Snow removal determination support device

Abstract

A model learning device 34 according to the present disclosure determines the necessity for snow removal at a point at a given distance on a road on which a vehicle is traveling, on the basis of the frequency characteristics of vibration intensity at the point at a given distance on the road, the frequency characteristics being detected on or under the surface of the road.

Description

Snow removal judgment support device

[0001] The present disclosure relates to optimization of snow removal determination, which determines whether snow removal is necessary.

[0002] Snow removal work is carried out to remove obstacles and hazards to traffic. Although the standards for dispatching snow removal work vary among local governments, the standard for dispatching snow removal work on roadways is generally set at 5 to 10 cm of snow accumulation. In order to determine the amount of snow accumulation, it is possible to use the amount of snowfall up to the time of inspection and subsequent weather forecast data.

[0003] However, in reality, in order to complete snow removal work at night before people start commuting to work or school the next day, it is often necessary to monitor the road surface conditions of each road in real time and issue instructions for snow removal work on the same day.

[0004] For example, Patent Document 1 proposes a system that collects data on road conditions, traffic conditions, and weather information from a terminal device mounted on a vehicle via a network to assist in making decisions about snow removal.

[0005] Patent No. 7144586

[0006] Distributed Acoustic / Vibration Sensing (DAS / DVS), APSENSING advanced photonic, Internet <URL: https: / / www.apsensing.com / > com / technology / distributed-acoustic-sensing-das-dvs>NEC Provides AI-based traffic monitoring system with fiber-optic sensing technology for NEXCO CENTRAL -Smart road traffic monitoring through existing optical fiber cable for communications-, NEC Corporation, Tokyo, May 24, 2022, Internet <URL: https: www. nec.com / en / press / 202205 / global_20220524_01.html> Tatsuro Matsuoka, "Ground Characteristics and Road Traffic Vibrations," Noise Control, Vol. 3, No. 2, 1979

[0007] On the other hand, snow removal requires significant costs, including equipment maintenance costs and labor costs, and it is necessary to perform snow removal work effectively within the allocated budget. However, the system described in Patent Document 1 requires the deployment of numerous terminals and sensors to acquire information on many vehicles or roads, and also requires workers to board vehicles and travel to the site to acquire data. This makes it impossible to reduce costs, and economic efficiency and maintainability become issues.

[0008] Therefore, in order to solve the above problem, the present disclosure aims to make it possible to determine remotely whether snow removal is necessary without deploying a large number of terminals or sensors to detect snow accumulation.

[0009] In order to solve the above problem, the determination device according to the present disclosure employs a method of remotely determining snow accumulation based on vibration intensity while the vehicle is traveling.

[0010] Specifically, the present disclosure provides a determination device that determines the need for snow removal at an arbitrary distance on a road on which a vehicle is traveling, based on the frequency characteristics of vibration intensity at the arbitrary distance on the road detected on the surface or underground of the road.

[0011] With this configuration, it is possible to remotely determine whether snow removal is necessary based on the frequency characteristics of vibration intensity, without having to go to the site and check the actual road surface conditions, etc.

[0012] In the above configuration, it is preferable to determine the need for snow removal at the arbitrary distance point on the road based on the frequency characteristics and the speed of the vehicle traveling on the road obtained based on the time change of the vibration intensity along the length of the road.

[0013] According to this configuration, by using a plurality of feature quantities, namely, the vehicle speed and the frequency characteristics of the vibration intensity, which are independent of each other, in combination, it is possible to suitably determine whether snow removal is necessary.

[0014] In the above configuration, it is preferable that the relationship between the frequency characteristics and the speed of the vehicle traveling on the road, and the necessity of snow removal at the arbitrary distance point on the road, be learned.

[0015] With this configuration, a prediction model can be suitably generated based on the frequency characteristics, the vehicle speed, and the determination result of the need for snow removal.

[0016] In the above configuration, it is preferable that the vibration intensity is obtained by distributed optical vibration sensing using an optical fiber cable laid under the road as a measurement medium.

