Vehicle management system and vehicle management method

Abstract

To estimate a vehicle position with high accuracy, while also reducing effort and cost.SOLUTION: A vehicle management system manages a vehicle within a predetermined area. The vehicle management system holds a machine learning model that estimates a position of the vehicle as shown in an image. The vehicle management system acquires a target-specific parameter, which is a parameter of a machine learning model trained specifically for a category of a target vehicle, from a parameter providing device. The vehicle management system acquires a target-specific machine learning model that is specialized for a category of the target vehicle since the vehicle management system applies the target-specific parameter to the machine learning model. The vehicle management system acquires an image containing the target vehicle, which is captured by a camera installed in a predetermined area. The vehicle management system estimates the position of the target vehicle based on the image and the target-specific machine learning model.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present disclosure relates to a technology for managing vehicles in a predetermined area, and to a technology for estimating the position of a vehicle based on an image captured by a camera. [Background technology]

[0002] Patent Document 1 discloses a driving assistance control device that supports the driving of a vehicle to be controlled. The driving assistance control device determines the position of the vehicle to be controlled based on image information acquired by a camera. At this time, the driving assistance control device acquires vehicle identification information (vehicle dimensions, wheelbase length, tread width, etc.) from the vehicle to be controlled. The driving assistance control device then references the acquired vehicle identification information and determines the position of the vehicle to be controlled from the image information. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-57677 Summary of the Invention [Problem to be solved by the invention]

[0004] When estimating the location of a vehicle in an image captured by a camera, it is possible to use a machine learning model to improve the accuracy of the location estimation. However, creating a machine learning model that can handle all vehicle models with sufficient accuracy requires a great deal of effort and cost. In addition, the machine learning model needs to be updated every time a new vehicle model is released. [Means for solving the problem]

[0005] The first aspect relates to a vehicle management system that manages vehicles in a predetermined area. The vehicle management system is one or more processors; A storage device that stores a machine learning model for estimating the position of a vehicle in an image. Equipped with. The one or more processors acquire, from the parameter providing device, target-specific parameters, which are parameters of a machine learning model trained specifically for the category of the target vehicle. The one or more processors apply object-specific parameters to the machine learning model stored in the storage device to obtain an object-specific machine learning model specialized for the category of the target vehicle. The one or more processors acquire images of the target vehicle captured by a camera installed in a predetermined area. The one or more processors estimate a position of the target vehicle based on the images and the target-specific machine learning model.

[0006] The second aspect relates to a vehicle management method for managing vehicles in a predetermined area by a computer. Vehicle management methods include: Obtaining a machine learning model for estimating the location of a vehicle in an image; Acquiring target-specific parameters, which are parameters of a machine learning model trained specifically for a target vehicle category, from a parameter providing device; Applying subject-specific parameters to the machine learning model to obtain a subject-specific machine learning model specialized for the category of the subject vehicle; Acquiring an image of a target vehicle captured by a camera installed in a predetermined area; Estimating the position of a target vehicle based on images and a target-specific machine learning model Includes. [Effects of the Invention]

[0007] According to the present disclosure, object-specific parameters learned specifically for the category of the target vehicle are obtained from a parameter providing device. By applying the object-specific parameters to a machine learning model, an object-specific machine learning model specialized for the category of the target vehicle is obtained. The position of the target vehicle is then estimated based on the object-specific machine learning model. This makes it possible to estimate the position of the target vehicle with high accuracy. Furthermore, according to the present disclosure, a general-purpose machine learning model that can be used for all categories is not required. Since there is no need to generate or update a general-purpose machine learning model that can be used for all categories, effort and cost are significantly reduced. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a conceptual diagram for explaining an overview of a vehicle management system. [Figure 2] FIG. 1 is a conceptual diagram illustrating an example of vehicle management in a predetermined area. [Figure 3] FIG. 2 is a conceptual diagram for explaining a basic configuration related to a vehicle position estimation process performed by a vehicle management system. [Figure 4] FIG. 10 is a conceptual diagram for explaining a comparative example. [Figure 5] FIG. 1 is a block diagram for explaining a vehicle position estimation process using an object-specific machine learning model. [Figure 6] FIG. 1 is a conceptual diagram for explaining a subject-specific machine learning model. [Figure 7] FIG. 1 is a block diagram illustrating a first example of a parameter providing device. [Figure 8] FIG. 10 is a block diagram illustrating a second example of a parameter providing device. [Figure 9] FIG. 10 is a block diagram for explaining generation of object-specialized parameters. [Figure 10] FIG. 1 is a conceptual diagram for explaining a specific example of vehicle management by a vehicle management system. DETAILED DESCRIPTION OF THE INVENTION

[0009] Embodiments of the present disclosure will be described with reference to the accompanying drawings.

