle collision is greater than the number of segments (Seg. A1 to Seg. An) generated from the data before the vehicle collision.

[0216] Additionally, the control unit (230) can generate segments (Seg. A1 to Seg. An) generated from data prior to a vehicle collision such that the time length becomes shorter as the collision time approaches. Additionally, the control unit (230) can generate segments (Seg. B1 to Seg. Bn) generated from data prior to a vehicle collision such that the time length becomes shorter as the collision time approaches and longer as the collision time moves away.

[0217] Additionally, the control unit (230) can generate multiple driving speed segments by dividing the driving speed data of the vehicle before the collision and the driving speed data of the vehicle after the collision into time intervals.

[0218] Additionally, the control unit (230) can generate multiple acceleration segments by dividing the acceleration data of the vehicle terminal (100) before the vehicle collision and the acceleration data of the vehicle terminal (100) after the vehicle collision into time intervals.

[0219] Additionally, the control unit (230) can generate multiple ambient sound segments by dividing the ambient sound data before the vehicle collision and the ambient sound data after the vehicle collision into time intervals.

[0220] The control unit (230) compresses each of the multiple driving speed segments, multiple acceleration segments, and multiple ambient sound segments into multiple different sizes. The control unit (230) applies a convolutional neural network (CNN) to compress each of the segments into multiple different sizes.

[0221] According to an embodiment, the control unit (230) compresses each of the driving speed segments to a different size to generate a driving speed segment compression signal, compresses each of the acceleration segments to a different size to generate an acceleration segment compression signal, and compresses each of the surrounding sound segments to a different size to generate a surrounding sound segment compression signal.

[0222] The control unit (230) can adjust the degree of compression of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal by adjusting the number of pooling cycles of the convolutional neural network.

[0223] Whenever pooling 1, 2, 3, or 4 is applied to each of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal, each of the said compressed signals can be reduced to 1 / 2, 1 / 4, 1 / 8, or 1 / 16.

[0224] The control unit (230) can vary the number of times pooling is applied to each of the driving speed segment compression signal, the acceleration segment compression signal, and the ambient sound segment compression signal. As a result, the number of generated driving speed segment compression signals, the acceleration segment compression signals, and the ambient sound segment compression signals, as well as the degree of compression, may vary.

[0225] The control unit (230) can extract feature points from each of the driving speed segment compressed signal, acceleration segment compressed signal, and ambient sound segment compressed signal, and determine the type and severity of the accident by comparing the extracted feature points with the feature points of the accident type and severity that have been secured.

[0226] The control unit (230) can determine the type and severity of an accident by comparing the feature points extracted from the driving speed segment compression signal with the feature points of the driving speed already secured.

[0227] The control unit (230) can determine the type and severity of an accident by comparing the feature points extracted from the acceleration segment compression signal with the feature points of the acceleration already secured.

[0228] The control unit (230) can determine the type and severity of an accident by comparing the feature points extracted from the ambient sound segment compressed signal with the feature points of the ambient sound already secured.

[0229] The control unit (230) can determine the type of accident by comparing the feature points extracted from the driving speed segment compressed signal with the feature points extracted from the acceleration segment compressed signal.

[0230] The control unit can analyze the type and severity of the accident using [Equation 1].

[0231] In [Equation 1], the control unit can analyze the type and severity of accidents by making the weights for feature points extracted from the driving speed segment compressed signal and the weights for feature points extracted from the acceleration segment compressed signal greater than the weights for feature points extracted from the ambient sound segment compressed signal.

[0232] In addition, the control unit can convert ambient sound data before the vehicle collision and ambient sound data after the vehicle collision into acoustic data in the frequency domain, detect high-frequency acoustic components above a reference frequency in the acoustic data in the frequency domain, and generate multiple high-frequency acoustic segments by dividing the high-frequency acoustic components into preset time intervals.

[0233] The control unit can generate a compressed signal by applying a Convolutional Neural Network (CNN) to compress each of multiple high-frequency acoustic segments into multiple different sizes, convert the compressed signal into a vector form by applying a convolution operation, and generate a feature map by applying multiple filters to the converted vector data.

[0234] The control unit can adjust the degree of compression of the compressed signal by adjusting the number of pooling cycles of the convolutional neural network, extract feature points from each compressed signal, and determine the type and severity of the accident by comparing the feature points.

[0235] Once the analysis of accident type and severity is complete, the accident occurrence report generation step (S60) is performed.

[0236] The accident occurrence report generation step (S60) generates an accident occurrence report. The accident occurrence report includes driver information, accident type information, accident severity information, occupant injury severity information, accident occurrence time information, accident occurrence location information, vehicle identification information, vehicle insurance information, vehicle body movement information, weather information, temperature information, and satellite image information of the accident location.