[0017] With this configuration, it is possible to obtain the vibration intensity on the road simply by installing an optical fiber cable underground, which is economical.

[0018] The above disclosures can be combined as much as possible.

[0019] According to the present disclosure, it is possible to determine remotely whether snow removal is necessary without deploying a large number of terminals or sensors for detecting accumulated snow.

[0020] 1 is a diagram illustrating an overview of the configuration of a snow removal and removal determination system according to an embodiment of the present disclosure. FIG. 2 is a diagram illustrating a method for acquiring vibration data using a distributed optical vibration sensing device. FIG. 3 is a diagram illustrating a method for generating vehicle speed data. FIG. 4 is a spectrogram obtained by performing time-frequency analysis on the time series data of FIG. 4. FIG. 5 is a spectrogram obtained by performing time-frequency analysis on the time series data of FIG. 6. FIG. 6 is a graph illustrating the correlation between vehicle speed data and frequency characteristic data. FIG. 7 is a graph illustrating the relationship between the median vehicle speed and the median bandwidth power in the 50-100 Hz range. FIG. 8 is a flowchart illustrating a series of processes related to machine learning in a snow removal and removal determination system according to an embodiment of the present disclosure. FIG. 9 is a flowchart illustrating a snow removal and removal determination process based on a predictive model.

[0021] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Note that the present disclosure is not limited to the embodiments shown below. These implementation examples are merely illustrative, and the present disclosure can be implemented in various forms with various modifications and improvements based on the knowledge of those skilled in the art. Note that components with the same reference numerals in this specification and drawings indicate the same components.

[0022] (Overview of Snow Removal Determination System) A snow removal determination system 300 according to an embodiment of the present disclosure will be described with reference to FIGS. 1 to 6. FIG. 1 is a diagram illustrating an overview of the configuration of the snow removal determination system 300. The snow removal determination system 300 includes a distributed optical vibration sensing device 31, a data analysis device 32, a statistics calculation device 33, and a model learning device 34. The data analysis device 32 includes a vibration data storage unit 32A. The statistics calculation device 33 includes a vehicle speed database 33A and a frequency characteristics database 33B. The model learning device 34 includes a teacher data storage unit 34A, a prediction model generation unit 34B, and a determination unit 34C. The snow removal determination system 300 calculates various statistics from the vibration data acquired by the distributed optical vibration sensing device 31 and generates a prediction model based on the various statistics. The model learning device 34 functions as a "determination device."

[0023] Specifically, the snow removal determination system 300 uses a prediction model trained by machine learning using past statistics of vehicle speeds on the target road, statistics of frequency characteristics of vibration data detected on the road surface or underground, and actual snow removal data as training data, to determine whether snow removal is necessary on the road in question based on the current statistics of vehicle speeds on the road in question and statistics of frequency characteristics of vibration data detected on the road surface or underground.

[0024] The distributed optical vibration sensing device 31 outputs vibration intensity or power when a vehicle passes over a roadway as vibration data to a data analysis device 32 using an optical fiber cable. The vibration data output from the distributed optical vibration sensing device 31 is stored in a vibration data storage unit 32A of the data analysis device 32. The data analysis device 32 analyzes the vibration data and extracts the vehicle's travel trajectory, thereby generating vehicle speed data indicating the vehicle's speed and outputting the data to a statistics calculation device 33. The data analysis device 32 also generates frequency characteristic data from the results of FFT (Fast Fourier Transform) of the time-series data of vibration intensity at a certain distance point and outputs the data to the statistics calculation device 33.

[0025] The vehicle speed data and frequency characteristic data output from the data analysis device 32 are stored in a vehicle speed database 33A and a frequency characteristic database 33B, respectively, of the statistics calculation device 33. The vehicle speed database 33A converts the vehicle speed data into statistics. The frequency characteristic database 33B converts the frequency characteristic data into statistics. Each of the statistics is output to the model learning device 34.

