Information processing device, information processing method, and information processing program

Abstract

To provide an information processing device, an information processing method, and an information processing program, which enable accurate estimation of the injury level of a whole body of an injured person at the time of an automobile collision accident using a trained machine learning model.SOLUTION: A computing device 10 is configured to build a trained injury prediction ROM through machine learning based on the collision condition at the time of a vehicle accident and a result of collision simulation using a vehicle model, and then use the trained injury prediction ROM to estimate an injury risk for each part of the body of an injured person from the collision condition detected by a group of sensors of a vehicle 200 at the time of the vehicle accident.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program for estimating the injury level of an injured person's entire body in the event of a car collision. [Background technology]

[0002] In the event of a car accident, it is necessary to provide prompt emergency medical care to the injured as well as accurately ascertain the level of their injuries.

[0003] Patent Document 1 discloses an invention relating to a collision injury prediction method for predicting the injury level of an injured person in the event of a car accident. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2020-061088 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the invention described in Patent Document 1 uses actual accident data as training data for the machine learning model. Actual accident data records the most severely injured area, so while there is a relatively large amount of data related to the head, there is little data related to other areas. This means there is a paucity of data recording details of the accident, including injuries to the entire body, making it difficult to obtain an appropriate prediction model through machine learning. As a result, there are problems with the accuracy of estimating the injury level of the injured person's entire body.

[0006] In consideration of the above facts, the present invention aims to provide an information processing device, an information processing method, and an information processing program that can accurately estimate the injury level of an injured person's entire body in the event of a car collision accident using a trained machine learning model. [Means for solving the problem]

[0007] In order to achieve the above object, the information processing device according to claim 1 comprises: In collision simulation using vehicle model and human body model Collision conditions, and the injury value of each part of the human body model obtained by the collision simulation; Collision simulation results and 、 Based on By limiting the collision conditions that are variables By using machine learning , to estimate the injury risk at each part of the injured person based on the collision conditions. Reduced dimensionality model Building a learning department and detected by the vehicle's sensors. information from Vehicle accident Collision conditions Calculate and the trained reduced dimension model is calculation In the department calculation was Vehicle accident Enter the collision conditions , the vehicle accident and a calculation unit that estimates the injury risk at each part of the injured person.

[0008] According to the information processing device described in claim 1, by using a trained dimension-reduced model in which the dimensions of a machine learning model trained from the results of a collision simulation are reduced, it becomes possible to estimate the injury level of an injured person's entire body in a short period of time from the collision conditions at the time of the accident.

[0009] In the information processing device according to claim 2, the group of sensors includes a collision direction between the vehicle and the injured person, a collision severity, and a contact position of the injured person with the vehicle. information Detect.

[0010] According to the information processing device of claim 2, ,car Information can be obtained including the direction of impact between the vehicle and the injured person, the severity of the impact, and the location of the injured person's contact with the vehicle.

[0011] The information processing device according to claim 3 comprises: The learning unit performs machine learning by limiting the collision conditions that are variables by adopting collision conditions that are guaranteed to be linearly independent of each other as independent variables. .

[0012] According to the information processing device of claim 3, By using machine learning with limited variables, such as collision conditions, it is possible to construct a reduced-dimensional model.

[0013] In order to achieve the above object, the information processing method according to claim 4 includes the steps of: constructing a dimension-reduced model for estimating injury risks at each body part of an injured person from collision conditions by performing machine learning while limiting collision conditions, which are variables, based on collision conditions in a collision simulation using a vehicle model and a human body model and results of the collision simulation including injury values ​​for each body part of the human body model obtained in the collision simulation; calculating collision conditions of a vehicle accident from information detected by a group of sensors provided in a vehicle; and inputting the calculated collision conditions of a vehicle accident into the trained dimension-reduced model and estimating injury risks at each body part of an injured person in the vehicle accident. The computer executes the process .

