Drive recorder, program, learning model generation method, learned model, system, and the like

Abstract

To provide a technology for making it possible to easily operate a drive recorder, for example.SOLUTION: A drive recorder 1 comprises: a camera; a control unit 14 for performing control on video recording on the basis of movement of a wiper in video captured by the camera 10; and a video storage unit 15 for storing video.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a drive recorder, a program, a learning model generation method, a learned model, and a system. [Background technology]

[0002] Conventionally, drive recorders are attached to the windshields of private cars, commercial cars, etc., and capture images of the area in front of the vehicle for a specified period of time before and after an accident, for example, to provide useful information for investigating traffic accidents, etc.

[0003] A drive recorder having an event recording function is also known (Patent Document 1). By pressing a switch for executing event recording, it is possible to record video and audio of a vehicle passing a given point for a certain period of time with a single touch. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2018-196066 A Summary of the Invention [Problem to be solved by the invention]

[0005] A drive recorder may be installed in an inconspicuous place such as behind the rearview mirror. In this case, it is difficult for the driver to press the switch of the drive recorder, and it is difficult to take a picture at the desired timing. If the drive recorder can communicate with a smartphone or the like via Wi-Fi, it may be possible to give an instruction to record by placing the smartphone within reach of the user, but such an environment may not be possible.

[0006] In view of the above background, one object of the present invention is to provide a technique for enabling a driver to easily operate, for example, a drive recorder.

[0007] In addition, the purpose of the present invention is not limited to this, and the applicant intends to obtain rights for configurations that aim to obtain effects from parts of the configurations disclosed in this specification and drawings, etc., by filing a divisional application, amendment, etc. For example, this specification discloses a problem in which the part described as "can be" in this specification is read as "the problem is to be". The problems are described as independent, and the applicant intends to obtain rights for the configurations for solving each problem separately by filing a divisional application, amendment, etc. Even if the problems are implicitly understood from the description in the specification, the applicant intends to include part of the configurations described in this specification in the scope of the patent claim by amendment or divisional application. In addition, the applicant has disclosed a configuration that solves a problem that combines these independent problems, and intends to obtain rights for it. [Means for solving the problem]

[0008] (1) A drive recorder according to one aspect of the present invention includes a camera and a control unit that controls recording based on movement associated with a part of a vehicle in an image captured by the camera.

[0009] In this way, the motion related to the parts of the vehicle in the video can be used to give the drive recorder control instructions for recording, so that the drive recorder can be easily operated regardless of the location where the drive recorder is installed. The control related to recording may include starting and ending recording. The motion related to the parts of the vehicle may include the motion of the parts of the vehicle, or the motion of a person (e.g., the driver) relative to the parts of the vehicle. Furthermore, by using the video captured by the camera of the drive recorder, no additional camera is required to capture the motion related to the parts of the vehicle.

[0010] (2) A drive recorder according to one embodiment of the present invention includes a model memory unit that stores a trained model for detecting a predetermined movement associated with a part of the vehicle that triggers control over the recording, and the control unit may apply an image captured by the camera to the model and, when the predetermined movement is detected, perform control over the recording.

[0011] By using such a trained model, it is possible to detect specific movements that trigger control over recording from the video footage captured by the drive recorder, and to operate the drive recorder from movements related to parts of the vehicle.

[0012] (3) The drive recorder according to one aspect of the present invention may include a learning unit that learns the model by using images captured during learning as teacher data associated with control over recording by the drive recorder.

[0013] In this way, it is possible to learn the specific movements that trigger control of recording, and therefore to learn the characteristics of movements related to vehicle parts that differ depending on the vehicle and person. If the movements related to vehicle parts are the movements of a person relative to a vehicle part, the specific movements can reflect the habits and preferences of the individual.

[0014] (4) In a drive recorder according to one embodiment of the present invention, when the specified movement is a movement instructing the start of recording, the control unit may store in the image storage unit video footage from a first specified time before the specified movement occurred.

[0015] In this way, by storing the video from a predetermined time before the predetermined movement, it is possible to store the scene that a person wants to shoot, even if it takes some time to perform the predetermined movement.

[0016] (5) In one embodiment of the drive recorder of the present invention, the drive recorder has a detection unit that detects an impact occurring to the vehicle, and when the impact is detected, the control unit stores video footage going back to a second predetermined time from the time of the detection, and the first predetermined time may be longer than the second predetermined time.

[0017] In this way, even if the reaction time to start recording due to movement of parts of the vehicle is slow, it is possible to record footage from a long time ago.

[0018] (6) In the drive recorder according to one aspect of the present invention, the movement associated with a part of the vehicle may be the operation of a windshield wiper.

[0019] The wiper lever is ergonomically optimized, so that the wiper movement can be used to safely operate the vehicle while it is moving. Regarding which wiper movement should be used as a trigger for control related to recording, a simple and short movement is preferable to make it easier to operate the drive recorder. Conversely, in order to avoid malfunction, some pattern of movement may be used as a trigger. In addition to the wiper movement, spraying of washer fluid may also be included. For example, a series of actions such as spraying washer fluid and then operating the wiper several times may be used as a trigger for control related to recording.

[0020] (7) In the drive recorder according to one aspect of the present invention, the movement associated with a part of the vehicle may be a human movement on an operating unit of the vehicle.

[0021] In this way, by using the human action on the vehicle's operation unit as a trigger for control related to recording, it is possible to safely control the recording of the drive recorder even while driving. The vehicle's operation unit is, for example, an operation unit that is pre-installed in the vehicle, and is often located within reach of the person in the vehicle so that the person in the vehicle can easily operate it. Therefore, the person in the vehicle can easily perform an operation to instruct the execution of control related to recording of the drive recorder. The human action on the vehicle's operation unit may include, for example, an action of holding a hand over a hazard lamp or an air conditioner switch. In this case, it is easy for the driver of the vehicle to operate these operation units. In order to use the human action on the vehicle's operation unit, it is necessary that the camera can capture the state inside the vehicle.

[0022] (8) In the drive recorder according to one aspect of the present invention, the movement associated with a part of the vehicle may be a movement relative to a mark having a pattern that is distinguishable from other parts of the vehicle.

[0023] By using a mark having a pattern that can be distinguished from other parts in the vehicle, the movement relative to the mark can be easily detected. The detection of the movement relative to the mark may use a trained model of a neural network. The detection of the movement relative to the mark may be performed by pattern matching. In this way, it is possible to easily detect that a mark matching the pattern has been hidden or appeared, and the movement relative to the mark can be detected. In this case, the drive recorder does not need to store a trained model of a neural network or execute a process using the model, and it is expected to suppress the complexity and high cost of the drive recorder. As the mark, a design using a reflective material or a hologram may be adopted. This is based on the inventor's knowledge that it is unlikely that a design identical or similar to the design exists in the vehicle, and it is expected to reduce the possibility of erroneously detecting a human movement relative to the mark. In order to use a human movement relative to the operating unit of the vehicle, the camera needs to be able to capture the state inside the vehicle.

