Driving evaluation model generation device, driving evaluation model generation method and program

Abstract

To generate an operation evaluation model taking difference of feelings generated due to operation experience, living space or the like into consideration.SOLUTION: An operation evaluation model generation device comprises a first storage unit for storing image information on a location where a mobile body travels and a second storage unit for storing evaluation information on operation evaluated by a determination person for the image information, in association with a profile of the determination person. The generation unit then generates an operation evaluation model on the basis of the image information and the evaluation information. Here, when an operation evaluation model corresponding to one evaluation condition is generated, a generation unit generates the operation evaluation model using image information and evaluation information not corresponding to the one evaluation condition in addition to image information and evaluation information corresponding to the one evaluation condition.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a method for generating a model for making an assessment regarding driving. [Background technology]

[0002] Conventionally, there has been known a system that aggregates data from multiple vehicles into a server device and distributes information related to the learning results obtained by the server device through machine learning. For example, Patent Document 1 discloses a system that collects environmental data on conditions such as the weather and time of day when each vehicle is traveling, and performs learning for each condition indicated by the environmental data. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-133552 Summary of the Invention [Problem to be solved by the invention]

[0004] Regarding situation recognition and judgment while driving, there are individual differences based on the driver's experience, etc., and recognition and judgment may differ depending on the driver. In this regard, Patent Document 1 does not take such individual differences into consideration at all.

[0005] The above is one example of a problem to be solved by the present invention. The present invention aims to generate a driving evaluation model that takes into account differences in perception due to driving experience, living area, etc. [Means for solving the problem]

[0006] The invention described in claim 1 is a driving evaluation model generation device comprising: a first memory unit that stores image information of locations where a moving body is traveling; a second memory unit that stores evaluation information regarding driving evaluated by an evaluator based on the image information in association with the evaluator's profile; and a generation unit that, when generating a driving evaluation model corresponding to a certain evaluation condition, generates the driving evaluation model using image information and evaluation information when the certain evaluation condition is met, as well as image information and evaluation information when the certain evaluation condition is not met.

[0007] The invention described in claim 6 is a driving evaluation model generation method executed by a driving evaluation model generation device, and includes a first acquisition step of acquiring image information of a location where a moving body is traveling, a second acquisition step of acquiring evaluation information regarding driving evaluated by an evaluator based on the image information in association with the evaluator's profile, and a generation step of generating a driving evaluation model corresponding to a certain evaluation condition using image information and evaluation information when the certain evaluation condition is met, as well as image information and evaluation information when the certain evaluation condition is not met.

[0008] The invention described in claim 7 is a program executed by an apparatus having a computer, which causes the computer to execute the following steps: a first acquisition step of acquiring image information of a location where a moving body is traveling; a second acquisition step of acquiring evaluation information regarding driving evaluated by an evaluator based on the image information in association with the evaluator's profile; and a generation step of generating a driving evaluation model corresponding to a certain evaluation condition by using image information and evaluation information when the certain evaluation condition is met, as well as image information and evaluation information when the certain evaluation condition is not met. [Brief explanation of the drawings]

[0009] [Figure 1] 1 shows a configuration of a learning system according to an embodiment of the present invention. [Figure 2] 10 shows an example of data stored in the teacher data management unit. [Figure 3]10 is a flowchart of a first embodiment of a learning process. [Figure 4] 10 is a flowchart of a second embodiment of the learning process. DETAILED DESCRIPTION OF THE INVENTION

[0010] In one preferred embodiment of the present invention, a driving evaluation model generation device includes a first memory unit that stores image information of locations where a moving body is traveling, a second memory unit that stores evaluation information regarding driving evaluated by an evaluator based on the image information in association with the evaluator's profile, and a generation unit that, when generating a driving evaluation model corresponding to a certain evaluation condition, generates the driving evaluation model using image information and evaluation information when the certain evaluation condition is met, as well as image information and evaluation information when the certain evaluation condition is not met.

