JP2021151830A - Abnormity detection system, information process device, abnormity detection method and program for the same - Google Patents

Abstract

To provide a method by which promptly detects abnormity occurrence in a railway crossing by detecting actuation state of block bar and difference of prediction position.SOLUTION: When an electric train approaches a railroad crossing, an alarm makes alarming tone and lets plural red flashing lights blink alternately. When an information process device 30 detects blinking of the red flashing lights through an image of camera 10 or detects the alarming tone by sounds of a microphone 20 (S1), a timer starts timing (S2). The device detects a blockage bar by an image of the camera 10 (S3) and acquires distances in a predict coordinate and in real coordinate of positions P of the blockage bar at every elapsing time (S4). If the acquired distance is more than or equal to a prescribed value, the information process device 30 notifies an alarm to a monitoring center device 100 (S5). An attendant in the monitoring center confirms an alarming of the monitoring center device 100 and notifies the alarming to a display device of the train through an operation of the attendant (S6).SELECTED DRAWING: Figure 1

Description

The present invention relates to an abnormality detection system, an information processing device, an abnormality detection method, and an abnormality detection program capable of quickly detecting the occurrence of an abnormality at a railroad crossing.

Conventionally, railroad crossings where railroad tracks and roads intersect in a plane are dangerous areas where moving objects such as pedestrians or automobiles may collide with trains and cause serious accidents. For this reason, an alarm and a barrier are installed at the railroad crossing, and when a train approaches within a certain distance of the railroad crossing, two or more red flash lights provided on the alarm blink alternately and from the alarm. It makes a clicking sound, and then lowers the blocking bar of the automatic barrier to prevent moving objects from entering the railroad crossing.

However, even if the blocking bar is lowered, there are still many serious accidents at railroad crossings, so a technique for monitoring railroad crossings with a fixed camera is known. For example, in Patent Document 1 (Japanese Unexamined Patent Publication No. 2018-192844), when it is recognized that the railroad crossing blocking bar has been lowered by image detection by a fixed camera, the number of subjects in the monitoring area of the railroad crossing must be zero. A technique for outputting an alarm is disclosed.

JP-A-2018-192844

However, since this Patent Document 1 is a process after the blocking bar is lowered, it may take time to output an alarm. This is because the time from when the blocking bar goes down until the train passes the railroad crossing is short. Therefore, how to quickly detect the occurrence of an abnormality at a railroad crossing has become an issue.

The present invention has been made to solve the above problems, and provides an abnormality detection system, an information processing device, an abnormality detection method, and an abnormality detection program capable of quickly detecting the occurrence of an abnormality at a railroad crossing. The purpose is to do.

In order to solve the above problems, the present invention is an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a crossing including at least a flash lamp and a blocking bar. The information processing device includes a predictive means for predicting the operating state of the blocking rod that suppresses entry to the crossing based on the flash of the flash lamp or the alarm sound of the alarm, the prediction result by the predicting means, and the blocking rod. It is provided with an abnormality detecting means for detecting an abnormality in the monitoring area based on an operating condition.

Further, in the above invention, the present invention further includes an imaging means for capturing the monitoring region and a flash detecting means for detecting the flash of the flash lamp based on the image captured by the imaging means.

Further, in the above invention, the present invention excludes noise from the sound collecting means for collecting the sound in the monitoring area and the sound collected by the sound collecting means, and detects the alarm sound by the alarm device. Further provided with means.

Further, in the above invention, the predicting means predicts the position of the blocking bar every time a predetermined time elapses from the time when the flash of the flash lamp is generated or the time when the alarm sound is generated by the alarm device.

Further, in the above invention, in the above invention, when the difference between the prediction result by the prediction means and the operating state of the blocking rod is larger than a predetermined value, the abnormality detection means has an abnormality in the monitoring area. It is regarded as a thing and the occurrence of an abnormality is detected.

Further, in the above invention, the information processing device is a display device provided in a predetermined monitoring center device or a driver's seat of a train when an abnormality in the monitoring area is detected by the abnormality detecting means. The notification means for notifying the warning is further provided.

Further, the present invention is an abnormality detection system for detecting an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a crossing including at least a flashlight and a blocking bar. A trained model that inputs a partial image of the blocking bar included in the image of each predetermined elapsed time captured by the imaging device that images the monitoring area and outputs the risk probability for each predetermined elapsed time, and the trained model. An abnormal state with respect to an external device located outside the monitoring area based on the risk score calculating means for calculating the risk score from a plurality of risk probabilities output from the model and the risk score calculated by the risk score calculating means. It is provided with an abnormal state notification means for notifying.

Further, the present invention is an information processing device that is arranged in or near a railroad crossing monitoring area including at least a flashlight and a blocking bar and detects an abnormality in the monitoring area, and is a flash of the flashlight or an alarm sound by an alarm. An abnormality that detects an abnormality in the monitoring area based on a prediction means that predicts the operating state of the blocking rod that suppresses entry into the railroad crossing, a prediction result by the predicting means, and the operating status of the blocking rod. It is equipped with a detection means.

Further, the present invention is an abnormality detection method in an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a crossing including at least a flashlight and a blocking bar, and the information. A prediction step in which the processing device predicts the operating state of the blocking rod that suppresses entry to the crossing based on the flash of the flash lamp or the alarm sound by the alarm, and the information processing device predicts the prediction result by the prediction step. And an abnormality detection step of detecting an abnormality in the monitoring area based on the operating state of the blocking rod.

Further, the present invention is disposed in or near a monitoring area of a crossing including at least a flashlight and a blocking bar, and executes an abnormality detection process at the request of an information processing device that detects an abnormality in the monitoring area or the information processing device. An abnormality detection program executed by the server device, which is a prediction procedure for predicting the operating state of a blocking rod that suppresses entry to the crossing based on information related to the flash of the flash lamp or the alarm sound of the alarm. The computer is made to execute the abnormality detection procedure for detecting the abnormality in the monitoring area based on the prediction result by the prediction procedure and the operation state of the blocking rod.