[0026] The teacher data storage unit 34A of the model learning device 34 stores each statistical quantity output from the statistical quantity calculation device 33 as teacher data. The prediction model generation unit 34B performs machine learning based on the teacher data stored in the teacher data storage unit 34A to generate a prediction model for determining whether snow removal work is necessary. The determination unit 34C uses the generated prediction model to determine whether snow removal work is necessary from the real-time statistical quantities output from the statistical quantity calculation device 33. The determination result by the determination unit 34C is assigned to each statistical quantity as snow removal determination data and used in machine learning.

[0027] Specifically, the model learning device 34 of the snow removal determination system 300 determines the need for snow removal at any distance on a road based on the frequency characteristics of vibration intensity at that distance, detected on the surface or underground of the road on which a vehicle travels. The model learning device 34 also determines the need for snow removal at any distance on the road based on the frequency characteristics and the speed of the vehicle traveling on the road, which is obtained based on the time variation of vibration intensity along the length of the road. The model learning device 34 also learns the relationship between the frequency characteristics, the speed of the vehicle traveling on the road, and the need for snow removal at any distance on the road. The vibration intensity is obtained using distributed optical vibration sensing, using an optical fiber cable laid under the road as the measurement medium.

[0028] The functional configuration of each device will be described in detail below.

[0029] (Functions of the Distributed Optical Vibration Sensing Device) First, the functions of the distributed optical vibration sensing device 31 will be described. The distributed optical vibration sensing device 31 acquires various data via an optical fiber cable. The distributed optical vibration sensing device 31 outputs the change in the phase of scattered light at each point along the length of the optical fiber cable as vibration intensity or power through optical time domain analysis using pulsed light for sensing. The distance resolution along the optical length of the distributed optical sensing device 31 is determined by the pulse width used for optical vibration sensing and the analysis gauge length used to calculate the phase difference. As a result, the output data of the distributed optical vibration sensing device is three-dimensional data with measurement time in the elapsed time direction, position on the optical fiber in the distance direction, and vibration intensity in the signal intensity direction. Details of distributed optical vibration sensing are described, for example, in Non-Patent Document 1.

[0030] (Acquisition of Vibration Data) In this embodiment, as described above, when calculating the vehicle speed (vehicle speed data) and the frequency characteristics (frequency characteristic data), vibration data related to the vibration intensity generated when the vehicle travels on the roadway is used. The vibration intensity is acquired by the distributed optical vibration sensing device 31.

[0031] 2 is a diagram illustrating a method for acquiring vibration data using the distributed optical vibration sensing device 31. The distributed optical vibration sensing device 31 measures vibration intensity using an optical fiber cable OF connected to the distributed optical sensing device 31 and laid underground along the length of a road as a measurement medium. Specifically, the distributed optical vibration sensing device 31 detects vibrations along the optical fiber cable OF when a vehicle travels on the roadway, and acquires acoustic energy along the optical fiber cable OF.

[0032] The distributed optical vibration sensing device 31 outputs the acquired acoustic energy as vibration data to the vibration data storage unit 31A via the information transfer device 35. It is optional whether or not to provide the information transfer device 35. The distributed optical vibration sensing device 31 may also output the vibration data directly to the vibration data storage unit 31A.

[0033] (Generation of Vehicle Speed ​​Data) FIG. 3 is a graph illustrating a method for generating vehicle speed data by the data analysis device 32. FIG. 3 shows an example of vibration data output from the distributed optical vibration sensing device 31. Here, when a vehicle passes a certain point at a constant speed, a strong impulse-like vibration is applied to the road surface. In this case, as shown in FIG. 3, a linear trajectory is obtained in which the vibration generation point moves in the distance direction on the optical fiber cable OF over time. The slope of this linear trajectory is the vehicle speed (see, for example, Non-Patent Document 2). In this way, vehicle speed data is generated from the vibration data. The generated vehicle speed data is output to a vehicle speed database 33A of the statistics calculation device 33.