[0024] (9) In the drive recorder according to one aspect of the present invention, the mark may be a sticker that is attached inside the vehicle.

[0025] By configuring the mark as a sticker in this way, the mark can be affixed in a location where movement relative to the mark can be easily performed.

[0026] (10) In the drive recorder according to one aspect of the present invention, the predetermined movement that triggers control over the recording may be the mark not being detectable for a certain period of time or longer.

[0027] The case where the mark cannot be detected is, for example, when a person's hand or the like gets between the mark and the camera and the mark is blocked by the person. If this state continues for a certain period of time or more, it is considered that the person is intentionally blocking the mark, and in such a case, it is determined that this is a predetermined movement that triggers control of the recording of the drive recorder. Since the condition is set to a certain period of time or more, it is possible to prevent the control of the recording from malfunctioning when a person unintentionally blocks the mark temporarily. The length of the certain period of time can be determined by a trade-off between the responsiveness of the control of the drive recorder and the prevention of malfunction.

[0028] (11) In one aspect of the drive recorder of the present invention, a plurality of marks having different designs may be provided, and the control unit may control the drive recorder depending on which type of mark has been detected as the movement.

[0029] By providing marks with different designs in this way, it becomes possible to have the drive recorder perform multiple controls.

[0030] (12) A drive recorder according to one embodiment of the present invention may include an audio input unit that accepts audio input, and the control unit may control recording by the drive recorder based on a combination of movement related to parts of the vehicle and audio.

[0031] By using a combination of the motion and the voice related to the vehicle part, the trigger of the control related to the recording can be detected with high accuracy. For example, even if the motion related to the vehicle part coincidentally meets the condition for executing the control related to the recording, the control related to the recording can be prevented from being executed erroneously. Conversely, if the input voice coincidentally meets the condition for executing the control related to the recording, the control is not executed. For example, when the voice "one-touch recording" is input at the same time as the predetermined action that triggers the control related to the recording, the drive recorder may start recording. If the voice is the content of the control related to the recording, people can easily understand what kind of control is executed by what kind of voice.

[0032] (13) In one embodiment of the drive recorder of the present invention, the vehicle is used by unspecified persons, and the control unit may control the recording of the drive recorder based on the results of learning from the images captured when each of a plurality of users uses the vehicle.

[0033] In this way, each of a plurality of users can control recording by the drive recorder based on the video suitable for recording.

[0034] (14) In the drive recorder of one aspect of the present invention, the control unit may notify a user based on the movement of a person wearing a helmet in the video captured by the camera.

[0035] In this way, a notification can be given according to the person wearing the helmet in the image captured by the camera.

[0036] (15) In the drive recorder according to one aspect of the present invention, the control unit may notify a user when it detects an abnormality on a road on which the vehicle is traveling.

[0037] In this way, the user can be notified in advance of possible danger caused by an abnormality on the road.

[0038] (16) In the drive recorder according to an aspect of the present invention, the control unit may notify a user when it detects that luggage has been left behind inside the vehicle.

[0039] In this way, the user can be notified that luggage has been left behind in a vehicle such as a taxi.

[0040] (17) In one aspect of the drive recorder of the present invention, the control unit may alert a user based on the image captured by the camera and a plurality of images stored in a memory area in which the images captured by the camera are accumulated.

[0041] In this way, it is possible to determine which images should be recorded by the drive recorder from among the images stored in the memory area because the user likes them or for other reasons, and to control the recording.

[0042] (18) A program according to one aspect of the present invention is a program for executing processing of the control unit of the drive recorder described in any one of (1) to (17) above.

[0043] By executing the processing of the drive recorder using such a program, it is possible to give the drive recorder control instructions regarding recording by utilizing movements related to parts of the vehicle in the video.

[0044] (19) A learning model generation method according to one embodiment of the present invention is a learning model generation method for generating a learning model for detecting a specified movement that triggers control of recording by a drive recorder from an image captured by the drive recorder, the method including the steps of: receiving an input instructing the drive recorder to learn the specified movement; and generating a learning model using the image captured during the learning as teacher data associated with control of recording by the drive recorder.

[0045] In this way, since the teacher data associated with the control of the drive recorder's recording is captured during learning, the driver can make the drive recorder learn the predetermined action that will trigger the control of the recording by performing the action that he / she wants to trigger. By using the learning model generated in this way, the predetermined action that will trigger the control of the drive recorder's recording can be detected by inputting data showing the video, so that the drive recorder can be easily operated.

[0046] (20) In one aspect of the learning model generation method of the present invention, video footage captured within a predetermined time from the start of the learning may be used as the teacher data.

[0047] By using the video captured within a certain time as teacher data in this way, the action within the certain time can be used as the trigger action for the control of the recording. By shortening this certain time, the trigger action for the control of the recording can be a short-time action, and by lengthening the certain time, the trigger action can be a long-time action.

[0048] (21) In one aspect of the learning model generation method of the present invention, the step of generating the learning model may generate a learning model by using video footage captured at times other than the learning time as teacher data associated with not controlling recording by a drive recorder.

[0049] In this way, by using images that do not trigger control over recording by the drive recorder as negative training data, it is possible to generate a trained model with high inference accuracy.

[0050] (22) A trained model of one embodiment of the present invention has an input layer to which data representing video is input, an output layer that outputs data related to a trigger for performing specified control over recording by the drive recorder, and at least one intermediate layer whose parameters are trained using teacher data that receives as input data representing video of movements related to parts of the vehicle captured during training and outputs data related to a trigger for performing specified control over recording by the drive recorder.

[0051] In this way, by inputting data showing an image into the input layer, data related to a trigger that controls recording by the drive recorder is output from the output layer, thereby providing a technology that allows easy operation of the drive recorder.

[0052] (23) A system according to one embodiment of the present invention has a control unit that performs predetermined control regarding recording of video captured by a camera of a vehicle based on movement associated with parts of the vehicle in the video captured by the camera.

[0053] In this way, since a control instruction for recording can be given by utilizing a movement related to a part of the vehicle in the video, operations related to recording can be easily performed. The control related to recording may include starting and ending recording. The movement related to the part of the vehicle may include the movement of the part of the vehicle, or may include the movement of a person (e.g., a driver) with respect to the part of the vehicle.

[0054] (24) A system according to one embodiment of the present invention includes a memory unit that stores a trained model and an acquisition unit that acquires data representing an image, and the control unit inputs the acquired data representing the image to the input layer, performs calculations in the intermediate layer, and outputs a signal for performing specified control regarding recording by a drive recorder based on data related to the trigger output from the output layer.