Further, the present invention is arranged in or near a monitoring area of a crossing including at least a flashlight and a blocking bar, and executes an abnormality detection process at the request of an information processing device or the information processing device that detects an abnormality in the monitoring area. An abnormality detection program executed by the server device, in which a partial image of the blocking rod included in the image of each predetermined elapsed time captured by the imaging device that images the monitoring area is input, and every predetermined elapsed time is input. The above is based on a trained model that outputs a risk probability, a risk score calculation procedure that calculates a risk score from a plurality of risk probabilities output from the trained model, and a risk score calculated by the risk score calculation procedure. Have the computer execute the abnormal status notification procedure for notifying the external device located outside the monitoring area of the abnormal status.

According to the present invention, it is possible to quickly detect the occurrence of an abnormality at a railroad crossing.

FIG. 1 is an explanatory diagram of an outline of an abnormality detection system according to the first embodiment. FIG. 2 is a functional block diagram showing the configuration of the information processing apparatus shown in FIG. FIG. 3 is a diagram showing an example of the accumulated data, the prediction data, and the determination threshold data shown in FIG. FIG. 4 is an explanatory diagram of an outline of distance calculation between the current position and the predicted position of the blocking rod according to the first embodiment. FIG. 5 is a functional block diagram showing the configuration of the monitoring center device shown in FIG. FIG. 6 is a diagram showing an example of the in-vehicle display device data shown in FIG. FIG. 7 is a flowchart showing a processing procedure for generating predicted data in the information processing apparatus shown in FIG. FIG. 8 is a flowchart showing an abnormality detection procedure in the information processing apparatus shown in FIG. FIG. 9 is a flowchart showing a processing procedure of the monitoring center device shown in FIG. FIG. 10 is an explanatory diagram of an outline of the abnormality detection system according to the second embodiment. FIG. 11 is a functional block diagram showing the configuration of the information processing apparatus shown in FIG. FIG. 12 is an explanatory diagram for explaining the processing of the risk score calculation unit shown in FIG. FIG. 13 is a flowchart showing an abnormality detection procedure in the information processing apparatus shown in FIG.

Hereinafter, preferred embodiments of the abnormality detection system, the information processing device, the abnormality detection method, and the abnormality detection program according to the present invention will be described in detail with reference to the accompanying drawings. In the first embodiment shown below, a case where an image captured by the camera is image-processed to detect an abnormality will be shown. Further, in the second embodiment, a case where an abnormality is detected by using a trained model in which supervised learning is performed by deep learning will be shown.

[Embodiment 1]
<Overview of anomaly detection system>
First, an outline of the abnormality detection system according to the first embodiment will be described. FIG. 1 is an explanatory diagram for explaining an outline of the abnormality detection system according to the first embodiment.

Conventionally, railroad crossings where railroad tracks and roads intersect in a plane are dangerous areas where moving objects such as pedestrians, bicycles, carts, motorcycles or automobiles may collide with trains and cause serious accidents. For this reason, an alarm and a barrier are installed at the railroad crossing, and when a train approaches within a certain distance of the railroad crossing, two or more red flash lights provided on the alarm blink alternately and from the alarm. A warning sound is issued to notify the moving body of the arrival of the train. After that, the blocking bar of the automatic barrier is lowered to prevent the moving body from entering the railroad crossing.

However, when the number of trains passing through the railroad crossing is large, it may take time for the blocking bar to rise. For this reason, the moving body may start blinking red flash lights alternately, and even if a warning sound is issued from the alarm, the moving body may try to pass the railroad crossing forcibly and the moving body may get stuck around the railroad crossing. As a result, a serious accident occurs at the railroad crossing.

Here, as a conventional technique, when it is recognized that the railroad crossing blocking bar has fallen by image detection by a fixed camera installed at the railroad crossing, an alarm is output if the number of subjects in the railroad crossing monitoring area is not zero. However, since the train is already approaching the railroad crossing when the blocking bar descends, it is possible that even if the monitoring center etc. receives a warning of an abnormality at the railroad crossing, it will not be possible to take appropriate measures. There is sex.

Therefore, in the first embodiment, the prediction data that the blocking rod descends is generated in advance, it is determined whether or not the blocking rod is properly descending according to the prediction data, and the descending state of the blocking rod is the predicted data. If the difference is more than a predetermined threshold, an abnormality is reported. For example, if a truck or the like is present at a position where the tip of the blocking rod is lowered, a situation may occur in which the blocking rod cannot be lowered properly, but according to the configuration of the first embodiment, the blocking rod is completely lowered. Abnormality can be detected before. In addition, an abnormality can be detected even when a situation occurs in which a person manually holds down the blocking rod during descent. As described above, in the first embodiment, whether or not the blocking rod is properly lowered is regarded as an index for detecting an abnormality, and the abnormality is detected.

As shown in FIG. 1, a camera 10, a microphone 20, and an information processing device 30 are installed at a railroad crossing in order to detect the state of the railroad crossing, and are connected to each other by a communication line. In the description of FIG. 1, it is assumed that the predicted coordinates of the position P of the blocking bar for each elapsed time are calculated in advance. Specifically, the blinking of the red flash lamp is detected from the image of the camera 10, or the sound input to the microphone 20 is detected to include an alarm sound, and the blocking bar is detected from the image of the camera 10 for each elapsed time. The process of detecting the predicted coordinates of the position P of is repeated. After that, the average value of a plurality of predicted coordinates is obtained, and this average value is used as the predicted coordinate of the position P of the blocking bar for each elapsed time.

The camera 10 is fixedly installed, for example, on the upper part of an overhead wire pillar near a railroad crossing. If there is an overpass near the railroad crossing, the camera can be fixedly installed at a predetermined position on the overpass.

The information processing device 30 is installed near a railroad crossing or on the ground portion of an overhead wire pillar of an overpass, and is connected to the camera 10 by wire or wirelessly. The monitoring center is a center that monitors the status of a plurality of railroad crossings located on the route. The monitoring center device 100 is installed, and wireless communication is performed with the display device and the information processing device 30 mounted on the driver's seat of the train. Connected to be communicable.