[0014] According to the information processing method described in claim 4, by using a trained dimension-reduced model in which the dimensions of a machine learning model trained from the results of a collision simulation are reduced, it is possible to estimate the injury level of an injured person's entire body in a short period of time from the collision conditions at the time of the accident.

[0015] In order to achieve the above object, the information processing program according to claim 5 causes a computer to: In collision simulation using vehicle model and human body model Collision conditions, and the injury value of each part of the human body model obtained by the collision simulation; Collision simulation results and 、 Based on By limiting the collision conditions that are variables By using machine learning , to estimate the injury risk at each part of the injured person based on the collision conditions. Reduced dimensionality model Building a learning department , detected by the vehicle's sensors information from Vehicle accident Collision conditions Calculate and the trained reduced dimension model is calculation In the department calculation was Vehicle accident Enter the collision conditions , the vehicle accident It functions as a calculation unit that estimates the injury risk at each part of the injured person.

[0016] According to the information processing program described in claim 5, by using a trained dimension-reduced model in which the dimensions of a machine learning model trained from the results of a collision simulation are reduced, it becomes possible to estimate the injury level of an injured person's entire body in a short period of time from the collision conditions at the time of the accident. [Effects of the Invention]

[0017] As described above, the information processing device, information processing method, and information processing program according to the present invention make it possible to accurately estimate the injury level of an injured person's entire body in the event of a car collision accident using a trained machine learning model. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a schematic diagram illustrating a configuration of an information processing device according to an embodiment of the present invention. [Figure 2] FIG. 1 is a block diagram showing the configuration of a vehicle. [Figure 3] FIG. 2 is a block diagram showing an example of a specific configuration of a calculation device according to the present embodiment. [Figure 4] (A) is a schematic diagram showing an example of a simulation in which a human FE model riding a bicycle FE model collides with a vehicle FE model, (B) is a schematic diagram showing an example of a collision, (C) is a schematic diagram showing an example of the relationship between head contact position and head injury value in a parametric simulation, and (D) is a schematic diagram showing an example of injury risk by body part of a human FE model in a parametric simulation. [Figure 5] 10 is a flowchart showing an example of a process for estimating an injury risk using a learned injury prediction ROM. DETAILED DESCRIPTION OF THE INVENTION

[0019] An information processing device 100 according to this embodiment will be described below with reference to Fig. 1. The information processing device 100 shown in Fig. 1 includes a communication device 110 that acquires data from multiple vehicles 200, which are so-called connected cars equipped with a constant connection function to a network; a data storage 120 that stores the data received by the communication device 110; and a computing device 10 that requests data necessary for machine learning using a neural network from the data storage 120, constructs an injury prediction ROM (reduced order model) that estimates injury risks for collision conditions such as the vehicle longitudinal speed, the other vehicle's speed, the collision angle (collision direction), and the contact position through machine learning based on the data acquired from the data storage 120, and estimates injury risks for collision conditions based on the constructed injury prediction ROM and data on actual automobile accidents acquired from the data storage 120 and transmits the estimated injury risks to an external device. The injury prediction ROM is a reduced-order model that significantly reduces analysis time and data volume by reducing the dimensions of the model while maintaining the essential behavior of the 3D model. Furthermore, collision conditions such as the longitudinal speed of the vehicle, the speed of the other vehicle, the collision angle (collision direction), and the contact position can take on various values ​​over time, and can therefore be described as linear quantities in the form of vectors or the like.

[0020] As will be described later, the data storage 120 is a data server equipped with a database, and the computing device 10 is a computer capable of executing advanced computing processes at high speed. Each of the data storage 120 and the computing device 10 may be a standalone server, or may be a cloud that can distribute the processing load. The data storage 120 and the computing device 10 may be the same server.