[0034] (Generation of Frequency Characteristic Data) The frequency components contained in vibrations generated when a vehicle travels on a snow-covered road surface vary depending on the road surface conditions and the amount of snow. Therefore, in this embodiment, frequency characteristic data is generated from vibration data to determine the amount of snow that requires snow removal. Generally, the amount of snow that requires snow removal is 5 cm to 10 cm or more. Figure 4 is time-series data that graphs the change over time in vibration intensity at a certain distance point in Figure 3. Figure 4 shows time-series data obtained from vibration data acquired on a dry road surface with no snow on the road surface. As shown in the figure, multiple or single impulse-like changes in vibration intensity are observed as the vehicle passes a certain distance point.

[0035] FIG. 5 shows a spectrogram obtained by time-frequency analysis of the time series data shown in FIG. 4 . In FIG. 5 , the instantaneous frequency obtained by short-time Fourier transform of the time series data can be observed. In this embodiment, since instantaneous frequency data is difficult to handle, the data analyzer 32 calculates a vibration power spectrum from the results of FFT of the time series data acquired at regular intervals and uses this as the basis for frequency characteristic data. Furthermore, the data analyzer 32 smooths and removes noise by averaging the vibration power spectrum at each point within a distance range where the road surface conditions are considered to be approximately similar (average vibration power spectrum). Once the average vibration power spectrum is obtained, the data analyzer 32 divides the average vibration power spectrum into a certain bandwidth. The data analyzer 32 converts each divided average vibration power spectrum into a multidimensional feature as a bandwidth power, and treats this as frequency characteristic data. The generated frequency characteristic data is output to the frequency characteristic database 33B of the statistics calculation unit 33.

[0036] (Comparison of frequency characteristics on a dry road surface with frequency characteristics on a snowy road surface) Figure 6 is time series data graphed to show the change over time in vibration intensity at a certain distance point in Figure 3. Figure 6 shows time series data obtained from vibration data acquired on a road surface that is snowy and determined to require snow removal. As shown in the figure, multiple or single impulse-like changes in vibration intensity are observed as the vehicle passes a certain distance point. Figure 7 is a so-called spectrogram obtained by performing a time-frequency analysis on the time series data in Figure 6.

[0037] Here, we compare the time series data and spectrograms for a dry road surface (Figures 4 and 5) with the time series data and spectrograms for a snowy road surface (Figures 6 and 7). It can be seen that the instantaneous frequency of impulse-like vibration changes when a vehicle passes over a dry road surface is dominated by signals of approximately 20 Hz or less, whereas the instantaneous frequency of impulse-like vibration changes when a vehicle passes over a snow-covered road surface requiring snow removal contains more signals in the range of approximately 20 Hz to 200 Hz, which is higher than 20 Hz. Thus, by taking frequency characteristics into account, it is possible to appropriately determine whether snow removal is required.

[0038] The scope of the present disclosure is not limited to simply performing frequency analysis using FFT as described above. Frequency analysis may also be performed using octave band analysis, mel-cepstral analysis, or other methods widely known as acoustic analysis. However, when using such analysis methods, it is important to actually incorporate high frequency bands of 20 Hz or higher as multidimensional features into a prediction model, which will be described later.

[0039] (Statistical Calculation) Next, a description will be given of the statistical calculation process performed by the statistical calculation device 33. The statistical calculation device 33 converts each of the vehicle speed data stored in the vehicle speed database 33A and the frequency characteristic data stored in the frequency characteristic database 33B into a statistical quantity.

[0040] Generally, the need for snow removal is determined based on the amount of snowfall up to the time of the inspection and subsequent weather forecast data. However, in reality, it is often necessary to monitor the road surface conditions in real time and issue instructions for snow removal work on the same day in order to complete snow removal work at night before people start commuting to work or school the next day.

