JP2024011741A - Time correction device, time correction method, and time correction system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a time correction device capable of infrequently performing time correction on equipment without using a highly accurate clock that provides a reference time.

SOLUTION: In a time correction device in an embodiment, a correction module learns the relationship between information about a piece of equipment and the deviation (time difference) of the clock time included in the equipment (current time) from the reference time. The time correction device corrects the time of the clock included in the equipment using the learned correction module.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2024,JPO&INPIT

Description

The present invention relates to a time correction device, a time correction method, and a time correction system.

2. Description of the Related Art Conventionally, a technique for improving the accuracy of correcting the time of a clock provided in a device is known (for example, see Patent Document 1).

Japanese Patent Application Publication No. 2018-190216

However, the conventional technology has a problem in that it may not be possible to correct the time of the device at a low frequency without using a highly accurate clock that provides a reference time.

The present invention has been made in view of the above, and an object of the present invention is to correct the time of a device at low frequency without using a highly accurate clock that provides a reference time.

In order to solve the above-mentioned problems and achieve the objects, a time correction device according to the present invention is a time correction device that corrects the time of a clock provided in a device, and includes a controller. The controller causes the module to learn the relationship between information about the device and the deviation of the time of the clock installed in the device from the reference time, and uses the learned module to correct the time of the clock installed in the device. I do.

According to the present invention, it is possible to correct the time of a device at low frequency without using a highly accurate clock that provides a reference time.

FIG. 1 is a diagram illustrating a correction module. FIG. 2 is a diagram showing an example of the configuration of the time correction device according to the first embodiment. FIG. 3 is a flowchart showing the flow of learning processing. FIG. 4 is a flowchart showing the flow of the correction process. FIG. 5 is a diagram illustrating the effects of the first embodiment. FIG. 6 is a diagram illustrating a configuration example of a time correction system according to the second embodiment. FIG. 7 is a diagram illustrating a conventional time correction method.

Hereinafter, embodiments of a time correction device, a time correction method, and a time correction system will be described in detail with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments described below.

Here, there is a system that utilizes images, sounds, and other data (time-series sensor logs, etc.) recorded by multiple devices for processing such as synthesis, judgment, and selection. For example, a system that analyzes traffic accidents synchronizes images from in-vehicle cameras and images from roadside surveillance cameras to analyze the situation at the same time from multiple perspectives. be able to.

The time stamp of the image is the time indicated by the clock provided in the device. Therefore, in order to accurately synchronize the images of the vehicle-mounted camera and the surveillance camera, it is necessary that the clock time of the vehicle-mounted camera and the clock of the surveillance camera match.

Conventionally, as shown in FIG. 7, the clocks of each device were synchronized by providing each device with a reference time. In the example of FIG. 7, the providing device 50 provides the reference time to the device 51a and the device 51b.

The times of the clocks provided in the device 51a and the device 52a are corrected in accordance with the reference time provided by the providing device 50.

The device 51a is, for example, an in-vehicle camera. Further, the device 51b is, for example, a surveillance camera. Further, the providing device 50 is a server connected to the device 51a and the device 51b via the Internet.

Here, in order to accurately correct the time using the conventional technology, it is necessary to receive the reference time frequently. On the other hand, it may not be practical to receive reference time frequently due to communication costs and the like.

Furthermore, if the reference time is not provided frequently, it is difficult to synchronize the clock times among a plurality of devices that have different reference time reception timings, device characteristics, and operating environments.

For example, if the environment in which the device is used differs, the rate of deterioration of parts will differ. It is conceivable that the clock time of a device that has significantly deteriorated will have a large deviation from the reference time.

In contrast, according to the present embodiment, the time of the device can be corrected at low frequency without using a highly accurate clock that provides the reference time.

In this embodiment, the clock is corrected by time correction predicted using a correction module.

The correction module will be explained using FIG. 1. FIG. 1 is a diagram illustrating a correction module.