[0021] 2 is a block diagram showing the configuration of the vehicle 200. The vehicle 200 includes an airbag ECU (Electronic Control Unit). an image information processing unit 20 that calculates the behavior of the vehicle 200 from image information acquired by an imaging device 22; an input device 12 that receives input of information related to the behavior of the vehicle 200 calculated by the image information processing unit 20, the vehicle longitudinal speed detected by a vehicle speed sensor 24, the azimuth angle deviation and acceleration of the vehicle 200 detected by an IMU (Inertial Measurement Unit) 26, the steering angle of the vehicle 200 detected by a steering angle sensor 28, the throttle opening of the vehicle 200 detected by a throttle sensor, a collision of the vehicle 200 detected by a collision detection sensor 32, and information acquired by a V2X communication unit 34 via wireless communication; the control and arithmetic unit 14 that estimates the injury risk for collision conditions based on the input data input from the input device 12 and the data stored in the storage device 18; and an output device 16 that outputs the calculation results of the control and arithmetic unit 14 to the V2X communication unit 34.

[0022] As mentioned above, the vehicle 200 is a so-called connected car, but it does not have to be a connected car; the vehicle may be equipped with retrofitted communication equipment such as a translog that analyzes and utilizes driving data transmitted from onboard equipment installed in the vehicle 200, and various sensors that acquire driving data.

[0023] The imaging device 22 according to this embodiment is an in-vehicle camera or the like, and acquires image data of the periphery of the vehicle 200. It is desirable that the imaging devices 22 are provided on the front, rear, left and right sides of the vehicle 200, and be able to capture images of the entire circumference of the vehicle 200. The vehicle speed sensor 24 may be configured to be able to detect the speed of each of the four wheels of the vehicle 200. The imaging device 22, the vehicle speed sensor 24, the IMU 26, the steering angle sensor 28, the throttle sensor 30, and the collision detection sensor 32 constitute a group of sensors provided in the vehicle 200.

[0024] 3 is a block diagram showing an example of a specific configuration of the arithmetic device 10 according to an embodiment of the present invention. The arithmetic device 10 is configured to include a computer 40. The computer 40 includes a CPU (Central Processing Unit) 42, a ROM (Read Only Memory) 44, a RAM (Random Access Memory) 46, and an input / output port 48. As an example, it is desirable that the computer 40 be a model capable of executing advanced arithmetic processing at high speed.

[0025] In the computer 40, the CPU 42, ROM 44, RAM 46, and input / output port 48 are connected to one another via various buses such as an address bus, a data bus, and a control bus. The input / output port 48 is connected to various input / output devices, such as a display 50, a mouse 52, a keyboard 54, a hard disk (HDD) 56, and a disk drive 60 that reads information from various disks (e.g., CD-ROM, DVD, etc.) 58.

[0026] Furthermore, a network 62 is connected to the input / output port 48, enabling information to be exchanged with various devices connected to the network 62. In this embodiment, a data storage 120, which is a data server connected to a database (DB) 122, is connected to the network 62, enabling information to be exchanged with the DB 122.

[0027] DB 122 stores time-series data including data relating to accidents involving multiple vehicles 200 acquired via communication device 110. Data may be stored in DB 122 not only via communication device 110, but also by computer 40 or other devices connected to network 62.

[0028] In this embodiment, it is described that time series data of multiple vehicles 200 is stored in DB 122 connected to data storage 120, but the information in DB 122 may also be stored in an external storage device such as HDD 56 built into computer 40 or an external hard disk.

[0029] A program related to machine learning using a neural network is installed on the HDD 56 of the computer 40. In this embodiment, the CPU 42 executes the program to start machine learning and construct an injury prediction ROM, which is a trained machine learning model based on the machine learning. Furthermore, the constructed injury prediction ROM is used to estimate the injury risk for a collision condition. The CPU 42 also displays the processing results of the program on the display 50.

[0030] There are several ways to install the machine learning program of this embodiment into the computer 40. For example, the program can be stored on a CD-ROM, DVD, or the like together with a setup program, and the disk can be inserted into the disk drive 60, and the setup program can be executed by the CPU 42 to install the program into the HDD 46. Alternatively, the program can be installed into the HDD 46 by communicating with another information processing device connected to the computer 40 via a public telephone line or the network 62.