When the train passes a position separated from the railroad crossing by a predetermined distance, the alarm is controlled to emit an alarm sound and alternately blink a plurality of red flash lights. The alarm sound is emitted from the alarm device and the blinking control of the red flash lamp is controlled by an existing control device different from the information processing device 30, but detailed description thereof will be omitted here.

If the information processing device 30 detects the blinking of the red flash lamp from the image of the camera 10 or detects that the sound input to the microphone 20 contains an alarm sound (S1), the information processing device 30 starts timing the timer. (S2).

A blocking bar (partial image of the blocking bar) is detected from the image of the camera 10 (S3), and the distance between the predicted coordinates of the blocking bar position P and the actual coordinates for each elapsed time is obtained (S4). When the obtained distance is equal to or greater than a predetermined value, the information processing apparatus 30 notifies the monitoring center apparatus 100 of an alarm (S5). In the first embodiment, a case where an alarm is notified to the monitoring center device 100 will be described, but the train operation schedule data (scheduled train crossing time, train display device IP address) is described in the information processing device 30. Etc.), the display device of the train to be notified can be specified based on the operation schedule data, and the alarm can be directly notified to the display device of the specified train. As a result, the driver can grasp the situation without delay.

The staff of the monitoring center confirms the alarm in the monitoring center device 100, and notifies the alarm to the display device of the train by the operation of the staff (S6). In the first embodiment, a case where an alarm is notified to the display device of the train by the operation of a staff member will be described. However, the monitoring center device 100 automatically determines whether or not the alarm should be notified, and the train is operated. It can also be configured to automatically notify the display device of an alarm. For example, it is possible to determine whether or not to automatically notify based on whether or not the distance between the predicted coordinates and the actual coordinates is larger than the past actual value. As a result, it is possible to reduce the burden on the staff of the monitoring center and eliminate human error by the staff. Further, the monitoring center device 100 excludes noise-like notifications caused by the shaking of the blocking rod due to the influence of wind or the like, and erroneous notifications to the driver can be reduced.

As described above, the abnormality detection system according to the first embodiment is configured to output an alarm to the monitoring center device 100 when the difference between the operating status of the blocking rod and the predicted position is equal to or more than a predetermined value. It is possible to quickly detect the occurrence of an abnormality.

<Configuration of information processing device 30>
Next, the configuration of the information processing apparatus 30 shown in FIG. 1 will be described. FIG. 2 is a functional block diagram showing the configuration of the information processing apparatus 30 shown in FIG. As shown in FIG. 2, the information processing device 30 is connected to the camera 10, the microphone 20, and the wireless communication device 40, and has a storage unit 31, a control unit 32, and a timer 33.

The camera 10 is a device that captures the situation of the blocking bar. The microphone 20 is a device that collects alarm sounds. The wireless communication device 40 is a device that connects the information processing device 30 and the monitoring center device 100 by wireless communication. The timer 33 is a timekeeping means for measuring the elapsed time after receiving the timekeeping start instruction.

The storage unit 31 is a storage device such as a hard disk device or a non-volatile memory, and stores stored data 31a, prediction data 31b, and determination threshold data 31c. The accumulated data 31a is data obtained by accumulating the coordinates of the position P of the blocking bar for each elapsed time measured in advance, for example, 10 times. The number of measurements is not limited to 10.

The prediction data 31b is the prediction data of the coordinates when the blocking bar operates normally. For example, the average of the coordinates of the position P of the blocking bar for each elapsed time included in the accumulated data 31a can be used as the prediction data 31b. It is meaningless to calculate the average value when the plurality of coordinates of the position P of the blocking bar for each elapsed time are considerably different. Therefore, in such a case, the measurement can be repeated until the coordinates fall within a certain range. The determination threshold data 31c is threshold data for determining whether or not the difference between the current position and the predicted position of the blocking bar is within a certain range.

The control unit 32 is a control unit that performs overall control of the information processing device 30, and includes an alarm sound reception unit 32a, an image reception unit 32b, a flash lamp detection unit 32c, a blocking bar detection unit 32d, a determination unit 32e, an alarm notification unit 32f, and the like. It has a prediction data generation unit 32g. Actually, by loading and executing these programs in the CPU (Central Processing Unit), the alarm sound reception unit 32a, the image reception unit 32b, the flash lamp detection unit 32c, the blocking bar detection unit 32d, the determination unit 32e, and the alarm The notification unit 32f and the prediction data generation unit 32g are made to execute the corresponding processes, respectively.

The alarm sound receiving unit 32a is a processing unit that receives an alarm sound included in the sound received from the microphone 20. For example, if the frequency pattern data of the alarm sound is stored in advance, noise such as an environmental sound is removed from the sound received from the microphone 20, and then the frequency pattern data of the alarm sound includes the frequency pattern data of the alarm sound. It is determined that the alarm sound has been received, and the timer 33 is instructed to start timing.

The image receiving unit 32b is a processing unit that receives an image transmitted from the camera 10. When a moving image (video) compressed by MPEG or the like is received from the camera 10, an image is extracted from the moving image at regular intervals.

The flash light detection unit 32c is a processing unit that detects the blinking of the red flash light included in the image received from the image reception unit 32b. For example, a partial image of the red flash lamp portion is cut out from the image received from the image receiving unit 32b, and the light emission of the red flash lamp is detected depending on whether or not a set of red pixels exists in the cut out partial image. It is desirable to detect the blinking of the red flash lamp by changing the color of a plurality of partial images forming a time series.

The blocking bar detection unit 32d is a processing unit that detects the current position of the blocking bar from the image received from the image receiving unit 32b. For example, a template image of the blocking bar is prepared in advance, it is determined whether or not the template image is included in the image received from the image receiving unit 32b, and if the template image is included, the blocking bar is included. The coordinate of the tip P of is detected as the current position. For this template image, it is desirable to prepare a plurality of images including a blocking bar that descends according to the elapsed time.