[0055] In this way, if data showing an image is input to the input layer, data related to a trigger that controls the recording of the drive recorder is output from the output layer, so that a technology can be provided that allows the drive recorder to be easily operated. The system of one aspect of the present invention is a system having a drive recorder, a personal computer, a server device, and other information processing functions, and may be composed of one or more devices. In addition, the subject that realizes each function may be any of hardware resources, software resources, and a combination of these.

[0056] (25) A program according to one aspect of the present invention is a program for executing processing of the control unit of the system described in (23) or (24) above.

[0057] The inventions shown in (1) to (25) above can be arbitrarily combined. For example, the invention shown in (1) may be combined with at least a portion of the configuration of at least one of the inventions from (2) onwards. In particular, the invention shown in (1) may be combined with at least a portion of the configuration of at least one of the inventions from (2) onwards. Any configuration may be extracted from the inventions shown in (1) to (25) and the extracted configurations may be combined. The applicant of this application intends to obtain rights to the inventions including these configurations. In addition, even if there is a description such as "in the case of" or "when", this is not intended to describe the configuration as being limited to that case or time. These are examples of better configurations, and the applicant intends to obtain rights to configurations other than these cases or times. In addition, the parts described in order are not limited to this order. Configurations in which some parts have been deleted or the order has been changed are also disclosed, and the applicant intends to obtain rights. Effect of the Invention

[0058] According to the present invention, for example, a technique for enabling a drive recorder to be easily operated can be provided. Furthermore, the effects of the invention of this application are not limited to this, and effects obtained from the configuration disclosed in the present specification and drawings, etc. are also disclosed, and the applicant intends to obtain rights to the configuration that achieves the effects by divisional applications, amendments, etc. For example, in this specification, the phrase "can" is a description that clearly indicates the effect that is achieved, and there are also parts that show the effect even without the phrase "can". Also, there are effects that can be understood from the configuration even without such a description. [Brief description of the drawings]

[0059] [Figure 1] FIG. 1 illustrates a configuration of a drive recorder according to a first embodiment. [Diagram 2] FIG. 1 is a diagram illustrating an example of an installation position of a drive recorder. [Diagram 3] 4 is a flowchart showing the recording operation of the drive recorder. [Figure 4] 11 is a flowchart showing an operation for generating a trained model. [Diagram 5] FIG. 13 is a diagram for explaining a mark. [Figure 6] FIG. 13 illustrates a configuration of a drive recorder according to a fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0060] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that the embodiment described below shows an example of the case where the present invention is implemented, and the present invention is not limited to the specific configuration described below. In implementing the present invention, a specific configuration according to the embodiment may be appropriately adopted. The drive recorder of this embodiment is a system having a function of capturing an image by a camera and a function of recording (recording) the captured image. The drive recorder of this embodiment performs a predetermined control regarding the recording of the image captured by the camera based on the movement related to the part of the vehicle in the image captured by the camera.

[0061] (Configuration of the first embodiment) Fig. 1 is a diagram showing the configuration of a drive recorder 1 according to a first embodiment. The drive recorder 1 includes a camera 10, a monitor 11, an acceleration sensor 12, various switches 13, a control unit 14, and an image storage unit 15. Note that the drive recorder 1 also includes components other than those described above, but Fig. 1 shows only the components related to the description of the embodiment.

[0062] 2(a) to 2(c) are diagrams showing examples of installation positions of the drive recorder 1. As shown in Fig. 2(a) to 2(c), the drive recorder 1 may be installed behind the rearview mirror so as not to be conspicuous. In such a case, various switches provided in the drive recorder 1 are hidden behind the rearview mirror, so that it may be difficult for the driver to operate the drive recorder 1 while driving.

[0063] 1 again, the configuration of the drive recorder 1 will be described. The camera 10 is a camera 10 that captures an image of the area in front of the vehicle. The camera 10 may also be a camera 10 that can capture an image of the interior of the vehicle. For example, the camera 10 may be a combination of a 360-degree camera and a rear camera.

[0064] The monitor 11 has a function of displaying the image captured by the camera 10. The acceleration sensor 12 detects the acceleration applied to the drive recorder 1, and outputs data indicating the detected acceleration to the control unit 14. The control unit 14 performs event recording when it detects a temporal change in a predetermined acceleration or a certain magnitude of acceleration, which is a trigger. Event recording is to record (video record) an image including the timing at which a predetermined event occurs. Acceleration is detected in order to perform event recording at the timing when an impact is applied to the vehicle. In this way, the acceleration sensor 12 functions as a detection unit that detects an impact that occurs to the vehicle. Event recording is to record (video record) an image including the timing. The "image including that timing" is, for example, a period including before and after the timing (for example, 10 seconds before and after). This image is an image captured by the camera 10.

[0065] The various switches 13 are switches for issuing operational instructions to the drive recorder 1. For example, there is a switch for starting and stopping recording. The drive recorder 1 has a function for performing event recording to automatically save video when a strong impact is detected, but the switch for starting recording is a switch that is operated when a person inside the vehicle (hereinafter, unless otherwise specified, referred to as the driver) wants to keep the video as a record even if there is no strong impact. Note that in this embodiment, recording means saving the video in the video storage unit 15.

[0066] The image storage unit 15 is a recording medium that stores images captured by the camera 10. As a hardware configuration, the image storage unit 15 may be, for example, an SD card as a removable external storage means, or may be, for example, a hard disk as an internal storage means. The image storage unit 15 may be a storage means that exists outside the drive recorder 1, and such a storage means may be a storage means possessed by another drive recorder, or a storage means that the drive recorder 1 uses via a public network (for example, a storage means provided using a cloud computing environment, etc.).

[0067] The control unit 14 constantly acquires the video captured by the camera 10 and temporarily stores it in the internal memory. When the internal memory becomes full, overwrite recording is performed by overwriting the video. This overwrite recording may be performed using a recording area secured in the video storage unit 15. In this case, overwrite recording may be called continuous recording. When a predetermined acceleration is detected by the acceleration sensor 12 to the control unit 14, the control unit 14 performs event recording by storing a predetermined length of video including the timing in the video storage unit 15. In this way, the drive recorder 1 can store video necessary for analysis of an accident or the like in the video storage unit 15. Since the control unit 14 temporarily stores video in the internal memory, when a predetermined acceleration is detected, the control unit 14 can store video from a time preceding the timing of the detection in the video storage unit 15.

[0068] In addition to the above-mentioned configuration, the drive recorder 1 of this embodiment includes a model storage unit 16. The model storage unit 16 stores a trained model for detecting a predetermined movement that triggers control of recording by the drive recorder 1 from an image. The trained model used in this embodiment is a model trained by a machine learning method using a deep neural network. How to generate the trained model will be described later.