[0031] Next, various functions realized by the CPU 42 of the computing device 10 executing a machine learning program will be described. The machine learning program includes a simulation function for performing a collision simulation using FE (finite element) models of the vehicle and the human body, a learning function for constructing an injury prediction ROM by executing machine learning using calculation results from the simulation function as training data, an evaluation function for evaluating the performance of the injury prediction ROM, which is a trained model, using actual measured values ​​(measurement values) as training data, and a selection function for selecting an injury prediction ROM with excellent performance. By executing the machine learning program having these functions, the CPU 42 functions as a simulation unit, a learning unit, an evaluation unit, and a selection unit. In this embodiment, as an example of the AI ​​mathematical model used to construct the injury prediction ROM, a neural network, such as an RNN (Recurrent Neural Network), in which processing units that linearly transform inputs are connected in a network configuration is used.

[0032] The CPU 42, in which the learned injury prediction ROM has been constructed, has a data selection function that selects data relating to a collision from data transmitted from the vehicle 200, a collision condition calculation function that calculates collision conditions such as the vehicle longitudinal speed, the other vehicle's speed, the collision angle (collision direction), and the contact position from the selected data, and an injury risk estimation function that estimates the injury risk for the collision conditions using the learned injury prediction ROM. The CPU 42 executes a program related to collision detection that has each of these functions, so that the CPU 42 functions as a data selection unit, a collision condition calculation unit, and an injury risk estimation unit.

[0033] Figure 4 is a schematic diagram showing an example of constructing an injury prediction ROM that estimates injury risk. Figure 4(A) shows an example of a simulation in which a human body FE model 300 riding a bicycle FE model 310 collides with a vehicle FE model 200F. In the case shown in Figure 4(A), a parametric simulation is performed with variables such as the vehicle longitudinal speed, bicycle speed, the angle of bicycle FE model 310 relative to vehicle FE model 200F, and the contact position of bicycle FE model 310 on vehicle FE model 200F.

[0034] Fig. 4(B) is a schematic diagram showing an example of a collision. In Fig. 4(B), human body FE model 300 flies over hood 210 of vehicle FE model 200F due to the impact and crashes into windshield screen 220. The injury value of each part of human body FE model 300 is calculated from the calculation results of this parametric simulation. In this embodiment, the parts of human body FE model 300 are, for example, the head, chest, abdomen, arms, and legs. Each of the arms and legs may further be separated into left and right.

[0035] Figure 4(C) is a schematic diagram showing an example of the relationship between head contact position and head injury value in a parametric simulation. 15 indicates the Head Injury Criterion, and HIC 15 In FIG. 4(C), the head of the human body FE model 300 is in contact with the upper and left sides of the windshield screen of the vehicle FE model 200F, and the injury severity is higher. 15 indicates high values.

[0036] 4(D) is a schematic diagram showing an example of injury risk by body part in the parametric simulation of the human body FE model 300. In this embodiment, a mathematical model such as a neural network is trained using as training data the relationship between the collision conditions, such as the vehicle longitudinal speed, the other vehicle's speed, the collision angle (collision direction), and the contact position, and the injury value calculated by the parametric simulation, to construct an injury prediction ROM.

[0037] In Figure 4(D), for example, if collision condition (1) is the longitudinal speed of the vehicle and collision condition (2) is the speed of the other party, the risk of injury to the head, chest, abdomen, arms, and legs increases as the longitudinal speed of the vehicle and the speed of the other party increase.

[0038] Since the injury prediction ROM is a dimension-reduced model, in this embodiment, when constructing the model, for example, the collision conditions, which are variables, are limited. Various methods can be considered for limiting the collision conditions, but for example, collision conditions that are guaranteed to be linearly independent of each other are adopted as independent variables. This dimension reduction reduces the computational load of injury risk estimation.

