Driving skills evaluation method, driving skills evaluation system, and recording medium

Abstract

A driving skill evaluation method according to an embodiment of the present disclosure includes: generating a kernel density estimation image on the basis of time series data of a first parameter corresponding to directional change in the direction of travel of the vehicle, and time series data of a second parameter indicating the square of jerk in the direction of travel of the vehicle; and evaluating the driving skill of the driver of the vehicle by comparing the kernel density estimation image with a reference image.

Description

Technical Field

[0004] , , ,

[0005] , By calculating image similarity and evaluating whether the driving skills of a vehicle driver are similar to those of a skilled driver based on image similarity, , ,

[0001] The present disclosure relates to a driving skill evaluation method and a driving skill evaluation system for evaluating a driver's driving skill, and a recording medium on which software for evaluating a driver's driving skill is recorded.

Background Art

[0002] In recent years, technologies for evaluating a driver's driving skill have been developed in vehicles such as automobiles. For example, Patent Document 1 discloses a technology for evaluating a driver's driving skill based on longitudinal acceleration and lateral acceleration when a vehicle turns.

Prior Art Documents

Patent Documents

[0003] <​​​​​​​​​​​​​​​​​​A driving skill evaluation system according to one embodiment of the present disclosure comprises an image generation circuit and an evaluation circuit. The image generation circuit generates a kernel density estimation image based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter indicating the square of the jerk in the direction of travel of the vehicle. The evaluation circuit uses the kernel density estimation image to evaluate the kernel density estimation image. skilled By comparing with a reference image obtained based on the driver's driving data By calculating image similarity and evaluating whether the driving skills of a vehicle driver are similar to those of a skilled driver based on image similarity, This evaluates the driving skills of a vehicle driver. The first parameter is one of the following: a parameter related to yaw angular velocity, a parameter related to yaw angular acceleration, a parameter related to the square of yaw angular acceleration, a parameter related to lateral acceleration, or a parameter related to the square of lateral acceleration.

[0006] A recording medium according to one embodiment of the present disclosure generates a kernel density estimation image based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter indicating the square of the jerk in the direction of travel of the vehicle, and the kernel density estimation image skilled By comparing with a reference image obtained based on the driver's driving data By calculating image similarity and evaluating whether the driving skills of a vehicle driver are similar to those of a skilled driver based on image similarity, This is software recorded that causes a processor to evaluate the driving skills of a vehicle driver. The first parameter is one of the following: a parameter relating to yaw angular velocity, a parameter relating to yaw angular acceleration, a parameter relating to the square of yaw angular acceleration, a parameter relating to lateral acceleration, or a parameter relating to the square of lateral acceleration. [Brief explanation of the drawing]

[0007] The accompanying drawings are provided for further understanding of this disclosure and are incorporated herein and constitute part of this specification. The drawings illustrate one embodiment and, together with the specification, serve to illustrate the principles of this disclosure.

[0008] [Figure 1]This is an explanatory diagram showing an example configuration of a driving skills evaluation system in which the driving skills evaluation method according to one embodiment of this disclosure is used. [Figure 2] Figure 1 is a block diagram showing one example of a smartphone configuration. [Figure 3] Figure 1 is a block diagram showing one example configuration of the server device 30. [Figure 4] Figure 3 is an explanatory diagram showing the data stored in the memory unit 32. [Figure 5] Figure 4 is an explanatory diagram illustrating an example of yaw angular velocity data and curve data. [Figure 6] Figure 4 shows an example of a kernel density estimation image, which is represented by the image data shown. [Figure 7] Figure 6 is an explanatory diagram illustrating an example of parameters in the kernel density estimation image. [Figure 8] Figure 1 is an explanatory diagram illustrating one example configuration of the data processing system shown. [Figure 9] Figure 8 is a block diagram showing one example configuration of the in-vehicle device. [Figure 10] Figure 8 is a block diagram showing one example configuration of the information processing device. [Figure 11] Figure 8 is a flowchart illustrating an example of the operation of the information processing device shown. [Figure 12] This is another flowchart illustrating an example of the operation of the information processing device shown in Figure 8. [Figure 13] This is another flowchart illustrating an example of the operation of the information processing device shown in Figure 8. [Figure 14] This is another flowchart illustrating an example of the operation of the information processing device shown in Figure 8. [Figure 15] Figure 14 is an explanatory diagram illustrating an example of the process for generating the pre-processed image shown. [Figure 16] This is another flowchart illustrating an example of the operation of the information processing device shown in Figure 8. [Figure 17] Figure 8 is an explanatory diagram illustrating an example of the operation of the information processing device shown. [Figure 18]It is another flowchart showing an operation example of the information processing apparatus shown in FIG. 8. [Figure 19A] It is a flowchart showing an operation example of the server apparatus shown in FIG. 1. [Figure 19B] It is another flowchart showing an operation example of the server apparatus shown in FIG. 1. [Figure 20] It is an explanatory diagram showing an example of time-series data of longitudinal and lateral accelerations. [Figure 21] It is an explanatory diagram showing an example of parameters in the kernel density estimation image according to a modification example. [Figure 22] It is an explanatory diagram showing an example of parameters in the kernel density estimation image according to another modification example. [Figure 23] It is an explanatory diagram showing an example of parameters in the kernel density estimation image according to another modification example. [Figure 24] It is an explanatory diagram showing an example of parameters in the kernel density estimation image according to another modification example. [Figure 25] It is an image diagram showing an example of the kernel density estimation image according to another modification example. [Figure 26] It is an explanatory diagram showing an example of parameters in the kernel density estimation image shown in FIG. 25. [Figure 27] It is a block diagram showing a configuration example of a smartphone according to another modification example.

Embodiments for Carrying Out the Invention

[0009] In the evaluation of a driver's driving skills, it is desired that the evaluation accuracy be high, and further improvement in the evaluation accuracy is expected.

[0010] It is desirable to provide a driving skill evaluation method, a driving skill evaluation system, and a recording medium that can enhance the evaluation accuracy of a driver's driving skills.

[0011] Hereinafter, several exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. The following description is intended to illustrate specific examples of the present disclosure and should not be construed as limiting the disclosure. For example, elements such as numerical values, shapes, materials, parts, the location of each part, and the method of connecting each part are merely examples and should not be construed as limiting the disclosure. Furthermore, in the following exemplary embodiments, components not described in separate sections based on the highest-level concepts of the present disclosure are optional and may be provided as needed. The drawings are schematic and are not intended to be to scale. Throughout this specification and the drawings, components having substantially the same function and substantially the same configuration are denoted by the same reference numerals, and redundant descriptions are omitted. Furthermore, components not directly related to an embodiment of the present disclosure are not shown in the drawings.

[0012] <Embodiment> [Example Configuration] Figure 1 shows an example configuration of a driving skills evaluation system 1 that uses a driving skills evaluation method according to one embodiment. The driving skills evaluation system 1 comprises a smartphone 10, a server device 30, and a data processing system 2.

[0013] Smartphone 10 is a high-function mobile phone and is fixedly installed inside vehicle 9 in a predetermined orientation relative to vehicle 9. Smartphone 10 collects driving data of vehicle 9. Smartphone 10 connects to the internet (not shown) by communicating with a mobile phone base station (not shown) via, for example, mobile phone communication.

[0014] The server device 30 is an information processing device. The server device 30 evaluates the driving skills of the vehicle 9's driver based on the vehicle's driving data. The server device 30 is connected to the internet (not shown). The server device 30 is capable of communicating with the smartphone 10 via the internet.

[0015] The data processing system 2 includes an information processing device and generates data used in the evaluation of driving skills. The data processing system 2 is connected to the Internet (not shown). The data processing system 2 is capable of communicating with the server device 30 via the Internet.

[0016] In the driving skills evaluation system 1, for example, in an evaluation area with many curves, including mountain roads, the driver being evaluated drives the vehicle 9, and the smartphone 10 collects driving data of the vehicle 9 and transmits this driving data to the server device 30. This driving data includes information about the acceleration of the vehicle 9 in the direction of travel (longitudinal acceleration), information about the yaw angular velocity of the vehicle 9, and information about the position of the vehicle 9. Based on the time-series data of the yaw angular velocity of the vehicle 9, the server device 30 detects multiple curves in the road traveled by the vehicle 9. Then, based on the time-series data of longitudinal acceleration and the time-series data of yaw angular velocity, the server device 30 generates kernel density estimation images for each of the multiple curves. Then, for each of the multiple curves, the server device 30 evaluates the driver's driving skills by comparing the kernel density estimation image generated by the server device 30 with kernel density estimation images for skilled drivers that are generated by the data processing system 2 and registered in advance on the server device 30. The smartphone 10 then presents the driver with the evaluation results of their driving skills. This allows the driver to obtain an objective evaluation of their driving skills using the driving skills evaluation system 1.

[0017] Figure 2 shows an example configuration of a smartphone 10. The smartphone 10 includes a touch panel 11, a memory unit 12, a communication unit 13, an acceleration sensor 14, an angular velocity sensor 15, a GNSS (Global Navigation Satellite System) receiver 16, and a processing unit 20.

[0018] The touch panel 11 is a user interface and comprises, for example, a touch sensor and a display device such as a liquid crystal display or an organic EL (Electro-Luminescence) display. The touch panel 11 accepts user input from the smartphone 10 and displays the processing results of the smartphone 10.

[0019] The memory unit 12 is a non-volatile memory and is configured to store program data for various application software. In this example, the smartphone 10 has application software related to the driving skills evaluation system 1 installed. The program data for this application software is stored in the memory unit 12.

[0020] The communication unit 13 is configured to communicate with a mobile phone base station by performing mobile phone communication. As a result, the communication unit 13 communicates with a server device 30 connected to the internet via the mobile phone base station.