[0237] The automatic question-and-answer step (S70) analyzes an accident occurrence report based on a Large Language Model (LLM) to generate a query, and transmits the generated query to a user terminal (200).

[0238] The query is output as a voice signal from the user terminal (200), and when a passenger answers, the passenger's voice signal is transmitted to the communication module (210) through the user terminal (200).

[0239] The automatic question-and-answer step (S70) learns and analyzes the user's answer content based on LLM, and if it is determined that a re-question is necessary, it reproduces a query based on the user's answer content and information extracted from the accident report. The reproduced query is transmitted to the user terminal (200) via the communication module (310). Through the process described above, questions and answers are exchanged between the passenger and the vehicle accident notification server (300) via the user terminal (200).

[0240] The automatic question-and-answer step (S70) can analyze the passenger's emotions and whether there is an emergency situation based on the words included in the passenger's answer and the voice volume. Specifically, the automatic question-and-answer unit (336) can determine an emergency situation based on whether the passenger's answer includes words indicating an emergency situation and whether the voice volume appears at a decibel (dB) above a reference value.

[0241] The accident notification step (S80) transmits accident notification data, an accident occurrence report, and automatic question-and-answer data to a pre-configured accident response agency.

[0242]

[0243] Although the present invention has been described in detail using preferred embodiments, the scope of the invention is not limited to specific embodiments and should be interpreted by the appended claims. Furthermore, those skilled in the art will understand that many modifications and variations are possible without departing from the scope of the invention.

[0244] The present invention can be used to automatically report the occurrence of a vehicle accident to an emergency rescue agency when a vehicle accident occurs.

Claims

1. A vehicle terminal connected to an OBD connector installed in a vehicle; and It includes a user terminal owned by a user riding in the vehicle that is capable of communicating with the above vehicle terminal, The above vehicle terminal is, A driving information collection unit that collects driving data of the vehicle from the above OBD connector; An inertial measurement sensor for measuring acceleration data of the above vehicle terminal; A processor that determines a vehicle collision using acceleration data measured by the inertial measurement sensor, calculates the collision time, and extracts vehicle pre-collision data acquired during a preset time before the collision time and vehicle post-collision data acquired during a preset time after the collision time; and It includes a communication module capable of communicating with the vehicle terminal and transmitting the vehicle collision pre-data and vehicle collision post-data to the user terminal, The above user terminal is, An accident determination unit that uses a previously trained AI deep learning algorithm to train the vehicle collision pre-data and vehicle collision post-data as input data, and compares the training results with previously stored accident pattern information to determine whether an accident has occurred; and A vehicle accident detection and automatic reporting system comprising an accident type and severity analysis unit that, using a pre-trained AI deep learning algorithm based on the collision time, divides the vehicle collision pre-data and the vehicle collision post-data into pre-set time units to generate a plurality of segments, compresses each of the segments to different sizes to generate segment compression signals, extracts feature points from each of the segment compression signals, and determines the accident type and severity by comparing the feature points with pre-secured accident type and severity feature points.

2. In Paragraph 1, The above vehicle terminal It further includes a GPS module that receives location data of the above-mentioned vehicle terminal, The above communication module is a vehicle accident detection and automatic reporting system that transmits the location data at the time of the collision to the user terminal.

3. In Paragraph 1, The above-described inertial measurement sensor is a vehicle accident detection and automatic reporting system that switches the processor to an operating mode when acceleration data indicating a vehicle collision is detected.

4. In Paragraph 1, The above vehicle terminal The vehicle further includes a microphone for receiving ambient sound signals, The above vehicle collision pre-data includes ambient sound data received by the microphone during a preset time prior to the collision point, and The vehicle accident detection and automatic reporting system includes the vehicle accident post-collision data received by the microphone for a preset time after the collision point.

5. In Paragraph 1, The above accident type and severity analysis department A vehicle accident detection and automatic reporting system that generates segments such that the number of segments generated from the data after the vehicle collision is greater than the number of segments generated from the data before the vehicle collision.

6. In Paragraph 1, The above accident type and severity analysis unit, The closer the time of the collision is to the above collision point, the shorter the time length of the segment generated from the vehicle collision pre-collision data and the vehicle collision post-collision data becomes, and A vehicle accident detection and automatic reporting system that divides the segments such that the time length of the segments generated from the vehicle collision pre-data and vehicle collision post-data increases as the distance from the collision point increases.