[0011] The driving evaluation model generation device includes a first storage unit that stores image information of locations where a moving object travels, and a second storage unit that stores evaluation information regarding driving evaluated by an evaluator based on the image information, in association with the evaluator's profile. When generating a driving evaluation model corresponding to a certain evaluation condition, the generation unit generates the driving evaluation model using image information and evaluation information for cases where the certain evaluation condition is met, as well as image information and evaluation information for cases where the certain evaluation condition is not met. This makes it possible to generate a general-purpose driving evaluation model that shows appropriate behavior even in situations where the evaluation condition is not met.

[0012] In one aspect of the driving evaluation model generation device, the one evaluation condition includes information related to a region, and when generating a driving evaluation model corresponding to the one region, the generation unit generates the driving evaluation model using image information and evaluation information obtained in the one region, as well as image information and evaluation information obtained in a region different from the one region, more preferably a region opposite the one region. In this aspect, it is possible to generate a driving evaluation model that shows appropriate behavior even in regions other than the region indicated by the evaluation condition.

[0013] In another aspect of the driving evaluation model generation device, the one evaluation condition includes information about the environment of the moving body, and when generating a driving evaluation model corresponding to the one environment, the generation unit generates the driving evaluation model using image information and evaluation information obtained in the one environment, as well as image information and evaluation information obtained in an environment different from the one environment, more preferably an environment opposite to the one environment. In this aspect, it is possible to generate a driving evaluation model that shows appropriate behavior even in environments other than those indicated by the evaluation condition.

[0014] In another preferred embodiment of the present invention, a driving evaluation model generation method executed by a driving evaluation model generation device includes: a first acquisition step of acquiring image information of a location where a mobile object travels; a second acquisition step of acquiring evaluation information regarding driving evaluated by an evaluator based on the image information in association with a profile of the evaluator; and a generation step of generating a driving evaluation model corresponding to a certain evaluation condition using image information and evaluation information for cases where the certain evaluation condition is met, as well as image information and evaluation information for cases where the certain evaluation condition is not met. This method also makes it possible to generate a general-purpose driving evaluation model that shows appropriate behavior even in situations where the evaluation condition is not met.

[0015] In another preferred embodiment of the present invention, a program executed by an apparatus including a computer causes the computer to execute the following steps: a first acquisition step of acquiring image information of a location where a moving object travels; a second acquisition step of acquiring evaluation information regarding driving evaluated by an evaluator based on the image information in association with the evaluator's profile; and a generation step of generating a driving evaluation model corresponding to a certain evaluation condition using image information and evaluation information when the certain evaluation condition is met, as well as image information and evaluation information when the certain evaluation condition is not met. By executing this program on a computer, the above-mentioned driving evaluation model generation device can be realized. This program can be stored in a storage medium and used. [Example]

[0016] Preferred embodiments of the present invention will now be described with reference to the drawings. [System Configuration] 1 shows the configuration of a learning system that is one embodiment of a driving evaluation model generation device of the present invention. The learning system 10 is realized by a device equipped with a computer, such as a server device. As shown in the figure, the learning system 10 includes an image data storage unit 11, a teacher data management unit 12, a learning data control unit 13, a machine learning unit 14, and a driving evaluation model database (hereinafter, "database" will be abbreviated as "DB").

[0017] Image data captured by a vehicle camera is stored in the image data storage unit 11. In a typical example, the image data is captured by an on-board camera while the vehicle is traveling and is supplied to the learning system 10, which is configured as a server, via wireless communication over the Internet. The image data is an image of the environment in which the vehicle is traveling, specifically, an image of the road ahead of the vehicle and its surroundings. The image data is stored in the image data storage unit 11 in association with an image data ID, which is its identification information.