[0041] Therefore, the statistical quantity calculation device 33 converts the vehicle speed data and frequency characteristic data into statistical quantities in real time. Specifically, the statistical quantity calculation device 33 calculates the average value and median value, which are statistical quantities of each piece of data obtained during a certain period of time, for example, from 9:00 a.m. to noon on the day of monitoring, as the feature quantity for that day. The statistical quantities of the vehicle speed data are output to the teacher data storage unit 34A of the model learning device 34 as vehicle speed statistical data. Furthermore, the statistical quantities of the frequency characteristic data are output to the teacher data storage unit 34A as frequency statistical data. Each of these statistical data is also output to the determination unit 34C in real time.

[0042] (Machine Learning) The training data storage unit 34A stores vehicle speed statistical data and frequency statistical data. Specifically, the training data storage unit 34A stores training data obtained by labeling past data acquired during a given period of time (e.g., data acquired during a given period last year) with snow removal determination data related to the result of the determination of whether snow removal was required at that time.

[0043] The prediction model generation unit 34B generates a prediction model for the necessity of snow removal as a classifier for supervised problems through machine learning using the training data. Specifically, the prediction model is generated by learning the relationship between each statistical data and the corresponding snow removal determination data. The machine learning model can be any model with high prediction accuracy from among representative classification models such as K Nearest Neighbor, support vector machines, decision trees, logistic regression, and neural networks.

[0044] The determination unit 34C uses the prediction model generated by the prediction model generation unit 34B to determine in real time whether snow removal is necessary based on each piece of statistical data output from the statistical quantity calculation device 33, and outputs the result. For example, the result of the snow removal necessity determination is displayed on a monitor (not shown). In this embodiment, because the statistical quantity calculation device 33 calculates each piece of statistical data in real time as described above, it is possible to determine in real time whether snow removal is necessary. In addition, the result of the snow removal necessity determination is fed back to the teacher data storage unit 34A as snow removal determination data for labeling each piece of statistical data.

[0045] In this embodiment, the prediction model is generated using statistical data related to the vehicle speed data and the frequency characteristic data, but the scope of the present disclosure is not limited to this. For example, the prediction model generation unit 34B can also generate a prediction model by adding publicly available data, such as meteorological data on snowfall and snow accumulation, and weather forecast data, to the features. However, these meteorological data are merely auxiliary features added to improve the accuracy of the prediction model generated using the vehicle speed data and frequency characteristic data. This is because such meteorological data is generally not available at geographical units, such as roads or construction areas, where snow removal needs to be determined, but is available only at larger geographical units, such as municipalities.

[0046] Next, the correlation between the vehicle speed data and frequency characteristic data used to generate the prediction model in this embodiment and the snow removal determination, as well as the correlation between these data, will be described. FIG. 8 is a graph showing the correlation between vehicle speed data and frequency characteristic data. First, it is common knowledge that the vehicle speed associated with the vehicle speed data decreases when conditions, such as snow accumulation, impede driving. Furthermore, regarding the frequency characteristic data, the frequency analysis experimental data shown in FIGS. 5 and 7 , for example, reveal a certain correlation between the frequency characteristics of vibrations applied to a road when a vehicle passes and the road surface conditions when snow accumulates. While past experiments have not addressed road surface conditions due to snow accumulation, there have been reports of changes in vibration intensity and frequency during vehicle passage depending on road surface conditions, such as road irregularities (see, for example, Non-Patent Document 3). From this, it can be easily inferred that road surface conditions, such as road irregularities, are also important features for determining whether or not vehicle driving is impeded. As in this embodiment, by using these multiple independent feature amounts (vehicle speed, frequency characteristics) in combination, highly correlated results can be expected for snow removal decisions when snow accumulates.

[0047] In Figure 8, the plots indicated by triangles represent acquired data for which snow removal was determined to be necessary, and the plots indicated by circles represent acquired data for which snow removal was determined to be necessary. The vertical axis of the graph represents the median vehicle speed, and the results are shown normalized by the median value over a certain period of time on a dry, snow-free road surface. The horizontal axis of the graph represents the frequency characteristics, specifically, as an example, the median value of bandwidth power in the 50-100 Hz range, and the results are shown normalized by the median value over a certain period of time on a dry, snow-free road surface, as with vehicle speed.