[0010] 1. Vehicle management system FIG. 1 is a conceptual diagram for explaining an overview of a vehicle management system 100 according to this embodiment. The vehicle management system 100 manages vehicles 1 in a predetermined area. Examples of the predetermined area include a parking lot, a factory, a facility site, and a city (smart city). The vehicle 1 may be an autonomous vehicle. The vehicle management system 100 includes, for example, a management server. The vehicle management system 100 may include multiple nodes that perform distributed processing.

[0011] According to this embodiment, one or more infrastructure cameras CAM installed in a predetermined area are used to manage the vehicle 1. The infrastructure camera CAM is installed so as to be able to capture images of at least the situation in the predetermined area. The infrastructure camera CAM captures an image of the predetermined area and acquires an image IMG showing the situation in the predetermined area.

[0012] The vehicle management system 100 communicates with the infrastructure camera CAM and acquires the image IMG captured by the infrastructure camera CAM. The vehicle management system 100 detects the vehicle 1 captured in the image IMG by analyzing the image IMG. The vehicle management system 100 also estimates the position of the vehicle 1 captured in the image IMG. This vehicle position estimation process will be described in detail later. Furthermore, the vehicle management system 100 manages the vehicle 1 in a predetermined area based on the position of the vehicle 1. The vehicle management system 100 may also manage the traveling of the vehicle 1 in a predetermined area based on the position of the vehicle 1.

[0013] The vehicle management system 100 includes one or more processors 110 (hereinafter simply referred to as processors 110), one or more storage devices 120 (hereinafter simply referred to as storage devices 120), and a communication device 130. The processor 110 executes various processes. Examples of the processor 110 include a general-purpose processor, a special-purpose processor, a central processing unit (CPU), a graphics processing unit (GPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), an integrated circuit, and / or a combination thereof. The storage device 120 stores various information. Examples of the storage device 120 include a hard disk drive (HDD), a solid-state drive (SSD), a volatile memory, a non-volatile memory, etc. The communication device 130 communicates with the outside via a communication network. For example, the communication device 130 communicates with an infrastructure camera CAM. The communication device 130 may also communicate with the vehicle 1.

[0014] The vehicle management program 140 is a computer program for managing vehicles 1 in a predetermined area. The functions of the vehicle management system 100 may be realized by cooperation between the processor 110 that executes the vehicle management program 140 and the storage device 120. The vehicle management program 140 is stored in the storage device 120. Alternatively, the vehicle management program 140 may be recorded on a computer-readable recording medium.

[0015] Fig. 2 is a conceptual diagram illustrating an example of vehicle management in a predetermined area. In the example shown in Fig. 2, the predetermined area is a parking lot. This parking lot provides an automated valet parking (AVP) service. Vehicle 1 is compatible with the automated valet parking service.

[0016] The in-vehicle device 10 is mounted on the vehicle 1 and controls the vehicle 1. Specifically, the in-vehicle device 10 recognizes the situation around the vehicle 1 using a recognition sensor (e.g., a camera) mounted on the vehicle 1. The in-vehicle device 10 drives the vehicle 1 while recognizing the situation around the vehicle 1. A plurality of markers M (landmarks) may be placed in the parking lot. The in-vehicle device 10 may recognize the markers M using an in-vehicle camera and perform localization processing to estimate the position of the vehicle 1 based on the recognition result of the markers M. The in-vehicle device 10 may automatically drive the vehicle 1 in the parking lot based on the estimated vehicle position.

[0017] The vehicle management system 100 manages automatic valet parking and each vehicle 1 in a parking lot (predetermined area). The vehicle management system 100 is capable of communicating with the in-vehicle device 10 of each vehicle 1 in the parking lot. One or more infrastructure cameras CAM are installed in the parking lot. The vehicle management system 100 acquires images IMG captured by the one or more infrastructure cameras CAM. The vehicle management system 100 may estimate the position of the vehicle 1 in the parking lot based on the images IMG. The vehicle management system 100 may manage each vehicle 1 based on the position of each vehicle 1. The vehicle management system 100 may remotely operate the vehicle 1 in the parking lot. For example, the vehicle management system 100 may remotely operate the vehicle 1 to reach a destination based on the position of the vehicle 1.