The determination unit 32e calculates the distance d between the coordinates of the tip P at the elapsed time t1 detected by the blocking bar detection unit 32d and the predicted coordinates of the tip P at the elapsed time t1 in the prediction data 31b. If the distance d exceeds the determination threshold data 31c, the determination unit 32e determines that an abnormality has occurred in the alarm notification unit 32f. When the determination unit 32e determines that an abnormality has occurred, the alarm notification unit 32f notifies the monitoring center device 100 of an alarm via the wireless communication device 40.

The prediction data generation unit 32g is a processing unit that calculates the predicted position (coordinates) of the tip P of the blocking bar for each elapsed time using the accumulated data 31a. For example, the average value of each of the 10 sets of X-coordinate and Y-coordinate data for each elapsed time stored in the accumulated data 31a is calculated, and the calculation result is stored in the prediction data 31b.

Next, an example of the accumulated data 31a, the prediction data 31b, and the determination threshold data 31c shown in FIG. 2 will be described. FIG. 3 is a diagram showing an example of the accumulated data 31a, the prediction data 31b, and the determination threshold data 31c shown in FIG.

The accumulated data 31a shown in FIG. 3A is the measurement result at the measurement date and time, that is, the data including the X coordinate and the Y coordinate of the tip P of the blocking rod for each elapsed time. Specifically, the accumulated data 31a has an X coordinate of "0.00" and a Y coordinate of "10.00" when the elapsed time is "1" seconds with respect to the measurement date and time "2020/03/01 10:25". When the elapsed time is "7" seconds, the X coordinate is "0.00" and the Y coordinate is "10.00". When the elapsed time is "8" seconds, the X coordinate is "2.35" and the Y coordinate is "9.70". At "13" seconds, the situation in which the X coordinate "9.56" and the Y coordinate "0.25" are associated with each other is shown.

Further, in the accumulated data 31a, with respect to the measurement date and time "2020/03/01 21:40", when the elapsed time is "1" seconds, the X coordinate is "0.00", the Y coordinate is "10.00", and the elapsed time is "1" seconds. In 7 "seconds, the X coordinate is" 1.12 ", the Y coordinate is" 9.78 ", and in the elapsed time" 8 "seconds, the X coordinate is" 3.78 ", the Y coordinate is" 8.95 ", and the elapsed time is" 13 ". In seconds, the X coordinate "10.00" and the Y coordinate "0.00" are associated with each other.

The prediction data 31b shown in FIG. 3B is data including the prediction position (X coordinate and Y coordinate) of the tip P of the blocking bar for each elapsed time. Specifically, the prediction data 31b has an X coordinate of "0.00" and a Y coordinate of "10.00" when the elapsed time is "1" seconds, and an X coordinate of "0.00" when the elapsed time is "7" seconds. When the Y coordinate is "10.00" and the elapsed time is "8" seconds, the X coordinate is "2.59" and the Y coordinate is "9.66". When the elapsed time is "13" seconds, the X coordinate is "10.00" and the Y coordinate is Y coordinate. The prediction data of "0.00" is shown. The determination threshold data 31c shown in FIG. 3C indicates that the determination threshold is “2.00”.

<Calculation of the distance between the current position of the blocking bar and the predicted position>
Next, an example of calculating the distance between the current position of the blocking bar and the predicted position will be described. FIG. 4 is an explanatory diagram for explaining an example of calculating the distance between the current position of the blocking rod and the predicted position. Here, the attachment point of the blocking bar of the blocking device included in the image captured by the camera 10 is set as the origin of the coordinates, the vertical direction of the image is set as the Y axis, and the predetermined direction orthogonal to the Y axis is set as the X axis. The scales of the X-axis and the Y-axis may be set to match the actual length, or may be set to 10 equal parts in an easy-to-understand manner.

In the XY plane, the position of the point P at the tip of the blocking bar is represented by the X coordinate and the Y coordinate. Assuming that the predicted position of the point P is (x ₁ , y ₁ ) and the current position is (x ₂ , y ₂ ), the distance d between the two points can be obtained by the following equation.
d = {(x ₁ − x ₂ ) ² + (y _{1 −} y ₂ ) ² } ^1/2 (Equation 1)

Although the case where the distance between two points on the XY plane is directly obtained is shown here for convenience of explanation, the distance between the two points on the Hough plane can also be obtained by performing the Hough transform. The Hough transform expresses a point (x, y) in the XY coordinate system by an angle θ and a distance ρ, and transforms it into a θρ coordinate system. One point in the XY coordinate system is transformed as one point in the θρ coordinate system by the Hough transform, and the distance between the two points can be calculated in the same manner as in the XY coordinate system.

<Configuration of monitoring center device 100>
Next, the configuration of the monitoring center device 100 shown in FIG. 1 will be described. FIG. 5 is a functional block diagram showing the configuration of the monitoring center device 100 shown in FIG. As shown in FIG. 5, the monitoring center device 100 is connected to the display unit 101, the input unit 102, and the wireless communication device 110, and has a storage unit 103 and a control unit 104.

The display unit 101 is a liquid crystal panel, a display device, or the like. The input unit 102 is a keyboard, a mouse, or the like. The wireless communication device 110 is a device that connects the monitoring center device 100, the information processing device 30, and the display device mounted on the train vehicle by wireless communication.

The storage unit 103 is a storage device such as a hard disk device or a non-volatile memory, and stores the in-vehicle display device data 103a. The in-vehicle display device data 103a is data that stores the vehicle number and the display device number of the train.

The control unit 104 is a control unit that performs overall control of the monitoring center device 100, and includes an alarm receiving unit 104a, a display control unit 104b, and an alarm transfer unit 104c. Actually, by loading and executing these programs in the CPU, the alarm receiving unit 104a, the display control unit 104b, and the alarm transfer unit 104c are made to execute the corresponding processes, respectively.

The alarm receiving unit 104a is a processing unit that receives an alarm from the information processing device 30 via the wireless communication device 110. The display control unit 104b is a processing unit that displays the alarm received by the alarm receiving unit 104a on the display unit 101.