7. In Paragraph 1, The above segment compression signal includes a driving speed segment compression signal and an acceleration segment compression signal, and The above accident type and severity analysis unit, A vehicle accident detection and automatic reporting system that determines a light collision or a collision with a soft object when the difference between a feature point extracted from the above driving speed segment compressed signal and a feature point extracted from the above acceleration segment compressed signal exceeds a threshold value.

8. In Paragraph 1, The above segment compression signal includes a driving speed segment compression signal, an acceleration segment compression signal, and an ambient sound segment compression signal, and The above accident type and severity analysis unit is a vehicle accident detection and automatic reporting system that analyzes accident type and severity using [Formula 1]. [Formula 1] L = αL1 + βL2 + γL3 Here, α is a weight for feature points extracted from the driving speed segment compressed signal, and β is a weight for feature points extracted from the acceleration segment compressed signal, and γ is a weight for feature points extracted from the ambient sound segment compressed signal, and L1 is the difference value between the feature points extracted from the driving speed segment compressed signal and the previously secured feature points of the driving speed, and L2 is the difference value between the feature points extracted from the acceleration segment compressed signal and the previously secured acceleration feature points, and L3 is the difference between the feature points extracted from the ambient sound segment compressed signal and the previously secured feature points of the ambient sound.

9. A vehicle data collection step of collecting driving data of the vehicle from an OBD connector mounted on the vehicle and measuring acceleration data of a vehicle terminal connected to the OBD connector; A vehicle collision determination step that determines a vehicle collision using the above acceleration data, calculates the collision time, and extracts vehicle pre-collision data acquired during a preset time before the collision time and vehicle post-collision data acquired during a preset time after the collision time; A vehicle data transmission step of transmitting the vehicle collision pre-collision data and the vehicle collision post-collision data to a user terminal of a user riding in the vehicle; An accident determination step of determining whether an accident has occurred by comparing the learning results with previously stored accident pattern information, using a previously trained AI deep learning algorithm to learn the vehicle collision pre-data and vehicle collision post-data as input data; and A vehicle accident detection and automatic reporting method further comprising an accident type and severity analysis step of using a previously trained AI deep learning algorithm to generate a plurality of segments by dividing the vehicle collision pre-data and the vehicle collision post-data into pre-set time units based on the collision time, generating segment compression signals by compressing each of the segments to different sizes, extracting feature points from each of the segment compression signals, and determining the accident type and accident severity by comparing the feature points with pre-obtained feature points of accident type and severity.

10. In Paragraph 9, The above accident type and severity analysis steps are, A vehicle accident detection and automatic reporting method that generates segments such that the number of segments generated from the data after the vehicle collision is greater than the number of segments generated from the data before the vehicle collision.

11. In Paragraph 9, The above accident type and severity analysis steps are, The closer the time of the collision is to the above collision point, the shorter the time length of the segment generated from the vehicle collision pre-collision data and the vehicle collision post-collision data becomes, and A vehicle accident detection and automatic reporting method that divides the segments such that the time length of the segments generated from the vehicle collision pre-data and vehicle collision post-data becomes longer as the distance from the collision point increases.

12. In Paragraph 9, The above segment compression signal includes a driving speed segment compression signal and an acceleration segment compression signal, and The above accident type and severity assessment steps are, A vehicle accident detection and automatic reporting method that determines a light collision or a collision with a soft object when the difference between a feature point extracted from the driving speed segment compression signal and a feature point extracted from the acceleration segment compression signal exceeds a threshold value.

13. In Paragraph 9, Accident occurrence report generation step for generating an accident occurrence report; and It further includes an accident notification step of transmitting the above-mentioned accident occurrence report to the accident response agency, The above accident occurrence report is a vehicle accident detection and automatic reporting method including the above accident type and accident severity information.

14. In Paragraph 9, The above segment compression signal includes a driving speed segment compression signal, an acceleration segment compression signal, and an ambient sound segment compression signal, and The above accident type and severity analysis unit is a vehicle accident detection and automatic reporting method that analyzes accident type and severity using [Formula 1]. [Formula 1] L = αL1 + βL2 + γL3 Here, α is a weight for feature points extracted from the driving speed segment compressed signal, and β is a weight for feature points extracted from the acceleration segment compressed signal, and γ is a weight for feature points extracted from the ambient sound segment compressed signal, and L1 is the difference value between the feature points extracted from the driving speed segment compressed signal and the previously secured feature points of the driving speed, and L2 is the difference value between the feature points extracted from the acceleration segment compressed signal and the previously secured acceleration feature points, and L3 is the difference between the feature points extracted from the ambient sound segment compressed signal and the previously secured feature points of the ambient sound.

Images