[0018] An example of the entry process is as follows: Vehicle 1 stops at a predetermined entry position. Vehicle management system 100 authenticates vehicle 1. Vehicle management system 100 also allocates an available parking space to vehicle 1. The allocated available parking space becomes the target parking space (destination) for vehicle 1 when entering the parking space. Furthermore, vehicle management system 100 estimates the current position of vehicle 1 and sets a target trajectory (target route) from the current position of vehicle 1 to the target parking space. Vehicle management system 100 issues an entry instruction to on-board device 10. In response to the entry instruction, on-board device 10 drives vehicle 1 to the target parking space according to the target trajectory. Alternatively, vehicle management system 100 may drive vehicle 1 to the target parking space according to the target trajectory.

[0019] An example of the leaving process is as follows: The destination at the time of leaving is the specified leaving position. As with the time of entering, a target trajectory from the current position of the vehicle 1 to the leaving position is set. The vehicle management system 100 issues a leaving instruction to the in-vehicle device 10. In response to the leaving instruction, the in-vehicle device 10 drives the vehicle 1 to the leaving position according to the target trajectory. Alternatively, the vehicle management system 100 may drive the vehicle 1 to the leaving position according to the target trajectory.

[0020] 2. Vehicle position estimation processing 2-1.Basic configuration 3 is a conceptual diagram for explaining the basic configuration related to the vehicle position estimation process by the vehicle management system 100. The vehicle management system 100 includes a position estimation unit 160 as a functional block. The position estimation unit 160 is realized by cooperation between the processor 110 and the storage device 120. Information required for the process is stored in the storage device 120.

[0021] The position estimation unit 160 acquires an image IMG captured by an infrastructure camera CAM. Assume that the image IMG shows a vehicle 1. The vehicle 1 shown in the image IMG is the target of position estimation. For convenience, the vehicle 1 shown in the image IMG is referred to as a target vehicle 1-X. The position estimation unit 160 detects the target vehicle 1-X shown in the image IMG. For this vehicle detection process, for example, a machine learning model trained to detect targets and classes from the image IMG is used. The position estimation unit 160 may recognize the model of the target vehicle 1-X. A bounding box may be attached around the detected target vehicle 1-X.

[0022] The position estimation unit 160 extracts feature points of the target vehicle 1-X detected in the image IMG. The feature points may also be called key points. Examples of feature points include parts of the vehicle 1, such as tires, headlights, license plates, side mirrors, doors, windows, and decorative parts. The feature points may also include tire contact points, corners of a bounding box, and the like. The position estimation unit 160 acquires the position (coordinates) and type of each feature point in the image IMG. The position estimation unit 160 may also acquire the size of each feature point in the image IMG. The feature point information is information related to the feature points extracted in the image IMG. For example, the feature point information indicates the position and type of each of multiple feature points in the image IMG, the positional relationship between multiple feature points, and the like.

[0023] Furthermore, the position estimation unit 160 estimates the position (coordinates) and orientation of the vehicle 1 in the image IMG based on the feature point information. The position of the vehicle 1 in the image IMG is a position corresponding to the reference point of the vehicle 1. The reference point is not particularly limited. For example, the reference point may be the center point of the vehicle 1.

[0024] A machine learning model MDL is used in the feature point extraction process and position estimation process in the position estimation unit 160. That is, the position estimation unit 160 includes a machine learning model MDL for estimating the position of the target vehicle 1-X shown in the image IMG. The machine learning model MDL may estimate the position and orientation of the target vehicle 1-X shown in the image IMG. The machine learning model MDL may use a neural network. The machine learning model MDL is generated in advance through learning and stored in the storage device 120. The position estimation unit 160 then estimates the position and orientation of the target vehicle 1-X by using the machine learning model MDL. Note that the position estimation unit 160 may use information on the vehicle specifications (vehicle length, vehicle width, wheelbase, etc.) of the target vehicle 1-X as supplementary information.

[0025] As a modified example, the position estimation unit 160 may first use a common machine learning model to extract common feature points (e.g., tires) common to various vehicles 1 from the target vehicle 1-X, and calculate the approximate position of the target vehicle 1-X based on the common feature points. Thereafter, the position estimation unit 160 may use a machine learning model MDL trained for each vehicle type to extract characteristic feature points specific to the vehicle model of the target vehicle 1-X, and calculate the detailed position of the target vehicle 1-X based on the characteristic feature points.