[0048] 8 clearly shows that the lower the vehicle speed and the higher the bandwidth power, the more data is deemed necessary for snow removal. It also shows that even if one were to try to determine the need for snow removal based solely on vehicle speed data, it would be difficult to do so. For this reason, it is effective to generate a prediction model by combining vehicle speed data with at least one or more frequency characteristic data, as in this embodiment.

[0049] The above has described the configuration of an embodiment for implementing the present disclosure using this embodiment, but the configuration of the present disclosure is not limited to such an embodiment, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present disclosure.

[0050] Next, a series of processes related to machine learning in the snow removal determination system 300 will be described with reference to Figure 9. Figure 9 is a flowchart illustrating a series of processes related to machine learning in the snow removal determination system 300 according to an embodiment of the present disclosure. First, the data analysis device 32 acquires vibration data output from the distributed optical vibration sensing device 31 (step S101). Next, the data analysis device 32 analyzes the vibration data and generates vehicle speed data and frequency characteristic data (step S102). The data analysis device 32 outputs the generated vehicle speed data and frequency characteristic data to the statistics calculation device 33.

[0051] The statistical quantity calculation device 33 converts the vehicle speed data stored in the vehicle speed database 33A and the frequency characteristic data stored in the frequency characteristic database 33B into statistical quantities (step S103). Each statistical quantity is output to the teacher data storage unit 34A of the model learning device 34.

[0052] The predictive model generation unit 34B of the model learning device 34 performs machine learning based on the training data stored in the training data storage unit 34A, and generates a predictive model for determining whether or not snow removal work is necessary (step S104).

[0053] Next, the snow removal determination process of the snow removal determination system 300 will be described with reference to Fig. 10. Fig. 10 is a flowchart illustrating the snow removal determination process based on a prediction model.

[0054] The statistics calculation device 33 outputs each calculated statistic to the teacher data storage unit 34A as described above, and simultaneously outputs it to the determination unit 34C (step S201). The determination unit 34C uses the generated prediction model to determine whether snow removal work is necessary based on the real-time statistics output from the statistics calculation device 33 (step S202). The determination result by the determination unit 34C is assigned to each statistic as snow removal determination data and used in machine learning.

[0055] The device of the present invention can also be realized by a computer and a program, and the program can be recorded on a recording medium or provided via a network. The program of the present disclosure is a program for causing a computer to realize each function of the device according to the present disclosure, and a program for causing a computer to execute each procedure of the method executed by the device according to the present disclosure.

[0056] 31: Distributed optical vibration sensing device 32: Data analysis device 32A: Vibration data storage unit 33: Statistical quantity calculation device 33A: Vehicle speed database 33B: Frequency characteristic database 34: Model learning device 34A: Teacher data storage unit 34B: Prediction model generation unit 34C: Determination unit 35: Information transfer device 300: Snow removal determination system

Claims

1. A determination device that determines the necessity of snow removal at an arbitrary distance point on a road on which a vehicle travels, based on frequency characteristics of vibration intensity at the arbitrary distance point on the road, the vibration intensity being detected on the surface or underground of the road.

2. The determination device of claim 1, which determines the need for snow removal at the given distance point on the road based on the frequency characteristics and the speed of the vehicle traveling on the road obtained based on the time change in vibration intensity along the length of the road.

3. The determination device according to claim 2, further comprising: a determining unit that learns a relationship between the frequency characteristics and the speed of the vehicle traveling on the road, and the necessity of snow removal at the arbitrary distance point on the road.

4. A determination device as claimed in any one of claims 1 to 3, wherein the vibration intensity is obtained by distributed optical vibration sensing using an optical fiber cable laid under the road as the measurement medium.

Images