The alarm transfer unit 104c is a processing unit that transfers the alarm received from the information processing device 30 to the display device mounted on the train vehicle. When the alarm transfer unit 104c receives an alarm transfer instruction by the operation of a staff member from the input unit 102, the alarm transfer unit 104c changes from a monitoring system that monitors the operation status of a train vehicle (not shown) to a railroad crossing where the information processing device 30 is installed. Receive the vehicle number of the approaching train. The alarm transfer unit 104c searches for the vehicle number from the in-vehicle display device data 103a, and transfers the alarm received from the information processing device 30 to the associated display device.

Next, an example of the in-vehicle display device data 103a shown in FIG. 5 will be described. FIG. 6 is a diagram showing an example of the in-vehicle display device data 103a shown in FIG. In the in-vehicle display device data 103a shown in FIG. 6, the display device number “XYZ567” is associated with the vehicle number “AB1234”.

<Processing procedure for generating predicted data in the information processing device 30>
Next, the processing procedure for generating the predicted data in the information processing apparatus 30 will be described. FIG. 7 is a flowchart showing a processing procedure for generating predicted data in the information processing apparatus 30. First, if the blinking of the red flash lamp or the alarm sound is detected (step S101; Yes), the coordinates of the tip P of the blocking rod at regular intervals when the blocking rod descends are detected (step S102), and the detection is performed. The coordinates are stored in the accumulated data 31a (step S103).

Then, if a predetermined number (for example, 10) of coordinates for each elapsed time are stored in the stored data 31a (step S104; Yes), the predicted position (coordinates) is based on the plurality of coordinates stored in the stored data 31a. ) Is calculated (step S105), and the process is terminated. For example, the average value of each coordinate is calculated. If the coordinates for each elapsed time are less than a predetermined number (for example, 10) (step S104; No), the process proceeds to step S101 and the same process is repeated.

By performing such a series of processes, it is possible to generate prediction data. It is desirable to update this forecast data on a regular basis. This is because the shape of the blocking rod may change due to aging or the like.

<Processing procedure for detecting an abnormality in the blocking bar in the information processing device 30>
Next, the processing procedure for detecting an abnormality in the blocking bar in the information processing apparatus 30 will be described. FIG. 8 is a flowchart showing a processing procedure for detecting an abnormality in the blocking bar in the information processing apparatus 30. Here, it is assumed that the prediction data 31b is generated in advance.

First, if the blinking of the red flash lamp or the alarm sound is detected (step S201; Yes), the timer 33 starts timing (step S202). A partial image of the blocking bar included in the image received from the camera 10 is extracted in a situation where the elapsed time of the timer 33 is a predetermined time (step S203).

The coordinates of the tip P of the blocking bar are extracted from the partial image of the blocking bar, the distance d is calculated from the coordinates and the coordinates corresponding to the elapsed time in the predicted data 31b (step S204), and the distance d and the determination threshold data 31c are calculated. (Step S205).

If the distance d exceeds the determination threshold value (step S205; Yes), an alarm is notified to the monitoring center device 100 (step S206), and the process ends. If the distance d is equal to or less than the determination threshold value (step S205; No), the time t from the start of timing is compared with the preset detection set time (step S207).

If the time t is equal to or less than the detection set time (step S207; No), the process proceeds to step S203 and the same process is repeated. If the time t exceeds the detection set time (step S207; Yes), the process ends.

By performing the above series of processes, it is possible to detect an abnormality before the blocking rod is fully lowered, so that it is possible to prevent the occurrence of a serious accident at a railroad crossing.

<Processing procedure of monitoring center device 100>
Next, the processing procedure of the monitoring center device 100 will be described. FIG. 9 is a flowchart showing a processing procedure of the monitoring center device 100. First, when the monitoring center device 100 receives an alarm from the information processing device 30 (step S301; Yes), the staff member by displaying the warning content on the display unit 101 and issuing a warning sound as necessary. An alarm is notified to (step S302).

After that, the monitoring center device 100 receives the instruction from the staff (step S303; Yes), and if the instruction content is not the transfer of the alarm (step S304; No), the process ends as it is.

On the other hand, if the instruction content by the staff is the transfer of an alarm (step S304; Yes), the monitoring system that monitors the operation status of the train vehicle (not shown) approaches the railroad crossing where the information processing device 30 is installed. Receive the train vehicle number (step S305). The vehicle number is searched from the in-vehicle display device data 103a, an alarm is transferred to the associated display device (step S306), and the process ends.

As described above, the abnormality detection system according to the first embodiment can quickly detect the occurrence of an abnormality at a railroad crossing by detecting the difference between the operating state of the blocking rod and the predicted position.

[Embodiment 2]
By the way, in the above-described first embodiment, the case where the abnormality detection of the state of the blocking bar is performed based on the distance between the prediction data and the measured data of the blocking bar is shown. If this occurs, a situation may occur in which the monitoring center device 100 is notified of the abnormality even though it is not in an abnormal state. Therefore, in the second embodiment, a case where abnormality detection is performed using a trained model in which supervised learning is performed by deep learning will be described.

<About the trained model>
In the second embodiment, a convolutional neural network (CNN) is used when generating a trained model. A trained model is generated by inputting teacher data to this CNN and performing supervised learning. Since CNN and deep learning are well-known techniques, detailed description thereof will be omitted here, but an outline of supervised learning will be described here.

In the second embodiment, an image including the blocking bar is captured by the camera 10 at predetermined elapsed time from the time when the red flash lamp or the alarm sound is detected (the time when the timekeeping starts). For example, if images taken 1 second, 2 seconds, 3 seconds, ..., 20 seconds after the start of timekeeping are captured, there will be 20 images. After that, 20 partial images are acquired by cutting out a partial image each consisting of a rectangular region including a blocking bar from the 20 images. These 20 partial images are a series of images showing the process of descending the blocking bar.