[0026] The infrastructure camera information includes installation information and performance information of the infrastructure camera CAM. The installation information includes the installation position and installation orientation of the infrastructure camera CAM in an absolute coordinate system (world coordinate system). The performance information includes the angle of view, focal length, etc. of the infrastructure camera CAM. The infrastructure camera information for each infrastructure camera CAM is registered in advance in the vehicle management system 100. Alternatively, the infrastructure camera information may be provided by the infrastructure camera CAM.

[0027] The position estimation unit 160 acquires infrastructure camera information related to the infrastructure camera CAM that captured the image of the target vehicle 1-X. Then, by using the infrastructure camera information, the position estimation unit 160 converts the position of the target vehicle 1-X in the image IMG into the position of the target vehicle 1-X in the absolute coordinate system. Alternatively, by using the infrastructure camera information, the position estimation unit 160 converts the position and orientation of the target vehicle 1-X in the image IMG into the position and orientation of the target vehicle 1-X in the absolute coordinate system.

[0028] In this way, it is possible to estimate the position and orientation of the target vehicle 1-X captured in the image IMG captured by the infrastructure camera CAM. By using the machine learning model MDL, it is possible to estimate the position and orientation of the target vehicle 1-X with high accuracy.

[0029] 2-2.Comparative Example FIG. 4 is a conceptual diagram for explaining a comparative example. In this comparative example, the machine learning model MDL included in the location estimation unit 160 is a general-purpose machine learning model that is compatible with all vehicle models and manufacturers. However, creating a general-purpose machine learning model that can be used with sufficient accuracy for all vehicle models and manufacturers requires a great deal of effort and costs. Furthermore, every time a new vehicle model appears, the general-purpose machine learning model needs to be updated. This also leads to increased effort and costs.

[0030] 2-3. Target-specific vehicle position estimation processing FIG. 5 is a block diagram for explaining the vehicle position estimation process according to this embodiment.

[0031] The parameters of the MDL machine learning model define the characteristics of the MDL machine learning model, including, for example, weights that represent the strength of connections between nodes in the neural network.

[0032] The object-specific parameters PA-X are parameters of the machine learning model MDL that have been trained to be specialized for the category X of the object vehicle 1-X. Examples of the category X include the vehicle type, the manufacturer, etc. According to this embodiment, the object-specific parameters PA-X that have been trained to be specialized for the category X are prepared in advance. An example of generating the object-specific parameters PA-X will be described later in Section 4.

[0033] The parameter providing device 200 holds a target-specialized parameter PA-X prepared in advance. The parameter providing device 200 is capable of communicating with the vehicle management system 100 and provides the target-specialized parameter PA-X to the vehicle management system 100. The parameter providing device 200 may be an in-vehicle device 10 mounted on the target vehicle 1-X, or may be a predetermined management server.

[0034] According to this embodiment, the position estimation unit 160 includes a base machine learning model MDL-0. The base machine learning model MDL-0 has the same structure as the machine learning model MDL. The base machine learning model MDL-0 may be the machine learning model MDL before learning. Furthermore, the position estimation unit 160 acquires, from the parameter providing device 200, object-specific parameters PA-X that have been trained specifically for the category X of the target vehicle 1-X. The position estimation unit 160 acquires the object-specific machine learning model MDL-X specialized for the category X by applying the object-specific parameters PA-X to the base machine learning model MDL-0. In other words, the position estimation unit 160 acquires the object-specific machine learning model MDL-X specialized for the category X by replacing the parameters of the base machine learning model MDL-0 with the object-specific parameters PA-X. In yet other words, the position estimation unit 160 acquires the object-specific machine learning model MDL-X specialized for the category X by combining the base machine learning model MDL-0 and the object-specific parameters PA-X.