[0021] The acceleration sensor 14 is configured to detect acceleration in three directions within the coordinate system of the smartphone 10.

[0022] The angular velocity sensor 15 is configured to detect three angular velocities (yaw angular velocity, roll angular velocity, and pitch angular velocity) in the coordinate system of the smartphone 10.

[0023] The GNSS receiver 16 is configured to acquire the position of the vehicle 9 on the ground using a GNSS such as GPS (Global Positioning System).

[0024] The processing unit 20 is configured to control the operation of the smartphone 10 and is composed of, for example, one or more processors, one or more memories, etc. The processing unit 20 collects time-series data of acceleration detected by the acceleration sensor 14, time-series data of angular velocity detected by the angular velocity sensor 15, and time-series data of the position of the vehicle 9 obtained by the GNSS receiver 16. The processing unit 20 can operate as a data processing unit 21 and a display processing unit 22 by executing application software related to the driving skill evaluation system 1 installed on the smartphone 10.

[0025] The data processing unit 21 is configured to perform predetermined data processing based on the detection results of the acceleration sensor 14 and the angular velocity sensor 15. The predetermined data processing includes, for example, filtering the time-series data of acceleration detected by the acceleration sensor 14 and filtering the time-series data of angular velocity detected by the angular velocity sensor 15. Here, the filtering is performed using a low-pass filter. After the vehicle has finished driving, the communication unit 13 transmits the time-series data of acceleration and angular velocity processed by the data processing unit 21, along with the time-series data of the vehicle 9's position obtained by the GNSS receiver 16, to the server device 30.

[0026] The display processing unit 22 is configured to perform display processing based on data indicating the driving skill evaluation results transmitted from the server device 30. As a result, the touch panel 11 displays the driving skill evaluation results.

[0027] Figure 3 shows an example configuration of the server device 30. The server device 30 includes a communication unit 31, a storage unit 32, and a processing unit 40.

[0028] The communication unit 31 is configured to communicate with the smartphone 10 via the internet by performing network communication.

[0029] The storage unit 32 includes, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive), and is configured to store program data for various software. In this example, the server device 30 has server software related to the driving skills evaluation system 1 installed. The program data for this software is stored in the storage unit 32. The storage unit 32 also stores various data used by this software.

[0030] Figure 4 shows an example of data used by the server software, stored in the storage unit 32. The storage unit 32 stores area data DAR, multiple datasets DS (dataset DSA and multiple datasets DSB), and evaluation target curve data DTC. This data is generated by the data processing system 2 and stored in this storage unit 32.

[0031] Area data DAR is data that indicates the evaluation area for assessing driving skills. The evaluation area can be defined as an area with many curves, such as a mountain road. This area data DAR includes, for example, data on the latitude and longitude of this evaluation area.

[0032] Each of the multiple datasets DS is data corresponding to driving data obtained by a skilled driver driving a vehicle in the area indicated by the area data DAR. The multiple datasets DS may contain data corresponding to multiple driving data from different skilled drivers, or they may contain data corresponding to multiple driving data from a single skilled driver. Each of the multiple datasets DS includes acceleration data DA, yaw angular velocity data DY, curve data DC, and multiple image data DP.

[0033] Acceleration data DA is time-series data of the acceleration in the direction of travel (longitudinal acceleration) of a vehicle driven by a skilled driver.

[0034] The yaw angular velocity data DY is time-series data of the yaw angular velocity of a vehicle driven by a skilled driver.

[0035] Curve data DC is data containing curve numbers for multiple curves on a road. Curve data DC is generated based on yaw angular velocity data DY. In curve data DC, the curve numbers of multiple curves are set in correspondence with the time-series data of yaw angular velocity in yaw angular velocity data DY.

[0036] Figure 5 shows an example of yaw angular velocity data DY and curve data DC. The yaw angular velocity changes according to the curve of the road. The curve data DC includes curve numbers ("1" to "9" in Figure 5) that are set based on this time-series data of yaw angular velocity. In the curve data DC, the curve numbers of multiple curves are set in correspondence with the time-series data of yaw angular velocity in the yaw angular velocity data DY.

[0037] Multiple image data DPs are image data of kernel density estimation images at multiple curves.

[0038] Figure 6 shows an example of a kernel density estimation image for a certain curve. Figure 7 shows the coordinate axes of the kernel density estimation image shown in Figure 6. In this kernel density estimation image, the horizontal axis (X axis) represents time, and the vertical axis (Y axis) represents the square of the yaw angular acceleration. The yaw angular acceleration is the time derivative of the yaw angular velocity. The pixel value (Z axis) of the kernel density estimation image represents the square of the longitudinal jerk. The longitudinal jerk is the time derivative of the longitudinal acceleration. In the kernel density estimation image, darker colored areas indicate a large value of the square of the longitudinal jerk, and lighter colored areas indicate a small value of the square of the longitudinal jerk. In this example, a larger value of the square of the longitudinal jerk corresponds to a smaller pixel value, and a smaller value of the square of the longitudinal jerk corresponds to a larger pixel value. This kernel density estimation image may change depending on the driver's driving skill. Multiple kernel density estimation images corresponding to multiple curves are stored in the storage unit 32 as multiple image data DPs.

[0039] Each of the dataset DSA and multiple dataset DSBs thus contains acceleration data DA, yaw angular velocity data DY, curve data DC, and multiple image data DP. As described later, the data processing system 2 adjusts the curve numbers of the curve data DC in each of the multiple dataset DSBs based on the curve data DC in dataset DSA. That is, since the curve numbers are generated based on the yaw angular velocity data DY, different curve numbers may be assigned to different curves depending on the yaw angular velocity data DY. Therefore, the data processing system 2 uses dataset DSA as sample data and adjusts the curve numbers of the curve data DC in each of the multiple dataset DSBs based on the curve data DC in dataset DSA. This ensures that the curve numbers of the same curves in the curve data DC of multiple dataset DSBs are adjusted to be the same.

[0040] The evaluation target curve data DTC is data that shows the curve numbers of the multiple curves that are the subject of evaluation for driving skills, out of the multiple curves in the area indicated by the area data DAR.

[0041] Such data is stored in the storage unit 32. Although the above explanation describes an example where data for one area is stored in the storage unit 32, it is not limited to this; data for multiple areas may also be stored. In this case, the storage unit 32 stores multiple datasets DS, evaluation target curve data DTC, and area data DAR for each of the multiple areas.

[0042] The processing unit 40 (Figure 3) is configured to control the operation of the server device 30 and is composed of, for example, one or more processors, one or more memories, etc. By executing the server software related to the driving skill evaluation system 1 installed on the server device 30, the processing unit 40 can operate as a data processing unit 41, a curve detection unit 42, a data extraction unit 43, an image generation unit 44, an image similarity calculation unit 45, and a skill determination unit 46.

[0043] The data processing unit 41 is configured to generate acceleration data DA1 and yaw angular velocity data DY1 by performing predetermined data processing based on the time-series data of acceleration, time-series data of angular velocity, and time-series data of the position of the vehicle 9 received by the communication unit 31. The predetermined data processing includes, for example, a process to confirm whether the vehicle 9 is traveling in the evaluation area based on the time-series data of the position of the vehicle 9; a process to generate time-series data of acceleration in the direction of travel (longitudinal acceleration) of the vehicle 9 by performing coordinate transformation based on the time-series data of acceleration obtained by the smartphone 10; a process to generate time-series data of the yaw angular velocity of the vehicle 9 by performing coordinate transformation based on the time-series data of angular velocity obtained by the smartphone 10; filtering of the time-series data of longitudinal acceleration; and filtering of the time-series data of yaw angular velocity. Here, the filtering is performed using a low-pass filter.

[0044] The curve detection unit 42 is configured to generate curve data DC1 by detecting multiple curves based on the yaw angular velocity data DY1 generated by the data processing unit 41.

[0045] The data extraction unit 43 is configured to extract time-series data of longitudinal acceleration related to multiple curves that are the subject of evaluation for driving skills from among the time-series data of longitudinal acceleration included in acceleration data DA1, based on the evaluation target curve data DTC stored in storage unit 32, and to extract time-series data of yaw angular velocity related to multiple curves that are the subject of evaluation for driving skills from among the time-series data of yaw angular velocity included in yaw angular velocity data DY1.

[0046] The image generation unit 44 is configured to generate multiple image data DP1 by generating multiple kernel density estimation images for multiple curves, based on time-series data of longitudinal acceleration and yaw angular velocity for multiple curves extracted by the data extraction unit 43. Specifically, the image generation unit 44 generates a kernel density estimation image for a curve by performing kernel density estimation processing based on time-series data of longitudinal acceleration and yaw angular velocity for one curve. In the kernel density estimation processing, the original data, including data that has not yet been observed, is estimated as density data based on measured data. The image generation unit 44 generates multiple kernel density estimation images by performing this processing for each of the multiple curves. In this way, the image generation unit 44 generates multiple image data DP1 for multiple curves.

[0047] The image similarity calculation unit 45 is configured to calculate the average value of image similarity (average similarity) based on multiple kernel density estimation images generated by the image generation unit 44 and multiple kernel density estimation images included in multiple datasets DS stored in the storage unit 32. Specifically, the image similarity calculation unit 45 calculates the image similarity for each curve by comparing the kernel density estimation image generated by the image generation unit 44 with multiple kernel density estimation images included in multiple datasets DS. The image similarity calculation unit 45 calculates multiple image similarities by performing this process for each of the multiple curves. The image similarity calculation unit 45 then calculates the average value (average similarity) of these multiple image similarities.