The correction module outputs a time correction value based on the input information. The time correction value is a positive or negative time. The clock is corrected by setting the current time of the clock to the current time indicated by the clock plus the time correction value.

For example, assume that the current time indicated by the clock is "10:29:31" and the time correction value is "00:03:10" (both in hh:mm:ss format). In this case, since "10:29:31+00:03:10=10:42:41", correction is performed by setting "10:42:41" to the clock time.

The correction module may also be referred to as a function, a model, or the like. The correction module learns (is trained) the relationship between information about the device and the deviation of the time of the clock provided in the device from the reference time using a machine learning method.

The input information corresponds to the explanatory variables of the correction module. Further, the time correction value corresponds to the objective variable of the correction module. Note that the deviation between the current time and the reference time (corresponding to an objective variable) and the input information when the deviation is known (corresponding to an explanatory variable) can be said to be training data for learning of the correction module.

Here, the device of this embodiment may be any device provided with a clock, such as an in-vehicle camera, a surveillance camera, a sound collection device, or the like.

The input information shown in FIG. 1 is an example of information regarding equipment. The input information includes sensor information, device operating status, and cumulative operating time.

The sensor information is a sensor value measured by a sensor provided in the device. Sensor information includes temperature, humidity, and voltage. The device operating state is, for example, the frequency of starting and stopping the device's operation. Cumulative operating time is the amount of time a device has been operating since a certain point in time.

The current time and reference time are used for learning by the correction module 1211. The correction module 1211 may be trained or unlearned.

The current time is the time of the device's clock. The clock is a circuit using, for example, a crystal oscillator and an oscillation circuit, and outputs date and time. The reference time is a time provided by a highly accurate clock that is separate from the device.

It can be said that the current time is an estimated value and the reference time is a true value. The ideal situation is for the current time to match the reference time. Therefore, learning is performed so that the time correction value output by the correction module 1211 approaches the difference between the reference time and the current time.

Then, by inputting the input information to the learned correction module 1212, a time correction value can be obtained.

[Configuration of first embodiment]
FIG. 2 is a diagram showing an example of the configuration of the time correction device according to the first embodiment. A time correction device is an example of a device. The time correction device may be any device provided with a clock, such as an in-vehicle camera, a surveillance camera, a sound collection device, or the like.

As shown in FIG. 1, the time correction device 10 includes a communication section 11, a storage section 12, a clock section 13, a sensor section 14, and a control section 15. Further, the time correction device 10 has a camera, a microphone, etc. (not shown), and functions as a vehicle-mounted camera, a surveillance camera, a sound collection device, etc.

The communication unit 11 is an interface for communicating data with other devices. The communication unit 11 is, for example, a NIC (Network Interface Card). Further, the communication unit 11 is an interface connected to input devices such as a mouse and a keyboard. Further, the communication unit 11 is an interface connected to output devices such as a display and a speaker.

The storage unit 12 is realized by, for example, a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 12 stores correction module information 121 and device information 122.

The correction module information 121 is a parameter for constructing a correction module. For example, if the correction module is a regression function, the correction module information 121 is a regression coefficient. Further, for example, when the correction module is a neural network, the correction module information 121 includes weights, biases, and the like.

The device information 122 is information regarding the device (time correction device 10). For example, the device information 122 includes a history of starting and stopping the operation of the device, operating time, and the like.

The clock unit 13 is a clock that outputs the current time. It is assumed that the time output by the clock unit 13 can be corrected by setting.

The sensor section 14 is one or more sensors. For example, the sensor unit 14 includes a temperature sensor, a humidity sensor, and a voltage sensor.

The control unit 15 is a controller, and various programs (not shown) stored in the storage unit 12 are executed by, for example, a CPU (Central Processing Unit) and an MPU (Micro Processing Unit) using the RAM as a work area. This is achieved by Further, the control unit 15 can be realized by, for example, an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

The control unit 15 includes an acquisition unit 151, a calculation unit 152, an update unit 153, and a correction unit 154.