[0035] FIG. 6 shows various object-specific machine learning models MDL-X. For example, by combining object-specific parameters PA-A of category A with the base machine learning model MDL-0, an object-specific machine learning model MDL-A specialized for category A is obtained. By combining object-specific parameters PA-B of category B with the base machine learning model MDL-0, an object-specific machine learning model MDL-B specialized for category B is obtained. The object-specific machine learning model MDL-A specialized for category A and the object-specific machine learning model MDL-B specialized for category B are different from each other. The object-specific machine learning model MDL-A specialized for category A can estimate the position of at least the target vehicle 1-A of category A with high accuracy, but it is not necessarily capable of estimating the positions of target vehicles of other categories with high accuracy. Similarly, the object-specific machine learning model MDL-B specialized for category B can estimate the position of at least the target vehicle 1-B of category B with high accuracy, but it is not necessarily capable of estimating the positions of target vehicles of other categories with high accuracy.

[0036] According to this embodiment, the position estimation unit 160 estimates the position and orientation of the target vehicle 1-X by using an object-specific machine learning model MDL-X specialized for category X. This makes it possible to estimate the position and orientation of the target vehicle 1-X with high accuracy.

[0037] After the position of the target vehicle 1-X is estimated, the position estimation unit 160 may erase at least one of the object-specific parameter PA-A and the object-specific machine learning model MDL-X. The position estimation unit 160 may erase both the object-specific parameter PA-A and the object-specific machine learning model MDL-X. This reduces the amount of data used by the storage device 120 in the vehicle management system 100.

[0038] 2-4.Effects As described above, according to this embodiment, the object-specific parameters PA-X, which have been learned specifically for the category X of the target vehicle 1-X, are obtained from the parameter providing device 200. By applying the object-specific parameters PA-X to the base machine learning model MDL-0, an object-specific machine learning model MDL-X specialized for the category X of the target vehicle 1-X is obtained. Then, the position of the target vehicle 1-X is estimated based on the object-specific machine learning model MDL-X. This makes it possible to estimate the position of the target vehicle 1-X with high accuracy.

[0039] Furthermore, according to this embodiment, a general-purpose machine learning model that can be used in all categories is not required. Since there is no need to generate or update a general-purpose machine learning model that can be used in all categories, the effort and cost are significantly reduced.

[0040] After the position of the target vehicle 1-X is estimated, at least one of the target-specific parameters PA-A and the target-specific machine learning model MDL-X may be erased (deleted) from the vehicle management system 100. This allows the amount of space used by the storage device 120 of the vehicle management system 100 to be reduced.

[0041] 3. Example of a parameter provider 3-1. First example 7 is a block diagram illustrating a first example of the parameter providing device 200. In the first example, the parameter providing device 200 is an in-vehicle device 10 mounted on a target vehicle 1-X. The in-vehicle device 10 holds an object-specified parameter PA-X related to category X of the target vehicle 1-X. For example, the object-specified parameter PA-X is stored in the ECU of the in-vehicle device 10. The object-specified parameter PA-X may be stored in the ECU by the manufacturer when the target vehicle 1-X is manufactured.

[0042] The vehicle management system 100 communicates with the in-vehicle device 10 of the target vehicle 1-X and acquires the target-specific parameters PA-X from the in-vehicle device 10. Then, the vehicle management system 100 estimates the position and orientation of the target vehicle 1-X using the target-specific parameters PA-X.

[0043] According to the first example, the object-specific parameters PA-X specific to the object vehicle 1-X can be obtained from the object vehicle 1-X itself, which is efficient.

[0044] 3-2. Second example 8 is a block diagram illustrating a second example of the parameter providing device 200. In the second example, the parameter providing device 200 is a parameter management server 20. The parameter management server 20 holds multiple types of object-specific parameters PA-A, PA-B, etc. The multiple types of object-specific parameters PA-A, PA-B, etc. are parameters of a machine learning model MDL that has been trained to be specialized for each of multiple categories A, B, etc.

[0045] The vehicle management system 100 acquires information on category X of the target vehicle 1-X from the target vehicle 1-X. Furthermore, the vehicle management system 100 communicates with the parameter management server 20 and selectively acquires an object-specializing parameter PA-X specialized for category X of the target vehicle 1-X from multiple types of object-specializing parameters PA-A, PA-B, etc. The vehicle management system 100 then estimates the position and orientation of the target vehicle 1-X using the object-specializing parameter PA-X.

[0046] According to the second example, it is possible to collectively manage a plurality of types of object-specialized parameters PA-A, PA-B, etc. Also, it is possible to reduce the processing load on the object vehicle 1-X.