[0039] 5 is a flowchart showing an example of a process for estimating an injury risk using the injury prediction ROM after learning. In step 100, for example, data storage 120 detects the occurrence of an accident from data on vehicle 200 acquired via communication device 110. The occurrence of an accident is detected, for example, based on the output of collision detection sensor 32 provided in vehicle 200. Alternatively, the occurrence of an accident may be detected from a change in the longitudinal speed of vehicle 200 detected by vehicle speed sensor 24, or a change in acceleration of vehicle 200 detected by IMU 26. The detection of the occurrence of an accident in step 100 may be performed by calculation device 10.

[0040] In step 102, data relating to the occurrence of an accident is transmitted to the calculation device 10. If the detection of the occurrence of an accident in step 100 is performed by the calculation device 10, the procedure in step 102 is unnecessary.

[0041] In step 104, the arithmetic device 10 calculates collision conditions from the acquired data. The calculated collision conditions include the longitudinal speed of the vehicle, the speed of the other vehicle, the collision angle (collision direction), and the contact position. The longitudinal speed of the vehicle is calculated from the output of the vehicle speed sensor 24 provided in the vehicle 200, and the speed, collision angle, and contact position of the other vehicle are calculated from image data of the other vehicle acquired by the imaging device 22. If the imaging devices 22 are provided on the front, rear, left, and right sides of the vehicle 200 and are capable of capturing images of the entire circumference of the vehicle 200, the speed, collision angle, and contact position of the other vehicle can be calculated from the image data acquired by the imaging device 22.

[0042] In step 106, the computing device 10 estimates an injury risk using the injury prediction ROM constructed by machine learning. The injury risk is estimated separately for the head, chest, abdomen, arms, and legs of the other party that collides with the vehicle 200.

[0043] In step 108, the calculation device 10 transmits information about the estimated injury risk to a medical institution, a fire department, etc. from the input / output port 48 via the network 62. The medical institution, etc. that receives the injury risk information from the calculation device 10 arranges a medical system and an ambulance (or a doctor helicopter, etc.) according to the injury risk. After transmitting the injury risk in step 108, the information processing device 100 according to this embodiment ends the injury risk estimation process.

[0044] As explained above, this embodiment uses an injury prediction ROM with reduced dimensions of a mathematical model trained on the results of collision simulation obtained by parametric simulation, making it possible to quickly estimate the injury level of an injured person's entire body from collision conditions obtained at the time of an accident, such as the vehicle's longitudinal speed, the other vehicle's speed, the collision angle (collision direction), and the contact position.

[0045] In some cases, data from actual accidents was used to train a mathematical model that estimates the level of injury to the entire body of an injured person, but there was a lack of data recording the details of the accident, including the injuries to the entire body of the injured person, making it difficult to obtain an appropriate predictive model using machine learning.

[0046] In this embodiment, the results of parametric simulation are used as training data for learning the mathematical model, so that a trained model can be constructed that can estimate the injury level of an injured person's entire body.

[0047] Furthermore, as mentioned above, the trained model used to estimate the injury level of an injured person is an injury prediction ROM with reduced dimensions, which enables faster processing speeds and enables the injury level of the injured person's entire body to be estimated in a short period of time.

[0048] In this embodiment, the injured person is primarily assumed to be a person who collides with the vehicle 200, but is not limited to this. A parametric simulation may be performed assuming a collision between two vehicles 200, including whether or not the occupants of the vehicle 200 are wearing seat belts and whether or not there is a multiple collision as collision conditions, to construct an injury prediction ROM capable of estimating the injury risk of the occupants of the vehicle 200. As an example, whether or not the seat belts are worn is detected by a seat belt sensor, and whether or not there is a multiple collision is detected based on image data acquired by the imaging device 22 and the output of the IMU 26.