[0069] The control unit 14 reads out the trained model stored in the model storage unit 16, applies the image transmitted from the camera 10 to the trained model, and detects a predetermined movement from the image that will be a trigger for controlling recording. As an example, a movement related to a part of the vehicle is detected based on a (temporal) change in the image of the area where the part of the vehicle is located.

[0070] In this embodiment, the movement related to the vehicle part is the movement of the wiper. The control unit 14 determines that a trigger to start recording has occurred when the wiper shown in the image captured by the camera 10 performs a certain movement by inference using the learned model, and the drive recorder 1 starts recording. In order to make the drive recorder 1 easy to operate, a simple and short movement is preferable. Conversely, in order to avoid a malfunction, some movement pattern may be used as a trigger. In addition to the movement of the wiper, for example, spraying of the washer fluid may also be included. For example, a series of actions such as spraying the washer fluid and then operating the wiper several times may be used as a trigger for control related to recording.

[0071] Specifically, the control unit 14 performs event recording by storing the captured data from a predetermined time before the predetermined movement in the video storage unit 15. Here, the predetermined time before may be, for example, one minute before the triggering movement is detected. By storing the video from the predetermined time before in this way, even if it takes some time for the driver to execute the predetermined movement, the video can be stored at the timing when the driver wants to record.

[0072] For example, suppose that the driver witnesses a supercar or a falling meteorite and operates the wipers in order to record it in the drive recorder 1. In this case, the supercar may have driven away or the meteorite may no longer be visible during the time between when the wipers perform a predetermined operation and when the drive recorder 1 starts recording, but the configuration of this embodiment for storing video from a slightly earlier time makes it possible to store the car that the driver intended to record in the video storage unit 15.

[0073] When the trigger for starting recording according to the present embodiment is applied to a drive recorder that stores an event record in one file, when the start of recording is triggered based on a movement related to a vehicle part, the file immediately before the event file may also be incorporated into the event. In this way, the control unit 14 also has a function of performing event recording to store video going back to the second predetermined time from the timing of the detection when an impact on the vehicle is detected, but when there is the above-mentioned movement of the wiper, the video goes back to the first predetermined time, which is longer than the second predetermined time, and stores the video. In this way, even if the reaction to the start of recording due to the movement of the wiper is slow, video from a long time ago can be saved. In addition, in the event recording, in a file that can record one minute of video, if the event recording starts with 10 seconds remaining, for example, 30 seconds are added to the end to make a file of 1 minute 30 seconds, but when the start of recording is triggered based on a movement related to a vehicle part, 30 seconds may be added to the beginning of the file. In this way, it is possible to match the specifications of the event recording.

[0074] 3 is a flowchart showing the recording operation of the drive recorder 1. The drive recorder 1 captures an image of the area ahead of the vehicle with the camera 10 (S10). The camera 10 captures images at all times, and the captured image data is temporarily stored in the internal memory of the control unit 14.

[0075] The control unit 14 determines whether or not there is a wiper movement that triggers recording in the image captured by the camera 10 (S11). If there is no wiper movement that triggers recording (NO in S11), the control unit 14 repeats this determination until there is a wiper movement that triggers recording (S11).

[0076] When it is determined that there is a wiper movement that triggers recording (YES in S11), the control unit 14 saves the video for a predetermined time (S12). Saving the video here does not mean temporarily storing the video in the internal memory of the control unit 14, but rather saving the video data in the video saving unit 15. This allows the video data to be held in the drive recorder 1 without being overwritten.

[0077] Next, a method for generating a trained model will be described. In this embodiment, the drive recorder 1 includes a learning unit 17, which learns the movement of the wipers that triggers recording.

[0078] FIG. 4 is a flowchart showing an operation of generating a trained model. The drive recorder 1 judges whether or not there is an input to switch to the learning mode (S20). For example, one of the various switches 13 is a switch to switch to the learning mode, and by pressing this, the drive recorder 1 switches to the learning mode. During learning, the drive recorder 1 may be attached in a position different from that shown in FIG. 2 so that the various switches 13 can be easily operated. During the period set in the learning mode, the drive recorder 1 performs learning to generate a learning model. When the learning mode starts, the drive recorder 1 starts shooting video that will become teacher data (S21). The video that will become teacher data is video of a predetermined time (for example, 10 seconds), and shooting ends when the predetermined time has elapsed (S22).

[0079] The time for capturing the video to be used as the training data may be set by the driver. When the time for capturing is short, a model can be generated in which a short-term movement is used as a trigger to start recording, and when the time for capturing is long, a model that is less likely to malfunction and in which a long-term movement is used as a trigger to start recording can be generated.

[0080] Next, the learning unit 17 reads out the model stored in the model storage unit 16, and updates the model using the video of the teacher data obtained by shooting (S23). Specifically, the model parameters are updated so that the video obtained by shooting is regarded as the correct video that triggers recording, and the result that the video triggers the start of recording is obtained.

[0081] The following is an example of a method for generating a trained model performed by the learning unit 17. The learning unit 17 acquires, as teacher data, a video showing a video captured by the camera 10 when the learning mode is set, and generates a trained model. When the wiper is moving, the camera 10 captures the wiper. The learning unit 17 inputs data showing each frame of a plurality of frames constituting the video captured by the camera 10 (i.e., still image data) into the model. The model learns parameters using teacher data that inputs data showing the video and outputs data corresponding to the degree of accuracy that the wiper is captured in the input video. The data corresponding to the degree of accuracy that the wiper is captured may be a value indicating the probability that the wiper is determined to be captured in the video. In this case, if the probability is equal to or greater than a threshold value, it is determined that the wiper is captured. The data indicating the degree of accuracy that the wiper is captured may be information indicating whether or not the wiper is captured. The data corresponding to the degree of accuracy that the wiper is captured may be a value indicating the degree to which event recording should be performed. The trained model thus generated has an input layer to which data showing a video is input, an output layer to which data corresponding to the degree of certainty that a part of the vehicle is captured in the video is output, and at least one intermediate layer in which parameters are trained using teacher data that receives data showing a video of a movement related to a part of the vehicle captured during learning and outputs data corresponding to the degree of certainty that a part of the vehicle is captured in the video. The model is a trained model generated by deep learning technology, and may be, for example, a deep neural network (hereinafter referred to as DNN (Deep Neural Network)), but may also be a CNN (Convolutional Neural Network) or other learning models.

[0082] When detecting a predetermined movement that will be a trigger for recording an event, the control unit 14 inputs the image from the camera 10 to the input layer of the trained model, performs calculations in the intermediate layer, and determines whether to record the event based on the data output from the output layer. For example, when the control unit 14 determines that the wiper is continuously captured for a predetermined period (e.g., a predetermined number of frames) or more based on information indicating the probability that the wiper is captured, it may use this as a trigger to record the event.