The acquisition unit 151 acquires data used for learning the correction module and correcting the time. The acquisition unit 151 acquires the current time from the clock unit 13. The acquisition unit 151 also acquires sensor values from the sensor unit 14 . Further, the acquisition unit 151 acquires information regarding the device from the device information 122. Further, the acquisition unit 151 acquires the reference time from an external device via the communication unit 11.

The calculation unit 152 inputs the input information to the correction module and calculates a time correction value. The calculation unit 152 constructs a correction module based on the correction module information 121.

The update unit 153 updates the correction module information 121 based on the calculation result by the calculation unit 152. Learning of the correction module is performed by updating the parameters by the updating unit 153.

The updating unit 153 updates the parameters using a method depending on the correction module. For example, when the correction module is a regression function, the updating unit 153 updates the correction module information 121 using the least squares method. Further, for example, when the correction module is a neural network, the updating unit 153 updates the correction module information 121 using the error backpropagation method.

The correction unit 154 corrects the clock. The correction unit 154 sets the current time of the clock to the time that is the sum of the time correction value calculated by the calculation unit 152 to the current time indicated by the clock.

The control unit 15 performs learning of the correction module through processing by the calculation unit 152 and the update unit 153. Further, the control unit 15 corrects the clock through processing by the calculation unit 152 and the correction unit 154. The method of calculating the time correction value in learning and correction is as explained in FIG.

The control unit 15 causes the correction module to learn the relationship between the information regarding the device and the deviation (difference in time) of the time (current time) of the clock provided in the device from the reference time. The information regarding the device corresponds to the input information in FIG. 1, and is, for example, information acquired by the acquisition unit 151 from the device information 122 and the sensor unit 14. Furthermore, the control unit 15 uses the learned correction module to correct the time of the clock provided in the device.

More specifically, the control unit 15 receives input information including the frequency of starting and stopping the operation of the device, the operating time of the device, and the sensor value measured by the sensor provided in the device, and The correction module learns the relationship between the deviation of the clock time from the reference time. Then, the control unit 15 corrects the time of the clock provided in the device based on the predicted value of the deviation obtained by inputting the input information acquired after the learning of the correction module to the learned correction module. conduct.

Note that the correction module learning the relationship means that the updating unit 153 updates the correction module information 121 according to the output of the correction module.

This allows the device to correct its own clock by storing the learned correction module. As a result, the time of the device can be corrected at low frequency without using a highly accurate clock that provides a reference time.

Further, the control unit 15 may cause the correction module to learn the relationship between the degree of deterioration based on the operating environment of the device and the deviation of the time of the clock provided in the device from the reference time. In this way, by using an index common to each device, such as the degree of deterioration, as an input to the correction module, it is possible to use a correction module having a similar structure in each device.

In this case, the calculation unit 152 calculates the degree of deterioration of the device based on the information acquired by the acquisition unit 151. Calculation of the degree of deterioration is performed by a predetermined module or the like. For example, the degree of deterioration increases as the cumulative operating time increases.

Then, the calculation unit 152 inputs the degree of deterioration to the correction module and calculates a time correction value. In this case, the degree of deterioration corresponds to input information.

[Processing of the first embodiment]
The flow of the learning process will be explained using FIG. 3. FIG. 3 is a flowchart showing the flow of learning processing.

As shown in FIG. 3, first, the time correction device 10 inputs input information to a correction module and calculates a time correction value (step S101). The time correction device 10 acquires input information from the device information 122 and the sensor section 14 .

Next, the time correction device 10 calculates the difference between the current time and the reference time (step S102).

Then, the time correction device 10 updates the parameters of the correction module so that the time correction value approaches the difference (step S103). The time correction value is a value output by the correction module based on input information.

The flow of the correction process will be explained using FIG. 4. FIG. 4 is a flowchart showing the flow of the correction process.

As shown in FIG. 4, first, the time correction device 10 acquires input information (step S201).

Next, the time correction device 10 inputs the input information to the correction module and calculates a time correction value (step S202). Then, the time correction device 10 corrects the time of the clock based on the time correction value (step S203).