Here, the CNN is made to perform supervised learning by using the pair of these partial images and the risk probabilities as teacher data. This risk probability is normalized to 0 to 1, and the closer the risk probability is 0, the lower the risk level, and the closer the risk probability is 1, the higher the risk level. In addition, a plurality of pairs consisting of a partial image and a risk probability are input to the CNN to perform supervised learning. Further, the partial image is processed by the existing image processing to generate teacher data having a high risk probability, and the CNN is made to perform supervised learning. By performing such supervised learning using a large number of samples, the parameters in the CNN become appropriate values, and a trained model is generated. If i partial images are input to this trained model, the risk probability pi of each partial image is output from the trained model.

<Overview of anomaly detection system>
Next, the outline of the abnormality detection system according to the second embodiment will be described. FIG. 10 is an explanatory diagram for explaining the outline of the abnormality detection system according to the second embodiment. Here, it is assumed that the trained model has been generated in advance.

In the second embodiment, a partial image of the blocking bar is extracted from the image showing the descending state of the blocking bar captured by the camera 10, the partial image is input to the trained model, and the risk probability is output from the trained model. Then, the risk score is calculated based on this risk probability. If this risk score exceeds a predetermined threshold value, an alarm is notified to the monitoring center device 100.

As shown in FIG. 10, a camera 10, a microphone 20, and an information processing device 200 are installed at a railroad crossing and are connected to each other by a communication line. The camera 10 and the microphone 20 are the same as those in the first embodiment. The information processing device 200 has the same configuration as the information processing device 30 of the first embodiment, except that processing using the trained model is performed.

If the information processing device 200 detects the blinking of the red flash lamp from the image of the camera 10 or detects that the sound input to the microphone 20 contains an alarm sound (S11), the information processing device 200 starts timing the timer. (S12).

If a blocking bar (partial image of the blocking bar) is detected from the image of the camera 10 (S13), one or more partial images acquired from the start of the timer to the present time are input to the trained model (S14). .. As a result, the risk probability corresponding to the number of partial images is output from the trained model, and the risk score is calculated based on this risk probability (S15). For example, when the number of partial images is n, n risk probabilities are output from the trained model. Therefore, the risk score R can be obtained from the following equation. However, i = 1 to n.
R = Σ (pi) / n (Equation 2)

Here, for convenience of explanation, a case where a partial image of the blocking bar is input to the trained model will be described, but the present invention is not limited to this, and the partial image including the entire barrier is used as the trained model. It can also be applied when inputting. This is because the positional relationship between the barrier body and the barrier bar is considered to be an important parameter in calculating the risk score of the trained model. It can also be applied to a case where a feature amount image obtained by extracting a feature amount from a partial image of a blocking bar or a partial image including the entire blocking machine by differential processing or the like is input to a trained model.

After calculating the risk score in this way, if the risk score is equal to or higher than a predetermined value, the information processing device 200 notifies the monitoring center device 100 of an alarm (S16). As in the first embodiment, the information processing device 200 stores the train operation schedule data, identifies the train display device to be notified based on the operation schedule data, and identifies the specified train display device. It can also be configured to notify the alarm directly to. As a result, the driver can grasp the situation without delay.

The staff of the monitoring center confirms the alarm in the monitoring center device 100, and notifies the alarm to the display device of the train by the operation of the staff (S17). Similar to the first embodiment, the monitoring center device 100 can be configured to automatically determine whether or not to notify the alarm and automatically notify the train display device of the alarm. As a result, it is possible to reduce the burden on the staff of the monitoring center and eliminate human error by the staff. Further, the monitoring center device 100 excludes noise-like notifications caused by the shaking of the blocking rod due to the influence of wind or the like, and erroneous notifications to the driver can be reduced.

As described above, the abnormality detection system according to the second embodiment inputs the partial image of the blocking rod into the trained model, calculates the risk score from the risk probability output from the trained model, and the risk score is a predetermined value. Since it is configured to output an alarm to the monitoring center device 100 when the number exceeds the above, it is possible to quickly detect the occurrence of an abnormality at the crossing.

<Configuration of information processing device 200>
Next, the configuration of the information processing apparatus 200 shown in FIG. 10 will be described. FIG. 11 is a functional block diagram showing the configuration of the information processing apparatus 200 shown in FIG. The description of the functional unit similar to that of the information processing apparatus 30 shown in FIG. 2 will be omitted.

As shown in FIG. 11, the information processing device 200 is connected to the camera 10, the microphone 20, and the wireless communication device 40, and has a storage unit 31, a control unit 220, and a timer 33. The same parts as those in FIG. 2 are designated by the same reference numerals, and detailed description thereof will be omitted.

The learned model 210 is stored in the storage unit 31. The trained model 210 has already been described, but here, a case where the trained model 210 is implemented by software is shown. However, the trained model 210 can also be configured using a hardware circuit such as an ASIC or FPGA.

The partial image extraction unit 221 is a processing unit that cuts out a partial image of a rectangular region including a blocking bar from the image received by the image reception unit 32b. When performing such cutting, a template for the blocking bar can be created in advance, and template matching between this template and the image can be performed. Further, the trained model 1 in which the image of the railroad crossing is input 1 and the partial image of the blocking bar or the entire barrier is output 2, and the partial image of the blocking bar or the entire barrier that is output 2 is input 2 and the risk score is output. It is also possible to perform the cutting process using the trained model 2 set to 2. Further, it is also possible to perform the cutting process using the trained model 3 in which the image of the railroad crossing is input 3 and the risk score is output 3. This is because it is considered that the correlation between the input and the output can be estimated by the trained model.

The risk score calculation unit 222 is a processing unit that calculates the risk score based on the risk probability output from the trained model. Specifically, as shown in FIG. 12, if n partial images acquired so far are input to the trained model 210, n risk probabilities pi (i = 1 to 1) are input from this trained model. n) is output. Therefore, the risk score R is calculated by the formula of R = Σ (pi) / n.

The determination unit 223 is a processing unit that compares the danger score R with a predetermined value (determination threshold data 31c) and determines that an abnormality has occurred when the danger score R exceeds the predetermined value. Specifically, as shown in FIG. 12, when the predetermined value is "0.8" and the risk score R> 0.8, it is determined that an abnormality has occurred. When the determination unit 223 determines that an abnormality has occurred, the alarm notification unit 224 notifies the monitoring center device 100 of an alarm via the wireless communication device 40.