[0048] The skill determination unit 46 is configured to determine the driving skills of the vehicle 9 driver based on the average similarity calculated by the image similarity calculation unit 45. The communication unit 31 then transmits data indicating the driving skill evaluation result generated by the skill determination unit 46 to the smartphone 10.

[0049] Multiple datasets DS, evaluation target curve data DTC, and area data DAR stored in the memory unit 32 are generated by the data processing system 2. The data processing system 2 is described below.

[0050] Figure 8 shows an example configuration of the data processing system 2. The data processing system 2 comprises an in-vehicle device 110 and an information processing device 130. The in-vehicle device 110 is a device installed in a vehicle 109 driven by a skilled driver. In this example, the information processing device 130 is a so-called personal computer.

[0051] Figure 9 shows an example configuration of the in-vehicle device 110. The in-vehicle device 110 includes an acceleration sensor 114, a yaw angular velocity sensor 115, a GNSS receiver 116, and a processing unit 120.

[0052] The acceleration sensor 114 is configured to detect the acceleration (longitudinal acceleration) of the vehicle 109 in the direction of travel.

[0053] The yaw angular velocity sensor 115 is configured to detect the yaw angular velocity of the vehicle 109.

[0054] The GNSS receiver 116 is configured to acquire the position of the vehicle 109 on the ground using a GNSS such as GPS.

[0055] The processing unit 120 is a so-called ECU (Electronic Control Unit) and is composed of, for example, one or more processors, one or more memories, etc. The processing unit 120 collects time-series data of longitudinal acceleration detected by the acceleration sensor 114, time-series data of yaw angular velocity detected by the yaw angular velocity sensor 115, and time-series data of the position of the vehicle 109 obtained by the GNSS receiver 116.

[0056] For example, after the run is finished, the engineer stores the time-series data collected by the processing unit 120 into an external recording medium such as a semiconductor memory.

[0057] Figure 10 shows an example configuration of the information processing device 130. Based on the engineer's operation, the information processing device 130 reads data recorded on an external recording medium and generates multiple data sets DS, evaluation target curve data DTC, and area data DAR to be stored in the storage unit 32 of the server device 30 based on the read data. The information processing device 130 comprises a user interface unit 131, a storage unit 132, a communication unit 133, and a processing unit 140.

[0058] The user interface unit 131 includes, for example, a keyboard, a mouse, and a display unit such as a liquid crystal display or an organic EL display. The user interface unit 131 accepts operations from the user of the information processing device 130 (an engineer in this example) and displays the processing results of the information processing device 130.

[0059] The storage unit 132 includes, for example, an HDD or SSD, and is configured to store program data for various software. In this example, the information processing device 130 has software related to the data processing system 2 installed. The program data for this software is stored in the storage unit 132.

[0060] The communication unit 133 is configured to communicate with the server device 30 via the internet by performing network communication.

[0061] The processing unit 140 is configured to control the operation of the information processing device 130 and is configured, for example, using one or more processors, one or more memories, etc. By executing software related to the data processing system 2 installed on the information processing device 130, the processing unit 140 can operate as a data processing unit 141, a curve detection unit 142, an image generation unit 144, an image similarity calculation unit 145, an evaluation target data generation unit 147, an area data generation unit 148, and a data registration unit 149.

[0062] The data processing unit 141 is configured to generate acceleration data DA and yaw angular velocity data DY by performing predetermined data processing based on time-series data of longitudinal acceleration and time-series data of yaw angular velocity read from an external recording medium. The predetermined data processing includes, for example, filtering the time-series data of longitudinal acceleration, filtering the time-series data of yaw angular velocity, and processing to generate area data DAR based on the engineer's operations. Here, the filtering is performed using a low-pass filter.

[0063] The curve detection unit 142 is configured to generate curve data DC by detecting multiple curves based on the yaw angular velocity data DY generated by the data processing unit 141. The processing of the curve detection unit 142 is the same as the processing of the curve detection unit 42 in the server device 30.

[0064] The image generation unit 144 is configured to generate multiple image data DPs by generating multiple kernel density estimation images for multiple curves based on time-series data of longitudinal acceleration and time-series data of yaw angular velocity for multiple curves. The processing of the image generation unit 144 is the same as the processing of the image generation unit 44 in the server device 30.

[0065] The image similarity calculation unit 145 is configured to calculate image similarity based on multiple kernel density estimated images. The image similarity calculation process in the image similarity calculation unit 145 is the same as the image similarity calculation process in the image similarity calculation unit 45 of the server device 30.

[0066] The evaluation target data generation unit 147 is configured to generate evaluation target curve data DTC by determining multiple curves that are the target of evaluation for driving skills based on the processing results of the image similarity calculation unit 145.

[0067] The area data generation unit 148 is configured to generate area data DAR, which is data indicating the area to be evaluated, based on the engineer's operations.

[0068] The data registration unit 149 is configured to store a dataset DS, which includes acceleration data DA, yaw angular velocity data DY, curve data DC, and multiple image data DP, as well as evaluation target curve data DTC and area data DAR, in the storage unit 132.

[0069] In this configuration, the data processing system 2 generates multiple datasets DS, evaluation target curve data DTC, and area data DAR based on the driving data of skilled drivers. The data processing system 2 then transmits the multiple datasets DS, evaluation target curve data DTC, and area data DAR to the server device 30. The server device 30 then stores this data in the storage unit 32.

[0070] Here, the image generation unit 44 corresponds to one specific example of the "image generation circuit" in this disclosure. The image similarity calculation unit 45 and the skill determination unit 46 correspond to one specific example of the "evaluation circuit" in this disclosure. The square of the yaw angular acceleration corresponds to one specific example of the "first parameter" in this disclosure. The square of the forward and backward jerk corresponds to one specific example of the "second parameter" in this disclosure. The kernel density estimation image shown by the image data DP of the dataset DS stored in the storage unit 32 corresponds to one specific example of the "reference image" in this disclosure.

[0071] [Action and function] Next, the operation and function of the driving skills evaluation system 1 of this embodiment will be described.

[0072] (Overview of overall operation) Refer to Figures 1-4 and 8-10 to explain the operation of the driving skills evaluation system 1.

[0073] First, the driving skills evaluation system 1 generates multiple datasets DS, evaluation curve data DTC, and area data DAR based on driving data when a skilled driver drives vehicle 109.

[0074] Specifically, when a skilled driver operates the vehicle 109 of the data processing system 2, the processing unit 120 of the in-vehicle device 110 collects driving data from the vehicle 109. For example, after the driving is completed, the engineer stores the driving data collected by the processing unit 120 in an external recording medium such as a semiconductor memory. Based on this driving data, the information processing device 130 generates acceleration data DA, yaw angular velocity data DY, curve data DC, and multiple image data DP, and stores a dataset DS containing this data in the storage unit 132. The information processing device 130 stores multiple datasets DS in the storage unit 132 by repeating this process. The information processing device 130 also generates evaluation target curve data DTC, which is data indicating multiple curves that are the subject of driving skill evaluation, based on the multiple datasets DS, and generates area data DAR, which is data indicating the evaluation target area, based on the engineer's operations, and stores this data in the storage unit 132. The information processing device 130 then transmits the multiple datasets DS, evaluation target curve data DTC, and area data DAR to the server device 30. The server device 30 stores multiple datasets DS, evaluation target curve data DTC, and area data DAR transmitted from the information processing device 130 in the storage unit 32.

[0075] The driving skills evaluation system 1 evaluates the driver's driving skills using multiple datasets DS, evaluation curve data DTC, and area data DAR, which are generated when a skilled driver operates the vehicle 109 in this manner.

[0076] Specifically, the smartphone 10 collects driving data from the vehicle 9 as the driver being evaluated drives the vehicle 9. The smartphone 10 then transmits the collected driving data to the server device 30. Based on this driving data, the server device 30 generates acceleration data DA1, yaw angular velocity data DY1, curve data DC1, and multiple image data DP1. Based on the multiple kernel density estimation images shown by the multiple image data DP1 and the multiple kernel density estimation images shown by the multiple image data DP included in the multiple datasets DS stored in the storage unit 32, the server device 30 calculates the average similarity of the images (average similarity), and determines the driving skills of the vehicle 9 driver based on this average similarity. The server device 30 then transmits data showing the evaluation result of the driving skills to the smartphone 10. The smartphone 10 then displays the driving skill evaluation result transmitted from the server device 30.

[0077] (Detailed operation) The operation of the driving skills evaluation system 1 is described in detail below.

[0078] (Regarding the generation of multiple datasets / datasets) First, the data processing system 2 generates multiple datasets DS based on driving data obtained by a skilled driver operating vehicle 109. The operation of generating multiple datasets DS is described in detail below.

[0079] When a skilled driver operates vehicle 109, the vehicle 109's onboard device 110 detects the vehicle's acceleration in the direction of travel (longitudinal acceleration) using its acceleration sensor 114, detects the vehicle's yaw angular velocity using its yaw angular velocity sensor 115, and acquires the vehicle's position on the ground using its GNSS receiver 116. The processing unit 120 collects time-series data of longitudinal acceleration detected by the acceleration sensor 114, time-series data of yaw angular velocity detected by the yaw angular velocity sensor 115, and time-series data of the vehicle's position obtained by the GNSS receiver 116.

[0080] For example, after the run is finished, the engineer stores the time-series data collected by the processing unit 120 into an external recording medium such as a semiconductor memory.

[0081] The information processing device 130 generates a dataset DS based on the time-series data of longitudinal acceleration, time-series data of yaw angular velocity, and time-series data of the position of the vehicle 109, which are read from the external recording medium, and stores this dataset DS in the storage unit 132. The operation of this information processing device 130 will be described in detail below.

[0082] Figure 11 shows an example of the operation of the information processing device 130 in the data processing system 2. In this example, the area data generation unit 148 has already generated the area data DAR based on the engineer's operation.