Here, if the time correction value (predicted value of deviation) obtained from the correction module exceeds the threshold, the time correction device 10 adjusts the clock installed in the device based on the reference time obtained from another device. The time can be corrected.

As the threshold value, a value such as "5 minutes" is preset as an allowable range in which the time correction value can be used. In this case, if the absolute value of the time correction value obtained from the correction module exceeds "5 minutes," the time correction device 10 corrects the clock using the reference time without using the time correction value. do.

This is because the accuracy of the time correction value by the correction module may decrease due to deterioration of the device itself, changes in the environment, etc. According to the above method, even if the accuracy of the time correction value by the correction module decreases, the clock can be corrected with high accuracy.

Further, the time correction device 10 may perform additional learning or relearning of the correction module when the time correction value obtained from the correction module exceeds an allowable range. Additional learning means further learning of the current correction module using newly obtained data as training data. Relearning means performing learning after initializing the current correction module.

[Effects of the first embodiment]
The time correction device 10 causes a correction module to learn the relationship between information about the device and the deviation (difference in time) of the time (current time) of a clock provided in the device from a reference time. The time correction device 10 uses a learned correction module to correct the time of a clock provided in a device.

Thereby, the device (time correction device 10) can use the learned correction module to correct the time of the device at a low frequency without using a highly accurate clock that provides a reference time.

FIG. 5 is a diagram illustrating the effects of the first embodiment. FIG. 5 shows the difference between the reference time and the current time (the time indicated by the clock of the device) between the conventional technology and this embodiment.

Line 200A represents the current time of the prior art. Furthermore, line 200B represents the current time in this embodiment. In either case, the current time is corrected based on the reference time at the timing of receiving the reference time. Further, in this embodiment, the current time is corrected at any time using a learned correction module even at times other than the timing of receiving the reference time.

As shown in FIG. 5, in this embodiment, the deviation of the current time from the reference time is smaller than in the conventional technology. Therefore, in this embodiment, even if reception of the reference time is omitted, the current time can be corrected with high accuracy.

[Second embodiment]
A second embodiment will be described using FIG. 6. FIG. 6 is a diagram illustrating a configuration example of a time correction system according to the second embodiment.

In the second embodiment, a server calculates time correction values for a plurality of devices equipped with clocks and performs clock correction.

As shown in FIG. 6, the time correction system 2 includes a server 30, a device 40a, a device 40b, and a device 40c.

The server 30 is connected to each device via a network such as the Internet.

The device 40a is an in-vehicle camera. The device 40b is a surveillance camera installed in a residence. The device 40c is a surveillance camera installed in a factory. The device of the second embodiment is not limited to the one illustrated here.

Device 40a, device 40b, and device 40c all have a clock and a sensor. That is, the device 40a, the device 40b, and the device 40c have the same functions as the clock section 13 and the sensor section 14 of the time correction device 10 of the first embodiment.

The server 30 uses information regarding each device to learn the correction module and correct the time of each device. The server 30 has the same functions as the acquisition unit 151, calculation unit 152, update unit 153, and correction unit 154 of the time correction device 10. The server 30 also stores parameters of learned or unlearned correction modules.

The devices 40a, 40b, and 40c record images acquired by cameras, and transmit data of the recorded images to the server 30.

At this time, the devices 40a, 40b, and 40c add information regarding each device to the data to be transmitted to the server, and transmit the data. Thereby, the device 40a, the device 40b, and the device 40c provide the server 30 with information regarding each device.

For example, the device 40a, the device 40b, and the device 40c are given the usage environment, year of manufacture, and frequency of power ON (start of operation) of each device. Further, similarly to the first embodiment, the device 40a, the device 40b, and the device 40c may be provided with sensor information, device operating status, and cumulative operating time.

Moreover, the devices 40a, 40b, and 40c may transmit the information to be provided to the server 30 in advance, not at the same time as transmitting the data.

For example, the device 40a is assigned "inside a car, Japan (Hokkaido)" as the usage environment, "2017" as the year of manufacture, and "once a week for 2 hours" as the power-on frequency.