[0049] In the above embodiments, the processing performed by the CPU after reading the software (program) may be performed by various processors other than the CPU. Examples of such processors include programmable logic devices (PLDs) whose circuit configuration can be changed after fabrication, such as field-programmable gate arrays (FPGAs), and dedicated electrical circuits, such as application-specific integrated circuits (ASICs), which are processors with circuit configurations specifically designed to perform specific processing. The processing may be performed by one of these processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA). The hardware structure of these processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

[0050] In addition, in each of the above embodiments, the program is described as being stored (installed) in advance in the disk drive 60 or the like, but this is not limiting. The program may be provided in a form stored in a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. The program may also be downloaded from an external device via a network.

[0051] The "acquisition unit" described in the claims corresponds to the "data storage 120" described in the detailed description of the invention, and the "calculation unit" described in the claims corresponds to the "injury risk estimation unit" described in the detailed description of the invention.

[0052] (Additional note 1) Memory and at least one processor coupled to said memory; Including, The processor: A reduced-dimensional model is constructed by performing machine learning based on the collision conditions during a vehicle accident and the results of a collision simulation using a vehicle model. Obtaining collision conditions detected by a group of sensors equipped in the vehicle, inputting the acquired collision conditions into the trained dimensionally reduced model to estimate injury risks at each part of the injured person; An information processing device configured as follows. [Explanation of symbols]

[0053] 10 Arithmetic unit 12 Input Devices 14 Control and arithmetic unit 16 Output Devices 18 Storage device 20 Image information processing section 22 Imaging device 24 Vehicle speed sensor 26 IMU 28 Steering angle sensor 32 Collision detection sensor 34 V2X communication unit 40 Computer 42 CPU 44 ROM 46 RAM 48 input / output ports 50 displays 52 Mouse 54 keyboard 60 disk drives 62 Network 100 Information processing device 110 Communication equipment 120 Data Storage 200 vehicles 200F vehicle FE model 300 human body FE models 310 Bicycle FE Model

Claims

1. a learning unit that constructs a dimension-reduced model for estimating injury risks at each part of an injured person from the collision conditions by performing machine learning while limiting the collision conditions, which are variables, based on collision conditions in a collision simulation using a vehicle model and a human body model and results of the collision simulation including injury values ​​at each part of the human body model obtained in the collision simulation; and a calculation unit that calculates collision conditions of a vehicle accident from information detected by a group of sensors provided in the vehicle; a calculation unit that inputs the collision conditions of the vehicle accident calculated by the calculation unit into the trained dimensionally reduced model and estimates injury risks for each body part of a person injured in the vehicle accident; An information processing device comprising:

2. The information processing device according to claim 1 , wherein the group of sensors detects information including a direction of collision between the vehicle and the injured person, a degree of collision, and a contact position of the injured person with the vehicle.

3. The information processing apparatus according to claim 1 , wherein the learning unit performs machine learning by limiting the collision conditions that are variables by adopting collision conditions that are guaranteed to be linearly independent of each other as independent variables.

4. A process of constructing a dimension-reduced model for estimating injury risks at each part of an injured person from collision conditions by performing machine learning while limiting collision conditions, which are variables, based on collision conditions in a collision simulation using a vehicle model and a human body model and results of the collision simulation including injury values ​​at each part of the human body model obtained in the collision simulation; calculating collision conditions of a vehicle accident from information detected by a group of sensors equipped in the vehicle; a step of inputting the calculated collision conditions of the vehicle accident into the trained dimensional reduction model and estimating injury risks for each body part of a person injured in the vehicle accident; An information processing method in which a computer executes a process including the steps of:

5. Computer, An information processing program comprising: a learning unit that constructs a dimension-reduced model for estimating injury risks at each body part of an injured person from collision conditions by performing machine learning while limiting collision conditions, which are variables, based on collision conditions in a collision simulation using a vehicle model and a human body model and results of the collision simulation including injury values ​​for each body part of the human body model obtained in the collision simulation; a calculation unit that calculates collision conditions of a vehicle accident from information detected by a group of sensors equipped in the vehicle; and an information processing program that functions as an operation unit that inputs the collision conditions of a vehicle accident calculated by the calculation unit into the learned dimension-reduced model and estimates injury risks at each body part of an injured person in the vehicle accident.

Images