[0083] First, the data processing unit 141 of the information processing device 130 uses the area data DAR stored in the storage unit 132 to check whether the vehicle 109 has traveled through the evaluation area based on the time-series data of the vehicle's position (step S101). If the vehicle 109 has not traveled through the evaluation area ("N" in step S101), this process ends.

[0084] When vehicle 109 travels through the evaluation area ("Y" in step S101), the data processing unit 141 performs filtering on the time-series data of longitudinal acceleration and yaw angular velocity in the evaluation area to generate acceleration data DA and yaw angular velocity data DY, respectively (step S102).

[0085] Next, the curve detection unit 142 generates curve data DC by performing a curve division process based on the yaw angular velocity data DY generated in step S102 (step S103). The curve division process includes the following two-stage process.

[0086] Figure 12 shows a specific example of the first stage of the curve splitting process. The curve detection unit 142 reads the yaw angular velocity included in the yaw angular velocity data DY sequentially in chronological order and performs the processing shown in Figure 12.

[0087] First, the curve detection unit 142 checks whether a yaw angular velocity of a predetermined value A or greater (for example, 0.02 rad / sec. or greater) continues for a predetermined time B or longer (for example, 2 seconds or longer) (step S201). This condition is the basic condition for curve detection by the curve detection unit 142. In this example, a predetermined time B was used for evaluation, but a predetermined distance may be used instead of a predetermined time B. If this condition is not met ("N" in step S201), the process in step S201 is repeated until this condition is met.

[0088] If the conditions shown in step S201 are met (in step S201, "Y"), the curve detection unit 142 checks whether the polarity of the yaw angular velocity is the same as the polarity of the yaw angular velocity at the previous curve, and whether the vehicle 109 has been traveling on a straight road for less than a predetermined time C (for example, less than 9 seconds) after the previous curve (step S202). In this example, a predetermined time C was used for evaluation, but a predetermined distance may be used instead of a predetermined time C.

[0089] In step S202, if this condition is not met ("N" in step S202), the curve detection unit 142 detects a curve (step S203). That is, in step S201, the basic conditions for curve detection are met, and the distance from the previous curve is large, so the curve detection unit 142 detects a new curve separately from the previous curve. The curve detection unit 142 assigns a curve number to the detected curve. Then the process proceeds to step S205.

[0090] In step S202, if this condition is met ("Y" in step S202), the curve detection unit 142 considers the previous curve to be a continuation (step S203). That is, in step S201, the basic conditions for curve detection are met, but because the distance from the previous curve is short, it is considered that the previous curve is a continuation. Then the process proceeds to step S205.

[0091] Next, the curve detection unit 142 checks whether the yaw angular velocity is less than or equal to a predetermined value A (for example, less than or equal to 0.02 rad / sec.) (step S205). That is, the curve detection unit 142 checks whether the basic conditions for curve detection are no longer met. If the yaw angular velocity is not less than or equal to the predetermined value A ("N" in step S205), the process in step S205 is repeated until the yaw angular velocity becomes less than or equal to the predetermined value A.

[0092] If the yaw angular velocity is less than or equal to a predetermined value A (indicated as "Y" in step S205), the curve detection unit 142 checks whether the yaw angular velocity has remained below the predetermined value A (for example, less than 0.02 rad / sec.) for a period of less than a predetermined time B (for example, less than 2 seconds) and then become greater than or equal to the predetermined value A (for example, 0.02 rad / sec. or greater) (step S206). If this condition is met (indicated as "Y" in step S206), the curve detection unit 142 considers that the previous curve is continuing (step S207). In other words, if this condition is met, the curve detection unit 142 determines that the vehicle 109 has swayed due to the driver adjusting the steering operation immediately after the end of the curve, and considers that the previous curve is continuing. Then, the process returns to step S205.

[0093] If the conditions in step S206 are not met ("N" in step S206), the curve detection unit 142 detects the end of the curve (step S208).

[0094] The curve detection unit 142 then checks whether it has read all the yaw angular velocity data included in the yaw angular velocity data DY (step S209). If not all the data has been read yet ("N" in step S209), the process returns to step S201.

[0095] In step S209, if all data has been read (indicated as "Y" in step S209), this process terminates.

[0096] Thus, in the first stage of the curve division process, the curve detection unit 142 basically detects a curve when a yaw angular velocity of a predetermined value A or greater (e.g., 0.02 rad / sec. or greater) continues for a predetermined time B or greater (e.g., 2 seconds or greater) (step S201). Furthermore, the curve detection unit 142 considers the previous curve to be continuing if, after the curve ends, a yaw angular velocity of less than the predetermined value A (e.g., less than 0.02 rad / sec.) continues for a predetermined time B or less (e.g., less than 2 seconds), and then the yaw angular velocity becomes the predetermined value A or greater (e.g., 0.02 rad / sec. or greater) (steps S205-S207). In addition, if the straight road between two curves curving in the same direction is short, the curve detection unit 142 considers these two curves to be one curve (steps S201, S202, S204).

[0097] Figure 13 shows a specific example of the second stage of processing in the curve division process. The curve detection unit 142 performs this second stage of processing using the results of the first stage of processing. As a result, the curve detection unit 142 decides whether or not to include each of the multiple curves detected in the first stage of processing in the driving skill evaluation.

[0098] First, the curve detection unit 142 selects the first curve from among the multiple curves obtained in the first stage of processing (step S221).

[0099] Next, the curve detection unit 142 checks whether the average value of the yaw angular velocity in the selected curve is less than a predetermined value D (for example, less than 0.05 rad / sec.) (step S222).

[0100] In step S222, if the average value of the yaw angular velocity is less than a predetermined value D (Y in step S222), the curve detection unit 142 checks whether the maximum value of the yaw angular velocity is greater than or equal to a predetermined value E (for example, 0.07 rad / sec. or greater) (step S223).

[0101] In step S222, if the average value of the yaw angular velocity is not less than a predetermined value D ("N" in step S222), or in step S223, if the maximum value of the yaw angular velocity is greater than or equal to a predetermined value E ("Y" in step S223), the curve detection unit 142 will include this curve in the evaluation for the driving skills assessment (step S224). Also, in step S223, if the maximum value of the yaw angular velocity is not greater than or equal to a predetermined value E ("N" in step S223), the curve detection unit 142 will not include this curve in the evaluation for the driving skills assessment (step S225).

[0102] Next, the curve detection unit 142 checks whether all curves have been selected (step S226). If not all curves have been selected yet ("N" in step S226), the curve detection unit 142 selects one of the unselected curves (step S227). Then, the process returns to step S221. The curve detection unit 142 repeats steps S221 to S227 until all curves have been selected.

[0103] Then, if the curve detection unit 142 has already selected all curves ("Y" in step S226), this process ends.

[0104] In this way, as shown in step S103 of Figure 11, the curve detection unit 142 generates curve data DC by performing a curve division process.

[0105] Next, the curve detection unit 142 checks whether sample data exists (step S104). Specifically, the curve detection unit 142 checks whether the sample data set DSA is stored in the storage unit 132.

[0106] In step S104, if there is no sample data ("N" in step S104), the image generation unit 144 generates multiple image data DPs by generating kernel density estimation images for each of the multiple curves (step S105).

[0107] Figure 14 shows an example of the process for generating a kernel density estimation image.

[0108] First, the image generation unit 144 performs preprocessing (step S241). Specifically, the image generation unit 44 first calculates time-series data of longitudinal jerk by differentiating the time-series data of longitudinal acceleration included in the acceleration data DA, and then calculates time-series data of the square of longitudinal jerk based on this time-series data of longitudinal jerk. The image generation unit 144 also calculates time-series data of yaw angular acceleration by differentiating the time-series data of yaw angular velocity included in the yaw angular velocity data DY, and then calculates time-series data of the square of yaw angular acceleration based on this time-series data of yaw angular acceleration. Finally, the image generation unit 144 generates a preprocessed image based on the time-series data of the square of longitudinal jerk and the time-series data of the square of yaw angular acceleration.

[0109] Figure 15 shows an example of the process for generating a preprocessed image. In this preprocessed image, the horizontal axis (X-axis) represents time, and the vertical axis (Y-axis) represents the square of the yaw angular acceleration. In this example, the preprocessed image is divided into 100 regions in the X-axis direction and 100 regions in the Y-axis direction. Therefore, the preprocessed image has 10,000 regions.

[0110] The full scale in the X-axis direction of the preprocessed image is set to, for example, 5 seconds, assuming a travel time of 10⁹ vehicles on a single curve.

[0111] The full scale in the Y-axis direction in the preprocessed image is set based on time-series data of the square of the yaw angular acceleration. The value of the square of the yaw angular acceleration generally varies greatly, and may be significantly out of bounds due to factors such as detection accuracy. Therefore, in this example, the image generation unit 144 performs a process to remove the significantly out-of-bounds values ​​from the square of the yaw angular acceleration. The image generation unit 44 can remove the significantly out-of-bounds values, for example, by using a box plot. Then, the image generation unit 144 finds the minimum and maximum values ​​from the data from which the significantly out-of-bounds values ​​have been removed, and determines the full scale in the Y-axis direction such that the range R of values ​​from the minimum to the maximum value fits within the Y-axis direction in Figure 15 with an appropriate margin M. The margin M can be, for example, about 3% of the range of values ​​from the minimum to the maximum value.

[0112] For example, if the full scale in the X-axis direction is 5 seconds and the sampling period for the time series data is 10 msec., then the number of data points in the time series data of the square of the forward and backward jerk is 500 (= 5 sec. / 10 msec.), and the number of data points in the time series data of the square of the yaw angular acceleration is also 500. In practice, however, as described above, outlier data among these data is removed. The image generation unit 144 maps the square of the forward and backward jerk values ​​to 10,000 regions in the preprocessed image based on the time and the square of the yaw angular acceleration.