[0018] The teacher data management unit 12 stores the results of an evaluation of driving by an evaluator as a teacher for the driving environment indicated by the image data. FIG. 2 shows an example of data stored in the teacher data management unit 12. The teacher data management unit 12 stores multiple pieces of teacher data. Each piece of stored data is composed of an image data ID, a profile, environmental information, and a judgment value. The teacher data stored in the teacher data management unit 12 is used for learning by the machine learning unit 14, which will be described later. Note that the teacher data can be generated from, for example, probe data transmitted from a vehicle that is actually traveling or a car navigation device installed in the vehicle. Alternatively, the teacher data can be generated by a questionnaire method in which the evaluator is shown the image data and asked to evaluate the driving, and at the same time, the evaluator's profile and environmental information at the time are obtained.

[0019] "Image data ID" is identification information for the image data that is the subject of evaluation by the evaluator. "Profile" is the profile of the instructor when performing machine learning for driving evaluation, and actually indicates the profile of the evaluator who makes the evaluation on the image data. The profile includes the instructor identifier, gender, year of birth, place of residence, and driving history.

[0020] The "teacher identifier" is identification information that uniquely identifies the teacher, and specifically, the evaluator's ID. "Gender" is the evaluator's gender, and "year of birth" is the year the evaluator was born. Note that age may be used instead of year of birth. "Area of ​​residence" indicates the area in which the evaluator resides, which in this example is classified as urban or suburban. "Driving history" is the evaluator's driving history, and in this example, the number of years of driving experience is used.

[0021] "Environment information" is information indicating the environment when each image data was captured, and includes weather, road type, and time of day. "Weather" is the weather when the image data was captured, and examples include sunny, cloudy, rainy, and snowy. "Road type" is the type of road included in the image data, i.e., the type of road the vehicle was traveling on when the image was captured, and examples include expressway, national highway, prefectural road, and narrow street. "Time of day" indicates the time of day when the image data was captured.

[0022] The "judgment value" is the result of the judge's evaluation of the image data, and is the result of the judge's evaluation of the driving environment shown in the image data. In this example, the judgment value is assigned two values, "difficult to drive" and "easy to drive," but the ease of driving may be classified into three or five levels, and three or five judgment values ​​may be used.

[0023] In the example shown in Figure 2, the results of three judges with teacher identifiers "K1116," "H1117," and "T1110" each judging the ease of driving for two pieces of image data are stored as teacher data.

[0024] Returning to FIG. 1 , the learning data control unit 13 generates learning data based on the image data stored in the image data accumulation unit 11 and the teacher data stored in the teacher data management unit 12, and supplies the generated learning data to the machine learning unit 14. "Learning data" refers to data input to the machine learning unit 14 and used for machine learning, and includes teacher data and learning data. Here, "teacher data" refers to the judgment value stored in the teacher data management unit 12, and "learning data" refers to image data corresponding to the judgment value. In other words, the learning data is data that links image data with the judgment value of the judge for that image data. As will be described in detail later, the learning data control unit 13 generates teacher models by grouping the teacher data based on commonalities in their profiles or environmental information, and generates learning data for each teacher model and supplies it to the machine learning unit 14.

[0025] The machine learning unit 14 performs supervised learning using the learning data supplied from the learning data control unit 13. Then, the machine learning unit 13 generates, as a learning result, a model (hereinafter referred to as a "driving evaluation model") for determining whether the driving environment indicated by the image data is easy to drive in. In other words, the driving evaluation model is a pattern classifier obtained by supervised machine learning, and is a learning result obtained by learning using a neural network, for example.

[0026] Here, as described above, the learning data control unit 13 generates learning data for each teacher model and supplies it to the machine learning unit 14, and the machine learning unit 14 generates a learning result for each teacher model and stores it as a driving evaluation model in the driving evaluation model DB 15. For example, in the case of learning by deep learning, the driving evaluation model is made up of network data, a teacher model (a profile of the teacher used to create the teacher model), environmental data (parameters of environmental information used during learning), etc.