[0083] As described above, the drive recorder 1 also functions as a learning model generating device or a device that executes a learning model generating method. The learning model generating method described above is merely an example. The learning unit 17 may, for example, use deep learning technology to capture the state (scene) of each of a plurality of time points when the wiper is moving in the learning mode, and generate a learned model corresponding to each of the plurality of time points. When detecting a predetermined movement that triggers the event recording, the control unit 14 may input the image of each time point from the camera 10 to the input layer of each learned model, perform calculations in the intermediate layer, and determine whether or not a movement identical to or similar to the wiper movement learned in the learning mode (for example, a similarity degree equal to or greater than a threshold) has been detected based on the data output from the output layer. For example, the control unit 14 determines that the wiper movement is detected when the similarity degree is equal to or greater than a threshold at each of a plurality of time points when the wiper is moving. When the control unit 14 determines that the learned wiper movement has been detected, it may determine that an event recording should be performed, and use this as a trigger to perform the event recording.

[0084] Alternatively, the following method may be used. The learning unit 17 may estimate a movement (e.g., optical flow) from the image of each of a plurality of frames constituting a video in which the wiper is moving, for example, using a deep learning technique. In this way, the control unit 14 can detect the regular movement of the wiper from the image and perform event recording, since the movement of the wiper is regular. In this case, when detecting a predetermined movement that serves as a trigger for performing event recording, the control unit 14 may apply the image of the camera 10 to a learned model, and when detecting a movement that is the same as or similar to the movement estimated in the learning mode (for example, a similarity equal to or greater than a threshold), use this as a trigger to perform event recording.

[0085] According to the trained model described above, it is expected that the operation of the drive recorder 1 can be simplified, and it may be possible to use the trained model to simplify the operation of other electronic devices in a similar manner. The method is not limited to using deep learning technology, and the control unit 14 may use a method such as pattern matching to determine whether the wipers are shown in the image or to estimate the movement of the image. In this case, the drive recorder 1 does not need to have a function equivalent to the learning unit 17.

[0086] Furthermore, the learning unit 17 may generate a learning model by using as training data an image captured when the learning mode is not in effect, associated with no control over recording by the drive recorder 1. In this way, by using as negative training data an image that does not trigger control over recording by the drive recorder, a trained model with high inference accuracy can be generated.

[0087] The configuration of the drive recorder 1 of this embodiment has been described above, and an example of the hardware of the control unit 14 of the drive recorder 1 described above is a calculation device equipped with a CPU, RAM, ROM, etc. A program having modules for realizing each of the above-mentioned functions is stored in the RAM or ROM, and the above-mentioned functions of the drive recorder 1 are realized by executing the program by the CPU. Such a program is also included in the scope of the present invention.

[0088] The drive recorder 1 of the first embodiment can give the drive recorder 1 a control instruction for recording by using the movement related to the part of the vehicle in the image. The drive recorder 1 can be easily operated regardless of the location where the drive recorder 1 is installed. In addition, since the movement related to the part of the vehicle can be detected using the image captured by the camera 10 of the drive recorder 1, a new camera is not required to capture the movement related to the part of the vehicle. In addition, when the movement related to the part of the vehicle is the driver's movement with respect to the part of the vehicle, the function of the learning unit 17 can reflect the individual habits and preferences of the driver in the predetermined movement. In particular, since the wiper lever is optimized based on ergonomics, the wiper movement can be used to safely operate the wiper even while the vehicle is running.

[0089] In this embodiment, the start of recording has been described as an example of control related to recording, but for the end of recording, as in the above embodiment, a model that has learned a specific wiper movement may be used to detect the wiper movement that triggers the end of recording from the video. Note that the control related to recording is not limited to the start of recording, which is the same in the examples described below.

[0090] (Configuration of the second embodiment) Next, a drive recorder according to a second embodiment will be described. The basic configuration of the drive recorder according to the second embodiment is the same as that of the drive recorder 1 according to the first embodiment. In the second embodiment, the movement related to the vehicle parts is the driver's movement with respect to the hazard lamp switch, which is one of the vehicle's operating parts, and the movement that triggers the start of recording is, for example, holding a hand over the hazard lamp switch.

[0091] The model storage unit 16 stores a trained model using a video of a driver operating a hazard lamp switch as training data. This trained model is generated and stored in the model storage unit 16 before the drive recorder is shipped.

[0092] In the second embodiment of the drive recorder, when the driver performs a predetermined action, such as holding his / her hand over the hazard lamp switch, the control unit 14 detects the movement that triggers the start of recording from the image captured by the drive recorder, and saves the image for a predetermined period of time (e.g., one minute) including the timing of the movement.

[0093] The drive recorder of the second embodiment, like the first embodiment, allows the drive recorder to be easily operated regardless of the location where the drive recorder is installed. The vehicle operation unit is, for example, an operation unit pre-installed in the vehicle, and is often located within reach of the person in the vehicle so that the person can easily operate it. Therefore, the person in the vehicle can easily perform an operation to instruct the execution of control related to recording of the drive recorder. The person's action on the vehicle operation unit may include, for example, an action such as holding a hand over a hazard lamp or an air conditioner switch. These types of operation units and actions that the person should perform are merely examples and are not limited to these.

[0094] (Configuration of the third embodiment) Next, a drive recorder according to a third embodiment will be described. The basic configuration of the drive recorder according to the third embodiment is the same as that of the drive recorder 1 according to the first embodiment. In the third embodiment, the movement related to the vehicle part is a movement with respect to a mark prepared in the vehicle, and the movement that triggers the start of recording is, for example, holding a hand over the mark and the mark being unable to be detected for a certain period of time or more. By setting the condition to be a certain period of time or more, it is possible to prevent recording from being started erroneously when the user's hand temporarily blocks the image of the mark.

[0095] FIG. 5 is a diagram for explaining a mark prepared in a vehicle. The mark has a design that can be distinguished from other parts in the vehicle. The mark is a sticker with a design. This allows the mark to be affixed to any position in the vehicle. The mark can be prepared in a place that is easy for the driver to operate.

[0096] In FIG. 5, the mark has a pattern in which a yellow ellipse is arranged inside a red ellipse so that their centers coincide, but this is not limited to this. It is desirable that the design of the mark is one that does not have a similar design in the vehicle. Such designs include a pattern using a reflective material and a design using a hologram. The inventor has obtained the knowledge that these designs are unlikely to have the same or similar design in the vehicle, and that it is expected to reduce the possibility of erroneously detecting a person's action toward the mark. For example, a pattern using a reflective material is easy to notice even in an image. For example, a design using a hologram is easy to distinguish from other objects in the image.