[0113] The image generation unit 144, for each of the 10,000 regions, sets the square of the forward and backward jerk value for the region where one data point is mapped. Furthermore, for each of the 10,000 regions where multiple data points are mapped, the image generation unit 144 adds the squares of the forward and backward jerk values ​​for each of the multiple data points together and sets the summed value as the value for that region. In this way, the image generation unit 144 generates the pixel values ​​(Z-axis) of the preprocessed image. The image generation unit 144 then scales these pixel values, for example, to integers between 0 and 255, such that a larger square of the forward and backward jerk corresponds to a smaller pixel value, and a smaller square of the forward and backward jerk corresponds to a larger pixel value. In this way, the image generation unit 144 generates the preprocessed image.

[0114] Next, the image generation unit 144 performs kernel density estimation processing based on this preprocessed image, as shown in Figure 14 (step S242). In the kernel density estimation processing, the original data, including data that has not yet been observed, is estimated as density data based on the measured data. The image generation unit 144 performs kernel density estimation processing using known techniques. As a result, the image generation unit 144 generates a kernel density estimation image.

[0115] This completes the process.

[0116] In this way, as shown in step S105 of Figure 11, the image generation unit 144 generates multiple image data DPs by generating a kernel density estimation image for each of the multiple curves.

[0117] Then, the data registration unit 149 stores the acceleration data DA and yaw angular velocity data DY generated in step S102, the curve data DC generated in step S103, and the multiple image data DP generated in step S105 as a dataset DSA in the storage unit 132 (step S106). This completes the process.

[0118] In step S104, if sample data is available (indicated as "Y" in step S104), the curve detection unit 142 corrects the curve data DC by identifying the correspondence between the curves based on the similarity between the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY of the dataset DSA stored in the storage unit 132 (step S107).

[0119] Figure 16 shows an example of the process in step S107.

[0120] First, the curve detection unit 142 identifies the overall travel section to be processed (step S261) based on the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY of the dataset DSA. Specifically, the curve detection unit 142 identifies similar overall travel sections using Dynamic Time Wrapping (DTW) based on the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY of the dataset DSA. That is, since these yaw angular velocity data DY are both time-series data of yaw angular velocity in the area to be evaluated, it is desirable that they be almost the same. However, if the accuracy of the vehicle 109's position obtained by GNSS is not very good, the yaw angular velocity data DY may contain unnecessary data or have missing data at the beginning or end of the time-series data. Therefore, the curve detection unit 142 identifies overall travel sections with approximately the same number of curves using Dynamic Time Wrapping based on the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY of the dataset DSA. This allows the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY in the dataset DSA to be compared with each other over the entire travel section. In this way, the curve detection unit 142 identifies the entire travel section to be processed.

[0121] Next, the curve detection unit 142 identifies the correspondence between the multiple curves in the entire travel section to be processed, obtained from the yaw angular velocity data DY generated in step S102, and the multiple curves in the entire travel section to be processed, obtained from the yaw angular velocity data DY of the dataset DSA (step S262). Specifically, the curve detection unit 142 identifies the correspondence between the curves by identifying multiple sections that are similar to each other, using a dynamic time stretching method based on the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY of the dataset DSA.

[0122] Figure 17 shows an example of the processing in step S262, where (A) shows the yaw angular velocity data DY generated in step S102, and (B) shows the yaw angular velocity data DY of the dataset DSA. This Figure 17 shows the yaw angular velocity data DY for a portion of the travel section to be processed. In Figure 17, the numbers indicate the curve numbers in the curve data DC.

[0123] There is a difference in section W1 between these two yaw angular velocity data DYs. Specifically, in the yaw angular velocity data DY shown in Figure 17(A), one curve with curve number "6" is detected in section W1. On the other hand, in the yaw angular velocity data DY shown in Figure 17(B), three curves with curve numbers "6" to "8" are detected in section W1.

[0124] Therefore, in this example, the curve detection unit 142 changes the curve number of the curve in section W1 from "6" to "6, 7, 8" in the yaw angular velocity data DY (Figure 17(A)) generated in step S102, and also changes the curve numbers of the curves after this curve. That is, the curve detection unit 142 identifies the correspondence between curves so that the curve numbers of the multiple curves in the yaw angular velocity data DY (Figure 17(A)) generated in step S102 match the curve numbers of the multiple curves in the yaw angular velocity data DY (Figure 17(B)) of the dataset DSA. Then, based on this processing result, the curve detection unit 142 corrects the curve data DC generated in step S103.

[0125] Next, the curve detection unit 142 excludes curves with low similarity between the multiple curves in the yaw angular velocity data DY generated in step S102 and the multiple curves in the yaw angular velocity data DY of the dataset DSA from the evaluation of driving skills (step S263). Specifically, the curve detection unit 142 calculates the similarity of the curves that correspond to each other among the multiple curves in the yaw angular velocity data DY generated in step S102 and the multiple curves in the yaw angular velocity data DY of the dataset DSA, using the dynamic time stretching method. Then, the curve detection unit 142 excludes curves with a similarity lower than a predetermined amount from the evaluation of driving skills.

[0126] This completes the process.

[0127] In this way, as shown in step S107 of Figure 11, the curve detection unit 142 corrects the curve data DC by identifying the correspondence between the curves based on the similarity between the yaw angular velocity data DY generated in step S102 and the yaw angular velocity data DY of the dataset DSA.

[0128] Next, the image generation unit 144 generates multiple image data DPs by generating kernel density estimation images for each of the multiple curves (step S108). This process is the same as the process in step S105.

[0129] Then, the data registration unit 149 stores the acceleration data DA and yaw angular velocity data DY generated in step S102, the curve data DC generated in step S103 and corrected in step S107, and the multiple image data DP generated in step S108 as a dataset DSB in the storage unit 132 (step S109). This completes the process.

[0130] In this way, the data processing system 2 generates a dataset DS, which includes acceleration data DA, yaw angular velocity data DY, curve data DC, and multiple image data DP, based on driving data obtained by a skilled driver operating the vehicle 109. By repeating this process, the data processing system 2 can generate multiple datasets DS based on multiple driving data sets.

[0131] (Regarding the generation of DTC curve data to be evaluated) In the above, the data processing system 2 performs processing based on driving data obtained when a skilled driver operates the vehicle 109. However, it is also possible to perform processing based on driving data obtained when an unskilled driver operates the vehicle 109. In this case, the data processing system 2 can obtain kernel density estimation images related to the skilled driver and kernel density estimation images related to the unskilled driver.

[0132] This kernel density estimation image may change depending on the driver's driving skill. As described later, the driving skill evaluation system 1 evaluates the driver's driving skill based on the kernel density estimation image. In some curves, for example, there may be a large difference between the kernel density estimation image for a skilled driver and the kernel density estimation image for an unskilled driver. In other curves, for example, there may be almost no difference between the kernel density estimation image for a skilled driver and the kernel density estimation image for an unskilled driver. Therefore, the driving skill evaluation system 1 evaluates curves where there is a large difference between the kernel density estimation image for a skilled driver and the kernel density estimation image for an unskilled driver.

[0133] The following describes in detail the process of generating evaluation curve data (DTC) by determining the curves to be evaluated in the driving skills assessment.

[0134] Figure 18 shows an example of the process for generating the evaluation curve data DTC. In this example, the data processing system 2 has already acquired multiple datasets DS related to skilled drivers and multiple datasets DS related to unskilled drivers.

[0135] The evaluation data generation unit 147 selects one of several curves (step S301).

[0136] Next, the image similarity calculation unit 145 calculates the average value F1 of the image similarity between multiple kernel density estimation images related to skilled drivers and multiple kernel density estimation images related to unskilled drivers for one selected curve (step S302). Specifically, the image similarity calculation unit 145 calculates the image similarity of kernel density estimation images for all combinations between multiple kernel density estimation images related to one or more skilled drivers and multiple kernel density estimation images related to one or more unskilled drivers for one selected curve. Then, the image similarity calculation unit 145 calculates the average value F1 of these image similarities.

[0137] Next, the image similarity calculation unit 145 calculates the average value F2 of the image similarity between each of the multiple kernel density estimation images related to the skilled driver for the selected curve (step S303). Specifically, the image similarity calculation unit 145 calculates the image similarity of the kernel density estimation images for all combinations between each of the multiple kernel density estimation images related to one or more skilled drivers for the selected curve. Then, the image similarity calculation unit 145 calculates the average value F2 of these image similarities.

[0138] Next, the evaluation data generation unit 147 checks whether there is a significant difference between the average value F1 and the average value F2 (step S304). Specifically, the evaluation data generation unit 147 determines that there is a significant difference between the average value F1 and the average value F2 if the difference between them is greater than or equal to a predetermined amount. If there is a significant difference between the average value F1 and the average value F2 (indicated as "Y" in step S304), the evaluation data generation unit 147 uses the curve selected in step S301 as the subject of evaluation for the driving skills assessment (step S305). If there is no significant difference between the average value F1 and the average value F2 (indicated as "N" in step S304), the evaluation data generation unit 147 does not use the curve selected in step S301 as the subject of evaluation for the driving skills assessment (step S306).

[0139] Next, the evaluation data generation unit 147 checks whether all curves have been selected (step S307). If not all curves have been selected yet ("N" in step S307), the evaluation data generation unit 147 selects one of the unselected curves (step S308). Then, the process returns to step S302. The evaluation data generation unit 147 repeats the process from steps S302 to S308 until all curves have been selected.