[0047] 4.Generating subject-specific parameters FIG. 9 is a block diagram for explaining generation of the object-specific parameter PA-X. The model learning system 300 acquires learning data specialized for category X. The learning data specialized for category X includes many combinations of learning images showing vehicles 1 of category X and position information of the vehicles 1. For example, the learning images are obtained by photographing one or more vehicles 1 of category X from various directions. As another example, the learning images may be generated by a simulation using CAD data of a vehicle 1 of category X. The position information may be acquired by actual measurement or calculated by simulation.

[0048] The model learning system 300 holds a base machine learning model MDL-0. The model learning unit 310 trains the base machine learning model MDL-0 by using learning data specialized for category X, thereby generating a subject-specific machine learning model MDL-X specialized for category X. The parameters of the generated subject-specific machine learning model MDL-X are subject-specific parameters PA-X. The model learning system 300 provides the generated subject-specific parameters PA-X to the parameter providing device 200.

[0049] Note that the entity that generates the object specialization parameter PA-X is arbitrary. For example, a vehicle manufacturer that manufactures a vehicle 1 of category X may generate the object specialization parameter PA-X for category X. Then, the vehicle manufacturer may store information on the object specialization parameter PA-X in the in-vehicle device 10 (see FIG. 7) of the vehicle 1 that it manufactures. As another example, a business that provides an AVP service may generate multiple types of object specialization parameters PA-A, PA-B, etc. Then, the business may store multiple types of object specialization parameters PA-A, PA-B, etc. in the parameter management server 20 (see FIG. 8).

[0050] A user of vehicle 1 may customize his or her vehicle 1. For example, the user may attach additional decorative parts to his or her vehicle 1. In this case, object-specific parameters PA-X specialized for the user's vehicle 1 may be generated. For example, the user may provide the model learning system 300 with information about category X of the user's vehicle 1 and information about the customized vehicle appearance. The information about the vehicle appearance may be, for example, images of the customized vehicle 1 taken from various angles. The model learning system 300 performs the above-described learning process based on the information received from the user. As a result, an object-specific machine learning model MDL-X specialized for the user's customized vehicle appearance in addition to category X is obtained. The parameters of the object-specific machine learning model MDL-X thus obtained are object-specific parameters PA-X specialized for the user's vehicle 1. The model learning system 300 may provide the object-specific parameters PA-X specialized for the user's vehicle 1 to the in-vehicle device 10 (parameter providing device 200) of the user's vehicle 1. In this way, it is possible to accommodate user customization of the vehicle 1.

[0051] 5. Specific examples of vehicle management FIG. 10 is a conceptual diagram for explaining a specific example of vehicle management by the vehicle management system 100. Here, consider a scene in which the automatic valet parking described in FIG. 2 is performed. The predetermined area is a parking lot. Also, assume that the in-vehicle device 10 of the target vehicle 1-X holds a target-specified parameter PA-X (see FIG. 7).

[0052] The target vehicle 1-X arrives at the entrance position of the parking lot and stops there. An infrastructure camera CAM is installed at the entrance position. The vehicle management system 100 acquires an image IMG captured by the infrastructure camera CAM. The image IMG shows the target vehicle 1-X.

[0053] The vehicle management system 100 performs authentication processing. Specifically, the vehicle management system 100 establishes communication with the in-vehicle device 10 of the target vehicle 1-X. The vehicle management system 100 also authenticates the target vehicle 1-X. During this authentication processing stage, the vehicle management system 100 communicates with the in-vehicle device 10 of the target vehicle 1-X and acquires an object-specifying parameter PA-X from the in-vehicle device 10. The vehicle management system 100 then estimates the position, etc. of the target vehicle 1-X based on the acquired object-specifying parameter PA-X and image IMG.

[0054] The vehicle management system 100 manages the travel of the target vehicle 1-X in the parking lot based on the estimated position of the target vehicle 1-X, etc. For example, after the authentication process is completed, the vehicle management system 100 allocates an available parking space to the target vehicle 1-X. The allocated available parking space becomes the target parking space (destination). Furthermore, the vehicle management system 100 sets a target trajectory (target route) from the current position of the target vehicle 1-X to the target parking space. The vehicle management system 100 issues a parking entry instruction to the in-vehicle device 10. In response to the parking entry instruction, the in-vehicle device 10 drives the target vehicle 1-X to the target parking space according to the target trajectory.

[0055] As described above, in the authentication stage before the target vehicle 1-X starts traveling in the parking lot (predetermined area), the vehicle management system 100 acquires the target-specific parameter PA-X and estimates the position of the target vehicle 1-X. Therefore, after the authentication is completed, the target vehicle 1-X can quickly start traveling.