[0097] The model storage unit 16 stores a trained model using a video of a driver operating a hazard lamp switch as training data. This trained model is generated and stored in the model storage unit 16 before the drive recorder is shipped.

[0098] In the third embodiment of the drive recorder, when the driver performs a predetermined action, such as holding his / her hand over a mark, the control unit 14 detects the movement that triggers the start of recording from the image captured by the drive recorder, and saves the image for a predetermined period of time (e.g., one minute) including the timing of the movement.

[0099] The drive recorder of the third embodiment, like the first embodiment, allows the driver to easily operate the drive recorder regardless of the location where the drive recorder is installed. In addition, in this embodiment, the mark can be prepared in a location where the driver can easily operate, so that the operability and safety are higher than, for example, a case where a recording start button is prepared in a predetermined location.

[0100] In the above embodiment, the trigger for starting recording is when the mark is blocked (when the mark disappears from the image), but the trigger for starting recording may be when the mark appears. For example, it may be determined that the trigger for starting recording is when the folded mark is unfolded. In this way, the mark is configured to have two states: a state in which the mark is photographed by the camera 10 and a state in which the mark is not photographed. In addition, the mark is not limited to being formed on a sticker, but may be formed on an item brought into the car (for example, a driver's belongings), or may be displayed on a display device such as a smartphone. In addition, the mark may be freely determined by the user, and in this case, the drive recorder may read the mark when set to the learning mode and learn it. In this way, a mark with a design that suits the user's preferences can be used.

[0101] In the above embodiment, an example was given in which a predetermined action that triggers the start of recording is detected from the video using a trained model, but since the mark design is known in advance, it may be determined by pattern matching whether the mark is blocked or not, or whether the mark appears or not. In this case, since the pattern of the mark design is only required to be stored in advance, the driver does not need to learn. In this case, the drive recorder does not need to store the trained model or execute processing using the trained model. Therefore, it is expected to have the effect of suppressing the complexity and high cost of the drive recorder configuration.

[0102] In the above-described embodiment, a plurality of different types of marks may be used to trigger control corresponding to the marks. For example, the start of recording may be triggered when one mark is obstructed, and the end of recording may be triggered when another mark is obstructed. In addition to the above, the control corresponding to the mark may be the setting of a condition related to recording (e.g., resolution or recording time), or any other control related to recording. In this way, the variety of available controls is increased.

[0103] (Configuration of the Fourth Embodiment) 6 is a diagram showing the configuration of the drive recorder 4 according to the fourth embodiment. In addition to the configuration of the drive recorder 1 according to the first embodiment, the drive recorder 4 according to the fourth embodiment has a microphone 18 as an audio input unit. The drive recorder 4 according to the fourth embodiment passes audio data collected by the microphone 18 to the control unit 14. The control unit 14 also uses the audio data to detect a trigger for control related to recording.

[0104] In the drive recorder 1 of the first embodiment, it is determined that a trigger to start recording has occurred when it is determined that the wipers shown in the image captured by the camera 10 have performed a certain fixed operation, but the drive recorder 4 of the present embodiment determines that a trigger to start recording has occurred when it detects the input of predetermined audio data such as "one-touch recording" in addition to detecting a trigger in the image. In this way, by using both the image and the audio, it is possible to increase the accuracy of detection of the trigger to start recording.

[0105] For example, when the voice is a human voice (e.g., a predetermined keyword), the human will is involved in emitting the voice, so the possibility of the control related to the recording being executed at a timing not intended by the person can be reduced. In this respect, it is also possible to suppress the control related to the recording being executed erroneously. For example, if the input voice satisfies the condition for executing the control related to the recording within a predetermined period (e.g., simultaneously) from when the movement related to the part of the vehicle satisfies the condition for executing the control related to the recording, the control related to the recording can be executed. The voice for instructing the execution of the control related to the recording may be predetermined, or may be set by the user. Furthermore, if the voice for instructing the execution of the control related to the recording is a voice that means the content of the control related to the recording, such as "one-touch recording," people can easily understand what kind of control is executed by what kind of voice. On the other hand, if the voice for instructing the execution of the control related to the recording is not a voice that means the content of the control related to the recording, it is possible to reduce the possibility that the drive recorder is unintentionally operated by a person other than a specific person, such as a third party other than the driver. Furthermore, a plurality of different types of voices may be used to trigger the control corresponding to the voice. For example, it may be possible to selectively start and end recording, and set conditions related to recording (such as resolution and recording time) by using sound, etc. In this way, the variety of available controls is increased.

[0106] (Fifth embodiment) The drive recorder of the fifth embodiment has the following function for determining whether or not a trigger for starting recording has occurred based on an image captured by a camera. Hereinafter, this embodiment will be described on the assumption that the drive recorder of this embodiment has the same configuration as the drive recorder 4 shown in FIG. 6, as in the fourth embodiment described above. However, the configuration (function) of this embodiment may be realized even in a drive recorder that does not have a function for controlling recording based on a movement related to a part of the vehicle in the image captured by the camera. In this embodiment, the drive recorder 4 is mounted on a vehicle used by an unspecified person, such as a rental car. The drive recorder 4 has a function for optimally editing the recorded image of the drive recorder 4 after being used at a travel destination, for example, and providing it to a customer. The editing is automatic editing by learning processing (for example, AI processing). In the drive recorder 4, the control unit 14 controls the recording of the drive recorder based on the result of learning the image captured when each of the multiple users uses the vehicle. For this purpose, the learning unit 17 may learn the video captured by the camera 10 when each of the multiple users instructs one-touch recording. The learning may be performed using the video of the multiple users, or may be performed using the video for each user. For example, the learning unit 17 may learn at least one of the following situations (e.g., situations) as the video for event recording: a part with conversation (e.g., a conversation by the voices of people inside the car), a famous place (e.g., identified by location information), and a part with good scenery (e.g., a night view, a sunset, a rainbow). For the part with conversation, the learning unit 17 may learn the voice collected by the microphone 18. The learning unit 17 may perform this learning when operating in the learning mode. In this embodiment, it is necessary to take care to prevent the data of the video from being lost, particularly when the data is overwritten by continuous recording, so it is desirable for the drive recorder 4 to have an SD slot with the maximum capacity. For example, multiple SD slots may be installed.

[0107] In this embodiment as well, the drive recorder can be easily operated. Moreover, according to this embodiment, an event is recorded without the user's instruction, and the event is recorded so that it can be enjoyed later. Therefore, it is expected that the drive recorder 4 will be selected by customers who are sightseeing. Furthermore, the recording medium (for example, a DVD or an SD card) on which the event is recorded may be sold to customers.