[0140] In step S307, if all curves are selected ("Y" in step S307), the evaluation target data generation unit 147 generates evaluation target curve data DTC based on the processing results of steps S305 and S306 (step S309). Specifically, in steps S305 and S306, the evaluation target data generation unit 147 generates evaluation target curve data DTC that includes the curve numbers of the curves that were evaluated for the driving skill evaluation.

[0141] This completes the process.

[0142] In this way, the data processing system 2 generates evaluation curve data DTC based on driving data obtained by skilled and unskilled drivers operating the vehicle 109.

[0143] The data processing system 2 then transmits multiple datasets DS, evaluation target curve data DTC, and area data DAR to the server device 30. The server device 30 stores this data in the storage unit 32.

[0144] In this example, the processing unit 140 calculates mean values ​​F1 and F2 in steps S302 to S304 and determines whether to evaluate the selected curve by checking whether there is a significant difference between these mean values ​​F1 and F2, but it is not limited to this. Alternatively, the processing unit 140 may, for example, determine whether to evaluate the selected curve by performing a non-parametric U test based on multiple image similarity values ​​used to calculate mean value F1 and multiple image similarity values ​​used to calculate mean value F2, without using mean values ​​F1 and F2. Specifically, first, similar to step S302, the image similarity calculation unit 145 calculates the image similarity of kernel density estimation images for all combinations between multiple kernel density estimation images related to one or more skilled drivers and multiple kernel density estimation images related to one or more unskilled drivers for the selected curve. Next, similar to step S303, the image similarity calculation unit 145 calculates the image similarity of kernel density estimate images for all combinations between each of the multiple kernel density estimate images relating to one or more skilled drivers for the selected curve. Then, the evaluation target data generation unit 147 determines whether there is a significant difference between the multiple image similarities obtained in the first step and the multiple image similarities obtained in the later steps by performing, for example, a nonparametric U test based on the multiple image similarities obtained in the first step and the multiple image similarities obtained in the later steps. If there is a significant difference, the evaluation target data generation unit 147 makes the curve selected in step S301 the subject of evaluation for driving skill assessment (step S305). If there is no significant difference ("N" in step S304), the evaluation target data generation unit 147 does not make the curve selected in step S301 the subject of evaluation for driving skill assessment (step S306).

[0145] (Regarding the evaluation of driving skills) The driving skills evaluation system 1 evaluates the driver's driving skills based on driving data generated by the driver operating the vehicle 9, using multiple datasets DS, evaluation target curve data DTC, and area data DAR stored in the storage unit 32 of the server device 30. This operation is described in detail below.

[0146] When the driver operates vehicle 9, the acceleration sensor 14 of the smartphone 10 detects acceleration in three directions in the coordinate system of the smartphone 10, the angular velocity sensor 15 detects three angular velocities (yaw angular velocity, roll angular velocity, and pitch angular velocity) in the coordinate system of the smartphone 10, and the GNSS receiver 16 obtains the position of vehicle 9 on the ground.

[0147] The data processing unit 21 performs predetermined data processing, such as filtering, based on the detection results from the acceleration sensor 14 and the angular velocity sensor 15. Specifically, the data processing unit 21 filters the time-series data of acceleration detected by the acceleration sensor 14 and filters the time-series data of angular velocity detected by the angular velocity sensor 15. However, it is not limited to this, and the data processing unit 21 may also perform downsampling on these filtered time-series data.

[0148] Then, after the run is completed, the communication unit 13 transmits the time-series data of acceleration and time-series data of angular velocity processed by the data processing unit 21 to the server device 30, along with the time-series data of the vehicle 9's position obtained by the GNSS receiver 16.

[0149] In the server device 30, the communication unit 31 receives data transmitted from the smartphone 10. Based on the data received by the communication unit 31, the server device 30 evaluates the driver's driving skills. The operation of this server device 30 will be described in detail below.

[0150] Figures 19A and 19B illustrate an example of the operation of the server device 30.

[0151] First, the data processing unit 41 of the server device 30 uses the area data DAR stored in the storage unit 32 to check whether the vehicle 9 has traveled through the evaluation area based on the time-series data of the vehicle 9's position (step S401). If the vehicle 9 has not traveled through the evaluation area ("N" in step S401), this process ends.

[0152] When vehicle 9 travels through the evaluation area ("Y" in step S401), the data processing unit 41 generates time-series data of longitudinal acceleration and yaw angular velocity by performing a coordinate transformation (step S402). Specifically, the data processing unit 41 generates time-series data of acceleration (longitudinal acceleration) in the direction of travel of vehicle 9 by performing a coordinate transformation based on the time-series data of acceleration in the evaluation area. The data processing unit 41 also generates time-series data of the yaw angular velocity of vehicle 9 by performing a coordinate transformation based on the time-series data of angular velocity in the evaluation area. However, it is not limited to this, and for example, if downsampling is performed in the smartphone 10, the data processing unit 41 may perform upsampling on the time-series data of acceleration and angular velocity received by the communication unit 31, and perform a coordinate transformation based on the upsampled time-series data of longitudinal acceleration and yaw angular velocity.

[0153] Next, the data processing unit 41 performs a filter process on the time-series data of longitudinal acceleration and yaw angular velocity generated in step S402 to generate acceleration data DA1 and yaw angular velocity data DY1, respectively (step S403).

[0154] Next, the curve detection unit 42 generates curve data DC1 by performing a curve division process based on the yaw angular velocity data DY1 generated in step S403 (step S404). This curve division process is the same as the process in step S103 shown in Figure 11.

[0155] Next, the curve detection unit 42 corrects the curve data DC1 by identifying the correspondence between the curves based on the similarity between the yaw angular velocity data DY1 generated in step S403 and the yaw angular velocity data DY of the dataset DSA stored in the storage unit 32 (step S405). This process is the same as the process in step S107 shown in Figure 11.

[0156] The data extraction unit 43 extracts time-series data of longitudinal acceleration and time-series data of yaw angular velocity related to the curves that are the subject of evaluation for driving skills, based on the evaluation target curve data DTC stored in the storage unit 32 (step S406). Specifically, the data extraction unit 43 extracts time-series data of longitudinal acceleration related to multiple curves used for evaluating driving skills from the time-series data of longitudinal acceleration included in the acceleration data DA1. The data extraction unit 43 also extracts time-series data of yaw angular velocity related to multiple curves used for evaluating driving skills from the time-series data of yaw angular velocity included in the yaw angular velocity data DY1.

[0157] Next, the image generation unit 44 generates multiple image data DPs by generating kernel density estimation images for each of the multiple curves based on the time-series data of longitudinal acceleration and yaw angular velocity extracted in step S406 (step S407). This process is the same as the process in step S108. In step S406, the image generation unit 44 generates kernel density estimation images related to the curves that are the subject of evaluation for the driving skill assessment.

[0158] Next, the image similarity calculation unit 45 calculates the average value of the image similarity (average similarity) based on the multiple kernel density estimation images generated in step S407 and the multiple kernel density estimation images included in the multiple datasets DS related to skilled drivers stored in the storage unit 32 (step S408). Specifically, the image similarity calculation unit 45 calculates the image similarity for one of the multiple curves by comparing the kernel density estimation image generated by the image generation unit 44 with the multiple kernel density estimation images included in the multiple datasets DS related to skilled drivers. The image similarity calculation unit 45 calculates multiple image similarities by performing this process for each of the multiple curves. Then, the image similarity calculation unit 45 calculates the average value of these multiple image similarities (average similarity). In this example, the image similarity value is a positive value, and the more similar the kernel density estimation images are, the smaller the image similarity value, and the less similar the kernel density estimation images are, the larger the image similarity value.

[0159] Next, the skill determination unit 46 checks whether the number of curves detected by the curve detection unit 42 in steps S404 and S405 is equal to or greater than a predetermined number G1 (step S409). That is, when the curve detection unit 42 detects curves in steps S404 and S405, it may not be able to detect some curves depending on the yaw angular velocity data DY1. For example, if the driver's driving skills are low and the yaw angular velocity data DY1 differs greatly from the yaw angular velocity data DY of a skilled driver, the number of curves that cannot be detected may increase. Therefore, the skill determination unit 46 checks whether the number of curves detected based on the yaw angular velocity data DY1 is equal to or greater than a predetermined number G1.

[0160] In step S409, if the number of detected curves is greater than or equal to a predetermined number G1 (Y in step S109), the skill determination unit 46 checks whether the average similarity value calculated in step S408 is less than the threshold G2 (step S410).

[0161] In step S410, if the average similarity value is less than the threshold G2 ("Y" in step S410), the skill determination unit 46 determines that the driver of vehicle 9 has high driving skills (step S411). That is, in step S410, since the average similarity value is less than the threshold G2, the kernel density estimation image of the driver is similar to the kernel density estimation image of a skilled driver. Therefore, the skill determination unit 46 determines that the driver of vehicle 9 has high driving skills.

[0162] Furthermore, in step S409, if the number of detected curves is not greater than or equal to a predetermined number G1 ("N" in step S409), or in step S410, if the average similarity value is not less than the threshold G2 ("N" in step S410), the skill determination unit 46 determines that the driver of vehicle 9 has a low skill level (step S412). In other words, in step S409, if the number of detected curves is not greater than or equal to a predetermined number G1, the yaw angular velocity data DY1 of that driver differs greatly from the yaw angular velocity data DY of a skilled driver, so the skill determination unit 46 determines that the driver has low driving skills. Also, in step S410, if the average similarity value is not less than the threshold G2, the kernel density estimation image of the driver does not resemble the kernel density estimation image of a skilled driver, so the skill determination unit 46 determines that the driver has low driving skills.

[0163] This completes the process.