[0056] The target vehicle 1-X traveling toward the target parking space may be photographed by the infrastructure camera CAM. The vehicle management system 100 acquires an image IMG showing the traveling target vehicle 1-X from the infrastructure camera CAM. The vehicle management system 100 estimates the position, etc., of the traveling target vehicle 1-X based on the target-specific parameter PA-X and the image IMG. The vehicle management system 100 may check whether the target vehicle 1-X is traveling along the target trajectory based on the estimated position. The vehicle management system 100 may remotely control the target vehicle 1-X to cause the target vehicle 1-X to travel to the target parking space according to the target trajectory.

[0057] The target vehicle 1-X arrives at and stops at the target parking space. After the target vehicle 1-X has completed its journey through the parking lot, the vehicle management system 100 may erase at least one of the object-specific parameters PA-A and the object-specific machine learning model MDL-X related to the target vehicle 1-X. The vehicle management system 100 may erase both the object-specific parameters PA-A and the object-specific machine learning model MDL-X. This reduces the amount of data used by the storage device 120 of the vehicle management system 100. [Explanation of symbols]

[0058] 1 vehicle 1-X Target vehicle 10 Onboard equipment 100 Vehicle Management System 110 processors 120 Storage device 130 Communication equipment 140 Vehicle Management Program 160 Position estimation part MDL-0 based machine learning model MDL-X Subject-Specific Machine Learning Model PA-X Target Specific Parameters

Claims

1. A vehicle management system for managing vehicles in a predetermined area, one or more processors; A storage device that stores a machine learning model for estimating the position of a vehicle in an image. Equipped with the one or more processors: Obtaining target-specific parameters, which are parameters of the machine learning model trained specifically for the category of the target vehicle, from a parameter providing device; applying the object-specific parameters to the machine learning model stored in the storage device to obtain an object-specific machine learning model specialized for the category of the target vehicle; Acquire an image of the target vehicle captured by a camera installed in the predetermined area, Estimating a position of the target vehicle based on the image and the target-specific machine learning model. It was configured as Vehicle management system.

2. The vehicle management system according to claim 1, The one or more processors further delete at least one of the object-specific parameters and the object-specific machine learning model from the vehicle management system after estimating the position of the object vehicle. Vehicle management system.

3. The vehicle management system according to claim 1, The one or more processors further manage the travel of the target vehicle in the predetermined area based on the estimated position of the target vehicle. It was configured as Vehicle management system.

4. The vehicle management system according to claim 3, The one or more processors acquire the target-specific parameters from the parameter providing device before the target vehicle starts traveling in the predetermined area. Vehicle management system.

5. The vehicle management system according to claim 3, The one or more processors delete at least one of the object-specific parameters and the object-specific machine learning model from the vehicle management system after the travel of the object vehicle in the predetermined area is completed. Vehicle management system.

6. The vehicle management system according to claim 1, the parameter providing device is an in-vehicle device mounted on the target vehicle, The one or more processors communicate with the in-vehicle device and acquire the object-specific parameters from the in-vehicle device. Vehicle management system.

7. The vehicle management system according to claim 6, The one or more processors acquire the target-specific parameters from the in-vehicle device during an authentication phase of the target vehicle. Vehicle management system.

8. The vehicle management system according to claim 1, the parameter providing device holds a plurality of types of object-specific parameters, which are parameters of the machine learning model trained to be specialized for each of a plurality of categories; The one or more processors selectively acquire the object-specializing parameter specialized for the category of the target vehicle from the plurality of types of object-specializing parameters. Vehicle management system.

9. The vehicle management system according to claim 1, The subject-specific parameters are parameters of the machine learning model that are trained specifically for the category and the vehicle appearance customized by the user. Vehicle management system.

10. A vehicle management method for managing vehicles in a predetermined area by a computer, comprising: Obtaining a machine learning model for estimating the location of a vehicle in an image; Obtaining target-specific parameters, which are parameters of the machine learning model trained specifically for the category of the target vehicle, from a parameter providing device; applying the object-specific parameters to the machine learning model to obtain an object-specific machine learning model specialized for the category of the target vehicle; acquiring an image of the target vehicle captured by a camera installed in the predetermined area; estimating a position of the target vehicle based on the image and the target-specific machine learning model; and Contains Vehicle management methods.

Images