[0108] (Sixth embodiment) The drive recorder of the sixth embodiment will be described on the assumption that it has the same configuration as the drive recorder 1 described above, but the configuration (function) of this embodiment may be realized even in a drive recorder that does not have a function for controlling recording based on the movement related to the parts of the vehicle in the image captured by the camera. The camera 10 is preferably a camera that can capture the outside of the vehicle and can capture images with a wide viewing angle, such as a hemispherical camera or a celestial sphere camera. In this embodiment, the drive recorder 4 is also described as being mounted on a vehicle used by unspecified persons (passengers in this embodiment), such as a taxi. In this embodiment, the control unit 14 notifies the user based on the movement of a person wearing a helmet in the image captured by the camera 10. When the control unit 14 detects that a person is not wearing a helmet based on an image captured by the camera 10 of the outside of the vehicle, and detects a person raising his / her hand, the control unit 14 may notify the user (the taxi driver in this embodiment). The notification may be performed by a method that the user can perceive, such as display on the monitor 11, output of sound from a speaker not shown, or emission of light from a light-emitting body such as an LED. This is done to prevent a customer from being overlooked while the taxi is being driven. The control unit 14 does not issue the above-mentioned notification when it detects that the person is wearing a helmet. For example, when the control unit 14 detects that the person is wearing a helmet and that the person is a child, it may detect the person as a school child. The control unit 14 may issue a notification in this case differently from the notification issued when a customer is detected while the taxi is being driven, or may not issue a notification at all. By issuing a notification using the helmet as a landmark in this way, it is possible to determine the user's attributes, for example, whether the person needs a taxi.

[0109] Furthermore, the control unit 14 may perform congestion analysis based on an image taken of the outside of the vehicle by the camera 10. The control unit 14 performs a situation analysis of a stand such as a taxi stand (e.g., how many passengers are waiting and how long it will take to pick up passengers) and notifies a predetermined notification destination (e.g., a server of the same taxi company, or a drive recorder 1 of another vehicle of the same taxi company) of the analysis result. Furthermore, in the drive recorder 1, the control unit 14 may obtain the result of this congestion analysis and notify the user. For example, the control unit 14 may obtain the result of the situation analysis of a stand that exists within a predetermined range from the drive recorder 1 and notify the user.

[0110] Furthermore, when the control unit 14 detects an abnormality on the road on which the vehicle is traveling based on an image taken outside the vehicle by the camera 10, it may notify the user. This notification is for the purpose of notifying the user of safety. The abnormality on the road may be an abnormality in the road itself (e.g., a crack or a depression), or an abnormality that may hinder traffic safety, such as the presence of a puddle or an obstacle.

[0111] (Seventh embodiment) The drive recorder of the seventh embodiment will be described on the assumption that it has the same configuration as the drive recorder 1 described above, but the configuration (function) of this embodiment may be realized even in a drive recorder that does not have a function for controlling recording based on the movement related to the parts of the vehicle in the image captured by the camera. When the camera 10 is a camera capable of capturing images at a wide viewing angle, such as a hemispherical camera or a celestial sphere camera, or when the drive recorder 1 has two cameras, the drive recorder can also capture images inside the vehicle. In this case, the drive recorder 1 may have the following configuration. In this embodiment, the drive recorder 4 is mounted on a vehicle used by unspecified persons (here, passengers), such as a taxi. In this embodiment, the control unit 14 may detect that luggage has been left behind inside the vehicle. For example, the control unit 14 determines whether luggage has been left behind on the seat or at the foot of the passenger after the passenger gets off, based on the image of the camera 10. In this case, the control unit 14 detects that luggage has been left behind and notifies the user (for example, the driver).

[0112] Furthermore, in this embodiment, the drive recorder may have a function of notifying at a place where sudden braking is often performed. For example, the notification is performed for the purpose of announcing a recommendation regarding the vehicle's speed (current speed, speed limit, etc.) before passing that place. The learning unit 17 learns the video captured by the camera 10 when the vehicle suddenly brakes based on the acceleration detected by the acceleration sensor 12. The control unit 14 notifies the user based on the characteristics of the video at the time of sudden braking and the characteristics of the video captured by the camera 10 based on the results of learning by the learning unit 17. For example, the control unit 14 may notify when the similarity between the characteristics of the video at the time of sudden braking and the characteristics of the video captured by the camera 10 is equal to or greater than a threshold value. In this way, the drive recorder can make a notification such as an announcement of a speed recommendation even in a place where sudden braking has not yet been performed.

[0113] A drive recorder may also have an aspect of an action camera, and the captured images may be posted on SNS as hobby images or images. In a drive recorder, when images or images are recorded in a storage area using a one-touch recording function or the like, the storage area contains many images with the same characteristics. Therefore, the control unit 14 may perform event recording based on the image captured by the camera 10 and multiple images stored in a storage area (for example, the storage area of ​​the image storage unit 18) that stores the images captured by the camera. The storage area may be, for example, a folder in which images are stored. For example, when the similarity between the characteristics of the image captured by the camera 10 and the characteristics of the image stored in the storage area is equal to or greater than a threshold, the control unit 14 may notify the user with a message such as "Do you want to record?" and perform event recording according to the user's response, or may perform event recording without notification. In this way, the drive recorder can be used for purposes such as increasing the number of images of the user's hobbies.

[0114] [Variations] Although the configuration and operation of the drive recorder of the present invention have been described in detail above with reference to the embodiment, the drive recorder of the present invention is not limited to the above-described embodiment.

[0115] In the above embodiment, an example was given in which a predetermined action that triggers control related to recording is learned by the drive recorder, but the trigger action may be determined in advance, a trained model for detecting the action may be generated, and stored in the model storage unit 16 of the drive recorder. Also, when learning a model in the drive recorder, the trained model may be modified to change to a model suited to each driver. This may enable efficient model generation.

[0116] In the above embodiment, an example was given of using camera 10 that captures images to be recorded in a drive recorder as a means for capturing movements related to parts of the vehicle, but it is also possible to provide a camera other than camera 10.

[0117] In the above embodiment, an example was given in which a model was learned using a video captured in the learning mode. In actual operation, a model may be generated that detects an action that starts recording from a movement related to a part of the vehicle when a driver operates a switch of the drive recorder to start recording. In this case, the learning period may be about one week, or may be about two or three days. A model with high estimation accuracy can be generated if the learning period is long. The learning period may be determined by the user's settings. This allows the user to select a shorter learning period if, for example, the user actively wants to record manually. By setting the learning period in this way, it is not necessary to switch the drive recorder from the normal use mode to the learning mode.

[0118] The drive recorder of this embodiment can also be used for fixed-point observation. For example, the location is limited, the above-mentioned processing is performed, and recording is performed on the premise that it is a specific location. Also, the time is limited, and the above-mentioned processing is performed, and recording is performed on the premise that it is a specific time, such as nighttime. In this way, the driver can make the drive recorder record events only when it is necessary.