<Processing procedure for detecting an abnormality in the blocking bar in the information processing device 200>
Next, a processing procedure for detecting an abnormality in the blocking bar in the information processing apparatus 200 will be described. FIG. 13 is a flowchart showing a processing procedure for detecting an abnormality in the blocking bar in the information processing apparatus 200. Here, it is assumed that the trained model 210 is generated in advance.

First, if the blinking of the red flash lamp or the alarm sound is detected (step S401; Yes), the timer starts timing (step S402). A partial image of the blocking bar included in the image received from the camera 10 when the elapsed time has reached a predetermined time is extracted (step S403).

After that, by inputting n partial images up to the present time into the trained model 210 (step S404), n risk probabilities are output from the trained model 210. The risk score R is calculated using the n risk probabilities (step S405).

If the risk score R exceeds a predetermined determination threshold value (step S406; Yes), an alarm is notified to the monitoring center device 100 (step S407), and the process ends. If the risk score R is equal to or less than the determination threshold value (step S406; No), the process ends.

By performing the above series of processes, the trained model 210 can be used to detect an abnormality before the blocking bar is fully lowered, so that the occurrence of a serious accident at a railroad crossing can be prevented.

In the above-described first and second embodiments, the case where the abnormality detection of the blocking rod status is performed by the information processing apparatus has been described, but the present invention is not limited to this, and the abnormality detection of the blocking rod status is performed. It can also be performed by a server device, a cloud, an edge computer, or the like. In such a case, the information processing device transmits data to the server device, cloud, edge computer, or the like, and the server device, cloud, edge computer, or the like performs processing related to the information processing device according to the first or second embodiment. It will be. Therefore, the server device, cloud, edge computer, or the like has a configuration related to the information processing device according to the first or second embodiment. Further, the information processing device and the server device may be in charge of the processing in a distributed manner, and both devices may cooperate to perform the processing.

Further, each configuration shown in each of the above embodiments is a schematic function, and does not necessarily have to be physically illustrated. That is, the form of distribution / integration of each device is not limited to the one shown in the figure, and all or part of the device is functionally or physically distributed / integrated in an arbitrary unit according to various loads and usage conditions. Can be configured.

The abnormality detection system, information processing device, abnormality detection method, and prediction detection program according to the present invention can quickly detect the occurrence of an abnormality at a railroad crossing by detecting the difference between the operating status of the blocking rod and the predicted position. Suitable for.

10 Camera 20 Mike 30 Information processing device 31 Storage unit 31a Accumulated data 31b Prediction data 31c Judgment threshold data 32 Control unit 32a Alarm sound reception unit 32b Image reception unit 32c Flash light detection unit 32d Blocker bar detection unit 32e Judgment unit 32f Alarm notification unit 32g Prediction data generation unit 33 Timer 40 Wireless communication device 100 Monitoring center device 101 Display unit 102 Input unit 103 Storage unit 103a In-vehicle display device data 104 Control unit 104a Alarm reception unit 104b Display control unit 104c Alarm transfer unit 110 Wireless communication device 200 Information Processing device 210 Trained model 220 Control unit 221 Partial image extraction unit 222 Danger score calculation unit 223 Judgment unit 224 Alarm notification unit

In order to solve the above problems, the present invention is an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a crossing including at least a flash lamp and a blocking bar. the information processing apparatus, the particular Ri by the prediction means and said predicting means for specifying a predicted position of a predetermined point in advance the flashlamp flashes or blocking bar a predetermined time from the time it is detected an alarm sound by alarm has elapsed The monitoring is based on the predicted position of the predetermined location and the position of the predetermined location on the blocking rod after a predetermined time has elapsed from the time when the flash of the flash lamp or the alarm sound by the alarm is detected at the crossing. It is provided with an abnormality detecting means for detecting an abnormality in an area.

Further, in the above invention, the predetermined time is shorter than the time required to reach the closed position of the blocking rod .

Further, the present invention is an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near a monitoring area of a crossing including at least a flashlight and a blocking bar, and the information processing device is a method for detecting an abnormality in the monitoring area. It is generated by performing supervised learning by deep learning on a multi-layer neural network using the pair of the image including the blocking bar and the risk probability indicating the possibility of collision between the moving body and the train as training data. A trained model that inputs a partial image of the blocking bar included in the image of each predetermined elapsed time captured by the imaging device that images the monitoring area and outputs the risk probability for each predetermined elapsed time, and the trained model. Based on the danger score calculation means that calculates the danger score from the danger probability output from the model and the danger score calculated by the danger score calculation means, the abnormal state is notified to the external device located outside the monitoring area. It is provided with a means for notifying an abnormal state.

Further, the present invention is an information processing device that is arranged in or near a monitoring area of a crossing including at least a flashlight and a blocking bar and detects an abnormality in the monitoring area, and is an alarm by a flash of the flashlight or an alarm in advance. prediction means for identifying a predicted position of the predetermined point in blocking bar has elapsed a predetermined time from the time it is detected sound, and the predicted position of the predetermined portion of Ri identified by the prediction means, the flashlamp at said crossing It is provided with an abnormality detecting means for detecting an abnormality in the monitoring area based on the position of the predetermined position on the blocking rod in which a predetermined time has elapsed from the time when the flash of light or the alarm sound by the alarm device is detected.

Further, the present invention is an abnormality detection method in an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a crossing including at least a flashlight and a blocking bar, and is described in advance. a prediction step of specifying a predicted position of a predetermined point in flash or blocking bar a predetermined time from the time it is detected an alarm sound by the alarm device has passed the flashlamp, and the predicted position of the predetermined locations identified by said prediction step An abnormality detection step of detecting an abnormality in the monitoring area based on the position of the predetermined position on the blocking rod after a predetermined time has elapsed from the time when the flash of the flash lamp or the alarm sound by the alarm device is detected at the crossing. And include.