For example, for the device 40b, "inside the house, Japan (Hokkaido)" is assigned as the usage environment, "2012" is assigned as the year of manufacture, and "10 hours every day" is assigned as the power-on frequency.

For example, for the device 40c, "Outdoor, Japan (Hokkaido)" is assigned as the usage environment, "2015" is assigned as the year of manufacture, and "24 hours a day" is assigned as the frequency of power ON.

The server 30 further acquires the current time from the devices 40a, 40b, and 40c. Then, the server 30 performs learning of the correction module using the acquired information.

The server 30 may prepare a correction module for each device, or may prepare a correction module common to all devices.

Then, the server 30 uses the learned correction module to correct the clock of each device.

The server 30 also functions as a system that uses images, sounds, and other data (time-series sensor logs, etc.) recorded by the plurality of devices described above for processing such as synthesis, judgment, and selection.

For example, the server 30 not only corrects the clocks of each device, but also corrects and synchronizes the time stamps of images acquired from each device, and can use this for analysis.

Furthermore, the server 30 can predict when a device will fail based on the time correction value calculated for each device.

For example, the server 30 approximates a change in the time correction value for each device to a linear or nonlinear function, and predicts a time when the time correction value exceeds a threshold value as a failure time.

According to the second embodiment, since the server 30 has a function related to clock correction, there is no need to update the software of each device, and it is possible to analyze data without being affected by devices that have not been updated. It can be performed.

Further advantages and modifications can be easily deduced by those skilled in the art. Therefore, the broader aspects of the invention are not limited to the specific details and representative embodiments shown and described above. Accordingly, various changes may be made without departing from the spirit or scope of the general inventive concept as defined by the appended claims and their equivalents.

10 Time correction device 11 Communication unit 12 Storage unit 13 Clock unit 14 Sensor unit 15 Control unit 30 Server 40a, 40b, 40c, 51a, 51b Equipment 50 Providing device 121 Correction module information 122 Equipment information 151 Acquisition unit 152 Calculation unit 153 Update unit 200A, 200B lines 1211, 1212 correction module

Claims (6)

A time correction device that corrects the time of a clock provided in a device, comprising a controller,
The controller includes:
causing a module to learn the relationship between information about the device and a deviation of the time of a clock provided in the device from a reference time;
A time correction device that uses the learned module to correct the time of a clock provided in the device. The controller includes:
From input information including frequency of start and stop of operation of the device, operating time of the device, and sensor value measured by a sensor provided in the device, and a reference time of a clock provided in the device. Let the module learn the relationship between the deviation of and
The time of a clock provided in the device is corrected based on the predicted value of the deviation obtained by inputting the input information acquired after learning of the module to the learned module. The time correction device described in . The controller includes:
The time correction device according to claim 1, wherein a module is made to learn a relationship between a degree of deterioration based on the operating environment of the device and a deviation from a reference time of a clock provided in the device. The controller includes:
If the predicted value of the deviation obtained from the learned module exceeds a threshold, the time of a clock provided in the device is corrected based on a reference time obtained from another device. The time correction device described in . A time correction method performed by a controller of a time correction device that corrects the time of a clock provided in a device, the method comprising:
The controller includes:
Learning a module with information about the device as an explanatory variable and the deviation of the clock time from the reference time as the objective variable, using information about the device and the deviation from the reference time of the clock provided in the device as training data. conduct,
A time correction method, in which the time of a clock provided in the device is corrected based on a deviation that is an objective variable obtained by inputting information regarding the device as an explanatory variable into the learned module. A time correction system comprising a server and a plurality of devices connected to the server,
The device includes a clock and a first controller,
the first controller provides information regarding the device to the server;
The server includes a second controller,
The second controller causes the module to learn the relationship between the information provided by the plurality of devices and the deviation of clocks provided in the plurality of devices from a reference time,
A time correction system that uses the learned module to correct the time of clocks provided in the plurality of devices.

Images