[0119] The above-mentioned function of the learning unit 17 may be executed by a system other than a drive recorder. Such a system may be configured by one or more devices such as a personal computer, a server device, or other systems having information processing functions. In this case, the system may store the generated trained model in the model storage unit 16. In addition, the entity that realizes each function of the system may be any of hardware resources, software resources, and a combination of these.

[0120] The scope of the present invention is not limited to the configurations explicitly described in the specification, but includes combinations of various aspects of the present invention disclosed in this specification. The configurations of the present invention that are to be patented are specified in the attached claims, but it is our intention to include configurations disclosed in this specification in the claims in the future, even if they are not currently specified in the claims.

[0121] The present invention is not limited to the configuration described in the above-mentioned embodiment. The components of each of the above-mentioned embodiments and modifications may be arbitrarily selected and combined. Any components of each of the embodiments and modifications may be arbitrarily combined with any components described in the means for solving the invention or any components that embody any components described in the means for solving the invention. We intend to acquire rights to these as well through amendments to this application or divisional applications. Even if there is a description such as "in the case of" or "when", this is not intended to describe a configuration that is limited to that case or time. We also disclose configurations that are not in these cases or times, and we intend to acquire rights to them. Furthermore, the parts described in order are not limited to this order. We also disclose configurations in which some parts are deleted or the order is changed, and we intend to acquire rights to them.

[0122] In addition, we intend to obtain rights to the overall design or partial design by converting to a design registration application. The drawings depict the entire device in solid lines, but they include not only the overall design but also partial designs claimed for a part of the device. For example, a part of the device may be a partial design, or a part of the part may be a part of the device. We intend to obtain rights to the overall design, as well as partial designs in which any part of the solid line part of the drawing is drawn as a broken line part. In addition, the modules, parts, and components inside the device's casing, all of which are shown in the drawings, are subject to trade independently, and we intend to obtain rights to them by converting to a design registration application. [Explanation of symbols]

[0123] 1,4 Drive recorder 10. Camera 11 Monitor 12 Acceleration Sensor 13 Various switches 14 Control section 15 Video Storage Section 16 Model memory section 17 Learning Department 18. Mike

Claims

1. Camera and A control unit that controls recording based on a movement related to a part of the vehicle in the image captured by the camera; A drive recorder equipped with the following.

2. a model storage unit that stores a trained model for detecting a predetermined movement associated with a part of the vehicle that serves as a trigger for controlling the recording; The drive recorder according to claim 1 , wherein the control unit applies the image captured by the camera to the model, and performs control related to the recording when the predetermined movement is detected.

3. 3. The drive recorder according to claim 2, further comprising a learning unit that learns the model by using images captured during learning as teacher data associated with control of recording by the drive recorder.

4. The drive recorder according to claim 2 or 3, wherein, when the specified movement is a movement instructing the start of recording, the control unit stores in the image storage unit image data from a first specified time before the specified movement occurred.

5. A detection unit that detects an impact applied to the vehicle, The control unit stores images going back to a second predetermined time before the timing of the impact detection, when the impact is detected, The first predetermined time is longer than the second predetermined time. The drive recorder according to claim 4.

6. The drive recorder according to claim 1 , wherein the movement associated with the part of the vehicle is the operation of a windshield wiper.

7. The drive recorder according to claim 1 , wherein the movement associated with a part of the vehicle is a human action on an operating unit of the vehicle.

8. 6. The drive recorder according to claim 1, wherein the movement associated with a part of the vehicle is an action relative to a mark having a pattern that is distinguishable from other parts in the vehicle.

9. The drive recorder according to claim 8 , wherein the mark is a sticker attached inside the vehicle.

10. 10. The drive recorder according to claim 8, wherein the predetermined movement that triggers control regarding the recording is when the mark cannot be detected for a certain period of time or longer.

11. 11. The drive recorder according to claim 8, wherein a plurality of marks having different designs are provided, and the control unit controls the drive recorder depending on which type of mark has been detected as the movement.

12. A voice input unit is provided for receiving voice input, The drive recorder according to claim 1 , wherein the control unit controls recording by the drive recorder based on a combination of a motion and a sound related to a part of the vehicle.

13. The vehicle is used by unspecified persons, The control unit controls the recording of the drive recorder based on a result of learning the video captured by each of a plurality of users when the vehicle is used. The drive recorder according to any one of claims 1 to 12.

14. The control unit notifies a user based on a movement of a person wearing a helmet in an image captured by the camera. The drive recorder according to any one of claims 1 to 13.

15. When the control unit detects an abnormality on a road on which the vehicle is traveling, the control unit notifies the user. The drive recorder according to any one of claims 1 to 14.

16. The vehicle is used by unspecified persons, When the control unit detects that luggage has been left behind in the vehicle, the control unit notifies the user. The drive recorder according to any one of claims 1 to 15.

17. The control unit notifies a user based on the image captured by the camera and a plurality of images stored in a storage area in which the images captured by the camera are accumulated. The drive recorder according to any one of claims 1 to 15.

18. A program for executing processing of the control unit of the drive recorder according to any one of claims 1 to 17.

19. A learning model generation method for generating a learning model for detecting a predetermined movement that serves as a trigger for controlling recording of a drive recorder from a video captured by the drive recorder, comprising: receiving an input instructing the drive recorder to learn the predetermined movement; generating a learning model using the video captured during the learning as teacher data associated with control of recording by the drive recorder; A learning model generation method comprising:

20. The learning model generating method according to claim 19 , wherein video images captured within a predetermined time from the start of the learning are used as the training data.

21. The learning model generating method according to claim 19 or 20, wherein the step of generating the learning model generates a learning model as teacher data that associates video footage taken at times other than the learning time with not controlling recording by a drive recorder.

22. an input layer to which data representing an image is input; an output layer that outputs data according to a probability that a part of the vehicle is shown in the image; and at least one intermediate layer in which parameters are learned using teacher data that receives as input data representing a video of a movement related to a part of the vehicle during learning, and outputs data corresponding to a degree of accuracy that the part of the vehicle is captured in the video. Trained model.

23. A system having a control unit that performs predetermined control regarding recording of images captured by a camera of a vehicle based on movements related to parts of the vehicle in the images captured by the camera.

24. A storage unit that stores the trained model according to claim 22; An acquisition unit that acquires data representing a video; having The control unit is Data showing the acquired image is input to the input layer, and is calculated in the intermediate layer. Based on data corresponding to the degree of certainty that a part of the vehicle is shown in the image output from the output layer, a predetermined control is performed regarding the recording.

23. The system of claim 22.

Images