Further, the present invention is disposed in or near a monitoring area of a crossing including at least a flashlight and a blocking bar, and executes an abnormality detection process at the request of an information processing device that detects an abnormality in the monitoring area or the information processing device. An abnormality detection program executed by a server device, which is a prediction procedure for specifying a predicted position of a predetermined position on a blocking rod in which a predetermined time has elapsed from the time when a flash of the flash lamp or an alarm sound by an alarm is detected in advance. the predetermined portion of the prediction position of the predetermined portion of Ri identified by the prediction procedure, the blocking bar which the time from the time it is detected an alarm sound of a predetermined by flash or alarm of the flashlamp in crossing has passed The computer is made to execute an abnormality detection procedure for detecting an abnormality in the monitoring area based on the position of.

Further, the present invention is disposed in or near a monitoring area of a crossing including at least a flashlight and a blocking bar, and executes an abnormality detection process at the request of an information processing device or the information processing device that detects an abnormality in the monitoring area. An anomaly detection program executed by a server device, which uses a pair of an image including the blocking bar and a risk probability indicating the possibility of a moving object colliding with a train as training data for a multi-layer neural network. A partial image of the blocking rod, which is generated by performing supervised learning by deep learning and is included in the image for each predetermined elapsed time captured by the imaging device that images the monitoring area, is input, and every predetermined elapsed time is input. The monitoring area is based on a trained model that outputs a risk probability, a risk score calculation procedure that calculates a risk score from the risk probability output from the trained model, and a risk score calculated by the risk score calculation procedure. Have the computer execute the abnormal status notification procedure for notifying the external device located outside of the abnormal status.

Claims (11)

It is an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a railroad crossing including at least a flash lamp and a blocking bar.
The information processing device
A predictive means for predicting the operating state of the blocking bar that suppresses entry to the railroad crossing based on the flash of the flash lamp or the alarm sound of the alarm.
An abnormality detecting system including an abnormality detecting means for detecting an abnormality in the monitoring area based on a prediction result by the predicting means and an operating state of the blocking rod. An imaging means for imaging the monitoring area and
The abnormality detection system according to claim 1, further comprising a flash detection means for detecting a flash of the flash lamp based on an image captured by the image pickup means. The first aspect of claim 1, further comprising a sound collecting means for collecting sounds in the monitoring area and an alarm sound detecting means for detecting an alarm sound by an alarm device by excluding noise from the sounds collected by the sound collecting means. Anomaly detection system. The prediction means is
The abnormality detection system according to any one of claims 1 to 3, which predicts the position of a blocking bar each time a predetermined time elapses from the time when the flash of the flash lamp is generated or the time when the alarm sound is generated by the alarm device. The abnormality detecting means is
Any of claims 1 to 4 in which when the difference between the prediction result by the prediction means and the operating state of the blocking rod is larger than a predetermined value, it is considered that an abnormality has occurred in the monitoring area and the occurrence of the abnormality is detected. Anomaly detection system described in one. The information processing device
Any of claims 1 to 5, further comprising a notification means for notifying a predetermined monitoring center device or a display device provided in the driver's seat of a train when an abnormality is detected in the monitoring area by the abnormality detecting means. Anomaly detection system described in one. It is an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near the monitoring area of a railroad crossing including at least a flash lamp and a blocking bar.
The information processing device
A trained model that inputs a partial image of the blocking bar included in the image of each predetermined elapsed time captured by the imaging device that images the monitoring area and outputs the risk probability for each predetermined elapsed time.
A risk score calculation means for calculating a risk score from a plurality of risk probabilities output from the trained model, and
An abnormality detection system including an abnormal state notification means for notifying an external device located outside the monitoring area of an abnormal state based on the danger score calculated by the danger score calculating means. An information processing device that is arranged in or near a railroad crossing monitoring area including at least a flashlight and a blocking bar and detects an abnormality in the monitoring area.
A predictive means for predicting the operating state of the blocking bar that suppresses entry to the railroad crossing based on the flash of the flash lamp or the alarm sound of the alarm.
An information processing device including an abnormality detecting means for detecting an abnormality in the monitoring area based on a prediction result by the predicting means and an operating state of the blocking rod. It is an abnormality detection method in an abnormality detection system that detects an abnormality in the monitoring area in an information processing device arranged at or near a railroad crossing monitoring area including at least a flashlight and a blocking bar.
A prediction step in which the information processing device predicts the operating state of the blocking bar that suppresses entry to the railroad crossing based on the flash of the flash lamp or the alarm sound of the alarm.
An abnormality detection method including an abnormality detection step in which the information processing apparatus detects an abnormality in the monitoring area based on a prediction result by the prediction step and an operating state of the blocking rod. It is arranged in or near the monitoring area of the crossing including at least a flashlight and a blocking bar, and is executed by an information processing device that detects an abnormality in the monitoring area or a server device that executes an abnormality detection process at the request of the information processing device. Anomaly detection program
A prediction procedure for predicting the operating state of the blocking bar that suppresses entry to the railroad crossing based on the information related to the flash of the flash lamp or the alarm sound by the alarm device, and the prediction procedure.
An abnormality detection program that causes a computer to execute an abnormality detection procedure for detecting an abnormality in the monitoring area based on the prediction result of the prediction procedure and the operating status of the blocking bar. It is arranged in or near the monitoring area of the crossing including at least a flashlight and a blocking bar, and is executed by an information processing device that detects an abnormality in the monitoring area or a server device that executes an abnormality detection process at the request of the information processing device. Anomaly detection program
A trained model that inputs a partial image of the blocking bar included in the image of each predetermined elapsed time captured by the imaging device that images the monitoring area and outputs the risk probability for each predetermined elapsed time.
A risk score calculation procedure for calculating a risk score from a plurality of risk probabilities output from the trained model, and a risk score calculation procedure.
An abnormality detection program that causes a computer to execute an abnormal state notification procedure for notifying an external device located outside the monitoring area based on the risk score calculated by the risk score calculation procedure.

Images