[0164] The communication unit 31 of the server device 30 transmits data including the driving skill evaluation results to the smartphone 10. The communication unit 13 of the smartphone 10 receives the data transmitted from the smartphone 10. The display processing unit 22 performs display processing based on the data indicating the driving skill evaluation results transmitted from the server device 30. The touch panel 11 displays the driving skill evaluation results. This allows the driver to obtain an objective evaluation of their driving skills.

[0165] Thus, the driving skill evaluation method in the driving skill evaluation system 1 includes generating a kernel density estimation image based on time-series data of a first parameter (in this example, the square of the yaw angular velocity) corresponding to the change in the direction of travel of the vehicle 9, and time-series data of a second parameter representing the square of the jerk (forward / backward jerk) in the direction of travel of the vehicle 9, and evaluating the driving skills of the vehicle driver by comparing the kernel image estimation image with a reference image of a skilled driver. As a result, the driving skill evaluation method can improve the accuracy of evaluating the driver's driving skills. In other words, in the kernel density estimation process, the original data, including data that has not yet been observed, is estimated as density data based on the measured data. Therefore, the kernel image estimation image may include the unique characteristics inherent to that driver, corresponding to the driver's driving skills. Thus, this driving skill evaluation method can improve the accuracy of evaluating the driver's driving skills by using the kernel image estimation image.

[0166] Furthermore, this driving skill evaluation method generates a kernel density estimation image based on time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle 9, thereby improving the accuracy of the driver's skill evaluation. That is, for example, if a kernel density estimation image is generated based on time-series data of the acceleration in the direction of travel of the vehicle 9 (longitudinal acceleration), the pixel values ​​of the pre-processed image may become negative, making it impossible to perform kernel density estimation. Also, for example, as shown in Figure 20, this longitudinal acceleration is easily affected by gravitational acceleration, such as on a slope, so a trend occurs in the longitudinal acceleration. Therefore, for example, if a kernel density estimation image is generated based on time-series data of the absolute value of this longitudinal acceleration, the kernel density estimation image will be affected by gravitational acceleration, which may lower the accuracy of the driver's skill evaluation. On the other hand, this driving skill evaluation method generates a kernel density estimation image based on time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle 9 (longitudinal jerk). Specifically, the longitudinal jerk is calculated by differentiating the longitudinal acceleration with respect to time, the square of this longitudinal jerk is calculated, and a kernel density estimation image is generated based on the time-series data of this squared longitudinal jerk. This makes the pixel values ​​of the preprocessed image positive, which suppresses the effect of gravitational acceleration and thus improves the accuracy of evaluating the driver's driving skills.

[0167] Furthermore, in this driving skill evaluation method, the time-series data for the first parameter (in this example, the square of the yaw angular velocity) and the time-series data for the second parameter (in this example, the square of the longitudinal jerk) were taken when vehicle 9 was traveling on a curve. On a curve, the driver performs various operations such as steering, braking, and acceleration, so the time-series data on the curve contains information about the driver's various operations. Therefore, this driving skill evaluation method can improve the accuracy of evaluating the driver's driving skills.

[0168] Furthermore, in this driving skill evaluation method, as shown in Figures 6 and 7, the first image direction in the kernel density estimation image represents time, the second image direction in the kernel density estimation image represents the first parameter (in this example, the square of the yaw angular velocity), and the pixel value of the kernel density estimation image is set to a value corresponding to the data of the second parameter (in this example, the square of the longitudinal jerk). This makes it possible to improve the accuracy of evaluating the driver's driving skills, for example, on roads where the speed of vehicle 9 changes significantly. That is, for example, on a road with many undulations, the speed of vehicle 9 changes significantly, and the value of the square of the longitudinal jerk can change significantly, so the pixel value of the kernel density estimation image can change significantly. Therefore, this driving skill evaluation method can improve the accuracy of evaluating the driver's driving skills, especially on roads where the speed of vehicle 9 changes significantly.

[0169] Furthermore, this driving skills evaluation method now includes presenting the driver with the evaluation results of their driving skills. This allows drivers to obtain an objective evaluation of their own driving skills. As a result, drivers can pay more attention to their driving and strive to improve their driving skills.

[0170] [effect] As described above, in this embodiment, the driving skill evaluation method includes generating a kernel density estimation image based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter indicating the square of the jerk in the direction of travel of the vehicle, and evaluating the driving skills of the vehicle driver by comparing the kernel image estimation image with a reference image related to a skilled driver. Therefore, the accuracy of evaluating the driver's driving skills can be improved.

[0171] In this embodiment, kernel density estimation images are generated based on time-series data of a second parameter that represents the square of the jerk in the direction of vehicle movement, thereby improving the accuracy of evaluating the driver's driving skills.

[0172] In this embodiment, the time-series data for the first parameter and the time-series data for the second parameter are based on data collected when the vehicle is traveling around a curve, thereby improving the accuracy of the evaluation of the driver's driving skills.

[0173] In this embodiment, the first image direction in the kernel density estimation image represents time, the second image direction in the kernel density estimation image represents the first parameter, and the pixel values ​​in the kernel density estimation image are set to values ​​corresponding to the data of the second parameter. Therefore, for example, on roads where the vehicle speed changes significantly, the accuracy of evaluating the driver's driving skills can be improved.

[0174] In this embodiment, the driving skills evaluation method further includes presenting the driver with the evaluation results of their driving skills, so that the driver can obtain an objective evaluation of their own driving skills. As a result, the driver can pay attention to their driving and work to improve their driving skills.

[0175] [Example 1] In the above embodiment, as shown in Figures 6 and 7, the vertical axis (Y-axis) of the kernel density estimation image represents the square of the yaw angular acceleration, but it is not limited to this. Alternatively, for example, the vertical axis (Y-axis) of the kernel density estimation image may represent the yaw angular acceleration as shown in Figure 21, or the absolute value of the yaw angular acceleration. Also, for example, the vertical axis (Y-axis) of the kernel density estimation image may represent the yaw angular velocity as shown in Figure 22, or the absolute value of the yaw angular velocity. Furthermore, the vertical axis (Y-axis) of the kernel density estimation image may represent the square of the acceleration in the direction intersecting the direction of vehicle travel (lateral acceleration) as shown in Figure 23, or it may represent the lateral acceleration as shown in Figure 24, or it may represent the absolute value of the lateral acceleration.

[0176] [Differentiation 2] In the above embodiment, as shown in Figures 6 and 7, the vertical axis (Y-axis) of the kernel density estimation image represents the square of the yaw angular acceleration, and the pixel values ​​(Z-axis) of the kernel density estimation image represent the square of the longitudinal jerk. However, the embodiment is not limited to this. Alternatively, for example, as shown in Figures 25 and 26, the vertical axis (Y-axis) of the kernel density estimation image may represent the square of the longitudinal jerk, and the pixel values ​​(Z-axis) of the kernel density estimation image may represent the square of the yaw angular acceleration. Figure 25 shows an example of a kernel density estimation image on a curve. Figure 26 shows the coordinate axes of the kernel density estimation image shown in Figure 25. This makes it possible to improve the accuracy of evaluating the driver's driving skills, for example, especially on roads where the change in vehicle speed 9 is small. That is, for example, on a road with gentle undulations, the change in vehicle speed 9 is small, and the value of the square of the longitudinal jerk does not change significantly, so the pixel values ​​of the kernel density estimation image shown in Figures 6 and 7 do not change significantly. In such cases, the accuracy of evaluating the driver's driving skills can be improved by using the kernel density estimation images shown in Figures 25 and 26.

[0177] The driving skill evaluation system 1 may be capable of processing based on both the kernel density estimation images shown in Figures 6 and 7, and the kernel density estimation images shown in Figures 25 and 26. For example, the information processing device 130 of the data processing system 2 can decide, based on the engineer's operation, whether to generate the kernel density estimation images shown in Figures 6 and 7, or the kernel density estimation images shown in Figures 25 and 26. Alternatively, the information processing device 130 may decide which of these kernel density estimation images to generate based on acceleration data DA, which includes time-series data of longitudinal acceleration, and yaw angular velocity data DY, which includes time-series data of yaw angular velocity. Specifically, the information processing device 130 checks, based on the acceleration data DA and yaw angular velocity data DY, whether the road traveled by the vehicle 109 is a road where the speed of the vehicle 109 changes significantly. The information processing device 130 then decides to generate kernel density estimation images shown in Figures 6 and 7 if the road traveled by the vehicle 109 is a road where the speed of the vehicle 109 changes significantly, and to generate kernel density estimation images shown in Figures 25 and 26 if the road traveled by the vehicle 109 is not a road where the speed of the vehicle 109 changes significantly.

[0178] The information processing device 130 may decide to generate kernel density estimation images shown in Figures 6 and 7 for all curves, or it may decide to generate kernel density estimation images shown in Figures 25 and 26 for all curves. Alternatively, the information processing device 130 may individually decide for each of the multiple curves whether to generate the kernel density estimation image shown in Figures 6 and 7 or the kernel density estimation image shown in Figures 25 and 26.

[0179] If the kernel density estimation image for a skilled driver in the dataset DS is the kernel density estimation image shown in Figures 6 and 7, the server device 30 generates the kernel density estimation image shown in Figures 6 and 7 based on the driver's driving data. If the kernel density estimation image for a skilled driver in the dataset DS is the kernel density estimation image shown in Figures 25 and 26, the server device 30 generates the kernel density estimation image shown in Figures 25 and 26 based on the driver's driving data.

[0180] Furthermore, in the examples in Figures 25 and 26, the pixel values ​​(Z-axis) of the kernel density estimation image represent the square of the yaw angular acceleration, but this is not the only way to do so. Alternatively, for example, as in Modification 1, the pixel values ​​(Z-axis) of the kernel density estimation image may represent the absolute value of the yaw angular acceleration, the square of the transverse acceleration, or the absolute value of the transverse acceleration.