[0027] In the above configuration, the image data storage unit 11 is an example of a first memory unit of the present invention, the teacher data management unit 12 is an example of a second memory unit of the present invention, and the learning data control unit 13 and the machine learning unit 14 are examples of a generation unit of the present invention.

[0028] [Learning process] Next, the learning process performed by the learning system 10 will be described. (First Example) In the learning process of the first embodiment, a teacher model is set by specifying a profile related to a residential area, and a driving evaluation model corresponding to the teacher model is created.

[0029] Figure 3 is a flowchart of the learning process according to Example 1. This process is mainly executed by the learning data control unit 13 and machine learning unit 14 of the learning system 10 shown in Figure 1. If the learning system 10 is configured with a computer device such as a server, the learning process is realized by the computer device executing a program prepared in advance.

[0030] First, the learning data control unit 13 creates a teacher model by grouping teachers based on commonalities in their residential area profiles from the teacher data stored in the teacher data management unit 12 (step S11). In the example of FIG. 2, gender, year of birth, residential area, and driving history are stored as profiles. The learning data control unit 13 creates a group of evaluators who share a common residential area profile from among these. For example, the learning data control unit 13 creates a group called "urban residents" from the teacher data of evaluators who live in urban areas, and uses this as a teacher model.

[0031] Next, the learning data control unit 13 acquires image data and judgment values ​​for the residential area and a different or opposite area from the residential area from the teacher data belonging to the teacher model created in step S11 (step S12). In the above example, since the teacher model "urban resident" is created, the learning data control unit 13 acquires, as the image data and judgment values ​​to be used as learning data, not only image data taken in the urban area where the teacher model "urban resident" lives, but also image data and judgment values ​​taken in the "suburban area" which is a different or opposite area from the urban area. Now, in the example of Figure 2, it is assumed that the images with image data IDs "107" and "109" are image data taken in urban areas, and the images with image data IDs "108" and "110" are image data taken in suburban areas. In this case, for the teacher model "urban resident," the learning data control unit 13 acquires the teacher data of two judges (teacher identifiers = L1116 and H1117) who are urban residents, including not only image data taken in urban areas (image data IDs = "107," "109") and their judgment values, but also image data taken in the suburbs (image data IDs = "108," "110") and their judgment values.

[0032] Then, the learning data control unit 13 stores the acquired image data and the judgment value as learning data for the teacher model (step S13). In this way, the learning data to be used for the teacher model created in step S11 is prepared.

[0033] The learning data created in this way is supplied to the machine learning unit 14, and the machine learning unit 14 performs machine learning using the learning data (step S14). Then, the machine learning unit 14 saves the learning result obtained by the machine learning (step S15). Specifically, the learning result obtained by the machine learning becomes a driving evaluation model corresponding to the teacher model created in step S11. That is, in the above example, a driving evaluation model of the teacher model "urban resident" is obtained as the learning result. Then, the machine learning unit 14 saves this driving evaluation model in the driving evaluation model DB15.

[0034] In step S13, when image data and judgment values ​​are acquired for teachers (judges) belonging to the teacher model to generate learning data, different judgment values ​​may be obtained from judges belonging to the same teacher model. For example, there may be a case where judgment values ​​for the same image data by judges belonging to the teacher model "urban resident" are divided into "easy to drive" and "difficult to drive." In this case, the obtained image data and judgment values ​​cannot be used as they are in machine learning processing, in which image data and judgment values ​​are input one-to-one as learning data. Therefore, the learning data control unit 13 unifies the judgment values ​​corresponding to the same image data into one using the following method.

[0035] The learning data control unit 13 basically determines one judgment value by majority vote. That is, the learning data control unit 13 counts the number of times that judgment values ​​appear for the same image data, and adopts the judgment value that appears most frequently as the judgment value for that image data. In the above example, the number of judgment values ​​"difficult to drive" and "easy to drive" for the same image by evaluators belonging to the teacher model "urban resident" is counted, and the judgment value with the most votes is adopted as the judgment value for that image data.

[0036] It is also possible to add a reliability judgment to the above majority voting method to determine a single judgment value for the same image data. Specifically, if there is not much difference in the number of occurrences of different judgment values, the accuracy may be insufficient, so the image data and its corresponding judgment value may not be adopted as learning data.

[0037] Alternatively, the ratio of the number of occurrences may be used to determine whether the most frequent occurrence is different enough to be recognized as superior compared to the number of occurrences of other judgment values, and the judgment value of the most frequent occurrence may be adopted only if such a difference is found. For example, the judgment value of the most frequent occurrence may be adopted when conditions are met, such as the most frequent occurrence being 70% or more of the total, or the most frequent occurrence being 10% or more greater than the second most frequent occurrence.

[0038] In this way, in the first embodiment, when a teacher model is set based on a residential area profile and machine learning is performed, not only image data and its evaluation information obtained in the teacher model's residential area, but also image data and its evaluation information obtained in a different or opposite area are used as learning data. This makes it possible to generate a general-purpose driving evaluation model that exhibits appropriate behavior regardless of the area in which the vehicle is traveling. In other words, the generated driving evaluation model for an "urban resident" will exhibit appropriate behavior not only when driving in urban areas, but also when driving in suburban areas.

[0039] (Second Example) In the learning process of the second embodiment, a teacher model is set based on environmental information, and a driving evaluation model corresponding to the teacher model is created.

[0040] Figure 4 is a flowchart of the learning process according to Example 2. This process is mainly executed by the learning data control unit 13 and machine learning unit 14 of the learning system 10 shown in Figure 1. If the learning system 10 is configured with a computer device such as a server, the learning process is realized by the computer device executing a program prepared in advance.

[0041] First, the learning data control unit 13 creates a teacher model by grouping teachers based on commonality of environmental information from the teacher data stored in the teacher data management unit 12 (step S21). For example, the learning data control unit 13 creates a group called "raining" from teacher data in which the weather is "rainy," and uses this as a teacher model.

[0042] Next, the learning data control unit 13 acquires image data and a judgment value from the training data belonging to the training model "rainfall" created in step S21 (step S12). In the above example, the learning data control unit 13 acquires image data and a judgment value for the training data for which the weather is "rain" from among the training data shown in FIG.

[0043] Furthermore, the learning data control unit 13 also acquires image data and judgment values ​​for an environment different from or opposite to the environment of the teacher model (step S23). In the above example, the learning data control unit 13 acquires image data and judgment values ​​from the teacher data for "sunny" and "cloudy", which are environments different from or opposite to the environment "rainy" of the teacher model.

[0044] Then, the learning data control unit 13 stores the acquired image data and the judgment value as learning data for the teacher model (step S24). In this way, the learning data to be used for the teacher model created in step S21 is prepared.

[0045] The learning data created in this way is supplied to the machine learning unit 14, and the machine learning unit 14 performs machine learning using the learning data (step S25). Then, the machine learning unit 14 saves the learning result obtained by the machine learning (step S26). Specifically, the learning result obtained by the machine learning becomes a driving evaluation model corresponding to the teacher model created in step S21. That is, in the above example, a driving evaluation model of the teacher model "during rainfall" is obtained as the learning result. Then, the machine learning unit 14 saves this driving evaluation model in the driving evaluation model DB15.

[0046] In this way, in the second embodiment, when generating a driving evaluation model for the teacher model "when it's raining" related to the weather "rain" which is environmental information, image data and judgment values ​​for "sunny" and "cloudy" environments which are different or opposite to the environment are also used as learning data. This diversifies the learning data used by the machine learning unit 14, making it possible to generate a highly versatile driving evaluation model. In other words, the generated driving evaluation model for "when it's raining" will exhibit appropriate behavior not only when it's raining, but also when driving on sunny or cloudy days.