[0181] [Difference 3] In the above embodiment, as shown in Figure 2, the smartphone 10 transmitted time-series data of acceleration, time-series data of angular velocity, and time-series data of the vehicle 9's position to the server device 30, but it is not limited to this. Alternatively, for example, the smartphone 10 may also transmit time-series data of the geomagnetic field to the server device 30. This example will be described in detail below.

[0182] Figure 27 shows an example configuration of a smartphone 10A according to this modified example. The smartphone 10A has a geomagnetic sensor 17A and a processing unit 20A. The geomagnetic sensor 17A is configured to detect the Earth's magnetic field. The processing unit 20A collects time-series data of acceleration detected by the acceleration sensor 14, time-series data of angular velocity detected by the angular velocity sensor 15, time-series data of the vehicle 9's position obtained by the GNSS receiver 16, and time-series data of the Earth's magnetic field detected by the geomagnetic sensor 17A. After the vehicle has finished driving, the communication unit 13 transmits the time-series data of acceleration and angular velocity processed by the data processing unit 21, along with the time-series data of the vehicle 9's position obtained by the GNSS receiver 16 and the time-series data of the Earth's magnetic field detected by the geomagnetic sensor 17A, to the server device 30.

[0183] The data processing unit 41 of the server device 30 generates time-series data of acceleration (longitudinal acceleration) in the direction of travel of the vehicle 9 by performing coordinate transformations based on the time-series acceleration data received by the communication unit 31. It also generates time-series data of the yaw angular velocity of the vehicle 9 by performing coordinate transformations based on the time-series angular velocity data received by the communication unit 31. Then, the data processing unit 41 corrects the time-series data of longitudinal acceleration and yaw angular velocity based on the time-series geomagnetic data received by the communication unit 31. This improves the accuracy of the time-series data of longitudinal acceleration and yaw angular velocity.

[0184] [Other variations] Furthermore, two or more of these variations may be combined.

[0185] While several embodiments of this disclosure have been described above with reference to the accompanying drawings, this disclosure is by no means limited to the embodiments described above. Those skilled in the art will understand that various modifications and changes can be made without departing from the scope defined by the claims. This disclosure is intended to encompass such modifications and changes insofar as they fall within the scope of the claims and their equivalents.

[0186] For example, in the above embodiment, the server device 30 evaluates driving skills based on driving data transmitted from the smartphone 10, but it is not limited to this. Alternatively, for example, the smartphone 10 may evaluate driving skills based on driving data. In this case, the processing unit 20 of the smartphone 10 may operate as a data processing unit 41, a curve detection unit 42, a data extraction unit 43, an image generation unit 44, an image similarity calculation unit 45, and a skill determination unit 46. In addition, the storage unit 12 of the smartphone 10 stores multiple datasets DS (dataset DSA and multiple datasets DSB), evaluation target curve data DTC, and area data DAR.

[0187] Furthermore, the onboard device of vehicle 9 may collect driving data of vehicle 9 and evaluate driving skills based on this driving data. Similar to the onboard device 110 (Figure 9), this onboard device of vehicle 9 collects time-series data of longitudinal acceleration detected by an acceleration sensor, time-series data of yaw angular velocity detected by a yaw angular velocity sensor, and time-series data of the position of vehicle 9 obtained by a GNSS receiver. The processing unit of the onboard device of vehicle 9 may operate as a data processing unit 41, a curve detection unit 42, a data extraction unit 43, an image generation unit 44, an image similarity calculation unit 45, and a skill determination unit 46. In addition, the storage unit of this onboard device stores multiple datasets DS (dataset DSA and multiple datasets DSB), evaluation target curve data DTC, and area data DAR.

[0188] The effects described herein are illustrative only, and the effects of this disclosure are not limited to those described herein. Therefore, other effects may be obtained with respect to this disclosure.

[0189] Furthermore, this disclosure may take the following forms:

[0190] (1) A kernel density estimation image is generated based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle. By comparing the kernel density estimation image with a reference image, the driving skills of the vehicle's driver are evaluated. A method for evaluating driving skills, including the following. (2) The time-series data for the first parameter and the time-series data for the second parameter are data obtained when the vehicle is traveling on a curve. The driving skills evaluation method described in (1) above. (3) In the kernel density estimation image, the first image direction represents time. The second image orientation in the kernel density estimation image indicates the first parameter. The pixel values ​​of the kernel density estimation image are values ​​corresponding to the data of the second parameter. The driving skills evaluation method described in (1) or (2) above. (4) The first parameter mentioned above is a parameter relating to the yaw angular velocity. A driving skills evaluation method as described in any of (1) to (3) above. (5) The first parameter is a parameter relating to yaw angular acceleration. The driving skills evaluation method described in (4) above. (6) The first parameter mentioned above is a parameter relating to the square of the yaw angular acceleration. The driving skills evaluation method described in (5) above. (7) The first parameter is a parameter relating to lateral acceleration. A driving skills evaluation method as described in any of (1) to (3) above. (8) In the kernel density estimation image, the first image direction represents time. The second image orientation in the kernel density estimation image indicates the second parameter. The pixel values ​​of the kernel density estimation image are values ​​corresponding to the data of the first parameter. The driving skills evaluation method described in (1) or (2) above. (9) The further includes presenting the driver with the results of an evaluation of the driver's driving skills. A driving skills evaluation method as described in any of (1) to (8) above. (10) An image generation circuit generates a kernel density estimation image based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle. An evaluation circuit that evaluates the driving skills of the vehicle's driver by comparing the kernel density estimation image with a reference image. A driving skills evaluation system equipped with [features / equipment]. (11) A kernel density estimation image is generated based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle. By comparing the kernel density estimation image with a reference image, the driving skills of the vehicle's driver are evaluated. The software that instructs the processor to perform this task was recorded. Recording medium.

[0191] The processing unit 40 shown in Figure 3 can be implemented by a circuit including at least one semiconductor integrated circuit, such as at least one processor (e.g., a central processing unit (CPU)), at least one application-specific integrated circuit (ASIC) and / or at least one field-programmable gate array (FPGA). At least one processor can be configured to perform all or some of the functions of the processing unit 40 shown in Figure 3 by reading instructions from at least one non-temporary, tangible computer-readable medium. Such a medium can take various forms, including, but is not limited to, various magnetic media such as hard disks, various optical media such as CDs or DVDs, and various semiconductor memories (i.e., semiconductor circuits) such as volatile or non-volatile memory. Volatile memory may include DRAM and SRAM. Non-volatile memory may include ROM and NVRAM. An ASIC is an integrated circuit (IC) specialized to perform all or some of the functions of the processing unit 40 shown in Figure 3. An FPGA is an integrated circuit designed to be configurable after manufacturing to perform all or some of the functions of the processing unit 40 shown in Figure 3.

Claims

1. A kernel density estimation image is generated based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle. The kernel density estimation image is compared with a reference image obtained based on the driving data of a skilled driver to calculate image similarity, and the driving skills of the vehicle's driver are evaluated based on the image similarity to determine whether the driving skills of the vehicle's driver are similar to those of the skilled driver. Includes, The first parameter is one of the following: a parameter relating to yaw angular velocity, a parameter relating to yaw angular acceleration, a parameter relating to the square of yaw angular acceleration, a parameter relating to lateral acceleration, or a parameter relating to the square of lateral acceleration. Methods for evaluating driving skills.

2. The time-series data for the first parameter and the time-series data for the second parameter are data obtained when the vehicle is traveling on a curve. The method for evaluating driving skills according to claim 1.

3. In the kernel density estimation image, the first image direction indicates time. The second image direction in the kernel density estimation image indicates the first parameter. The pixel values ​​of the kernel density estimation image are values ​​corresponding to the data of the second parameter. The method for evaluating driving skills according to claim 1.

4. In the kernel density estimation image, the first image direction indicates time. The second image direction in the kernel density estimation image indicates the second parameter. The pixel values ​​of the kernel density estimation image are values ​​corresponding to the data of the first parameter. The method for evaluating driving skills according to claim 1.

5. The further includes presenting the driver with the results of an evaluation of the driver's driving skills. The method for evaluating driving skills according to claim 1.

6. An image generation circuit generates a kernel density estimation image based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter indicating the square of the jerk in the direction of travel of the vehicle. An evaluation circuit for evaluating the driving skills of a vehicle driver is provided, which calculates image similarity by comparing the kernel density estimation image with a reference image obtained based on the driving data of a skilled driver, and evaluates whether the driving skills of the vehicle driver are similar to those of the skilled driver based on the image similarity. Equipped with, The first parameter is one of the following: a parameter relating to yaw angular velocity, a parameter relating to yaw angular acceleration, a parameter relating to the square of yaw angular acceleration, a parameter relating to lateral acceleration, or a parameter relating to the square of lateral acceleration. Driving skills evaluation system.

7. A kernel density estimation image is generated based on time-series data of a first parameter corresponding to the change in the direction of travel of the vehicle, and time-series data of a second parameter representing the square of the jerk in the direction of travel of the vehicle. The kernel density estimation image is compared with a reference image obtained based on the driving data of a skilled driver to calculate image similarity, and the driving skills of the vehicle's driver are evaluated based on the image similarity to determine whether the driving skills of the vehicle's driver are similar to those of the skilled driver. Software that instructs the processor to perform this task is recorded. The first parameter is one of the following: a parameter relating to yaw angular velocity, a parameter relating to yaw angular acceleration, a parameter relating to the square of yaw angular acceleration, a parameter relating to lateral acceleration, or a parameter relating to the square of lateral acceleration. Recording medium.

Images