[0047] "Different environments" in environmental information include weather such as "clear," "cloudy," "rainy," and "snowy," road types such as "expressway," "national highway," "prefectural road," and "narrow street," and time periods such as "morning," "afternoon," and "night." "Opposite environments" include weather such as "clear" and "rainy" or "snowy," road types such as "expressway" or "national highway" and "narrow street," and time periods such as "morning," "afternoon," and "night."

[0048] [Variations] In the above embodiment, the teacher model is set based on residential area or environmental information, but the teacher model may also be set based on a combination of residential area and environmental information, or a combination of multiple environmental information. For example, teacher models such as "urban resident (rainy)" and "clear night" may be set. Here, for the teacher model "urban resident (rainy)," the teacher data is for an "urban area" residential area, and not only image data taken in urban areas on rainy days but also image data taken in urban and suburban areas on clear or cloudy days are used as training data. Furthermore, for the teacher model "clear night," not only image data taken on clear nights but also image data taken on rainy or cloudy mornings or nights are used as training data. [Explanation of symbols]

[0049] 10 Learning System 11 Image data storage unit 12 Teacher Data Management Department 13 Learning data control section 14 Machine Learning Department 15 Driving evaluation model database

Claims

1. a first storage unit that stores image information of a location where the moving object travels; a second storage unit that stores evaluation information regarding the driving evaluated by an evaluator based on the image information in association with a profile of the evaluator; a generating unit that, when generating a driving evaluation model corresponding to one evaluation condition, generates the driving evaluation model using image information and evaluation information when the one evaluation condition is met, as well as image information and evaluation information when the one evaluation condition is not met; A driving evaluation model generation device comprising:

2. The one evaluation condition includes information about a region, 2. The driving evaluation model generation device according to claim 1, wherein, when generating a driving evaluation model corresponding to a certain region, the generation unit generates the driving evaluation model using image information and evaluation information obtained in the certain region, as well as image information and evaluation information obtained in a region other than the certain region.

3. A driving evaluation model generation device as described in Claim 2, wherein when generating a driving evaluation model corresponding to either an urban area or a suburban area, the generation unit generates the driving evaluation model using image information and evaluation information obtained in the one area, as well as image information and evaluation information obtained in another area different from the one area.

4. The one evaluation condition includes information about the environment of the moving body, The driving evaluation model generation device of claim 1, wherein when generating a driving evaluation model corresponding to a certain environment, the generation unit generates the driving evaluation model using image information and evaluation information obtained in the certain environment, as well as image information and evaluation information obtained in an environment different from the certain environment.

5. When generating a driving evaluation model corresponding to either a rainy or non-rainy environment, the generating unit generates the driving evaluation model using image information and evaluation information obtained in the one environment, as well as image information and evaluation information obtained in the other environment different from the one environment, The driving evaluation model generating device according to claim 4 , wherein the non-rainy environment is either a sunny environment or a cloudy environment.

6. A driving evaluation model generation method executed by a driving evaluation model generation device, comprising: a first acquisition step of acquiring image information of a location where the moving object is traveling; a second acquisition step of acquiring evaluation information regarding the driving evaluated by an evaluator with respect to the image information in association with a profile of the evaluator; a generation step of generating a driving evaluation model corresponding to one evaluation condition using image information and evaluation information when the one evaluation condition is met, as well as image information and evaluation information when the one evaluation condition is not met; A driving evaluation model generation method comprising:

7. A program executed by an apparatus having a computer, a first acquisition step of acquiring image information of a location where the moving object is traveling; a second acquisition step of acquiring evaluation information on the driving evaluated by an evaluator with respect to the image information in association with a profile of the evaluator; a generation step of generating a driving evaluation model corresponding to one evaluation condition using image information and evaluation information when the one evaluation condition is met, as well as image information and evaluation information when the one evaluation condition is not met; A program that causes the computer to execute the above.

8. A storage medium storing the program described in claim 7.

Images