JPH09245284A - Traffic situation monitoring device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To calculate traffic quantity information of a position with abnormality occurrence with high precision and to analyze a traffic situation with high precision even when abnormality occurs in a vehicle detecting equipment which is arranged at the important position by permitting the vehicle detecting equipment with abnormality occurrence to calculate tranffic quantity information to be outputted through the use of a neural network method. SOLUTION: When the vehicle detecting equipment 14d arranged at a comparatively important specified position 12d becomes abnormal in traffic monitoring in road networks 11a-11e, the vehicle running quantity QD' of the specified position 12d is obtained through the use of a neural network 19 from the respective vehicle running quantities which are detected by the respective vehicle detecting equipments 14a, 14b, 14c and 14e of the respective positions 12a, 12b, 12c and 12e adjacent to the specified position. Then, a jam occurrence detecting part 18 judges jam occurrence based on the calculated vehicle running quantity QD' and the respective vehicle running quantities of the other normal positions 12a-12c and 12e. Therefore, the traffic situation is analyzed with high precision and the proper road situation is provided to the drivers of respective vehicles 13.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traffic condition monitoring apparatus, and more particularly to a traffic condition monitoring method using traffic volume information of a position where a vehicle detector installed is out of order or a position where the vehicle detector is not permanently installed. The present invention relates to a traffic condition monitoring device that monitors traffic.

[0002]

2. Description of the Related Art Detecting traffic information at each position in a road network, analyzing traffic conditions such as occurrence of traffic congestion at each position, and displaying the traffic information on an electronic display device arranged at each position on the road. A traffic condition monitoring device has been put into practical use, which alerts the driver by broadcasting or broadcasting on a radio.

For example, as shown in FIG. 6, the main positions 2a, 2 of the roads 1a, 1b, 1c, 1d constituting the road network are shown.
Vehicle detectors 4a, 4b, 4c for detecting traffic volume information indicated by, for example, the vehicle traffic volume Q per unit time of the vehicle 3 passing through the corresponding positions 2a-2c are provided at b, 2c. The traffic volume information detected by the vehicle detectors 4a, 4b, 4c is collected at the monitoring center 5.

In the monitoring center 5, at each position 2a.n in the road network based on the traffic information detected by each vehicle detector 4a, 4b, 4c. The congestion information of 2b and 2c is grasped, the presence or absence of the congestion is judged, and when the congestion occurs, the occurrence information is displayed together with the occurrence position on the electronic display device or the like via the alarm output unit 6. In addition, the occurrence of traffic congestion may be predicted and displayed together with the occurrence position.

In this case, the judgment of traffic jam occurrence or the prediction of traffic jam occurrence uses not only the traffic information of the corresponding position but also the traffic information of each position before and after the vehicle traveling direction of the corresponding position.

In such a traffic condition monitoring apparatus, when one vehicle detector arranged at each of the positions 2a to 2c of the road network fails, there is a missing value in the traffic volume information.
There is a problem that an accurate traffic situation analysis cannot be implemented.

In order to eliminate such inconvenience, a simple arithmetic interpolation method is used to obtain the traffic volume information of the position of the faulty detector from the traffic volume information of each vehicle detector arranged at each position before and after it. The traffic volume information of the missing position is calculated by using, and the traffic situation analysis is performed using the calculated traffic volume information.

[0008]

However, the traffic condition monitoring apparatus which calculates the traffic information of the defective vehicle detector by using the arithmetic interpolation method as described above should be solved as follows. There was a problem.

That is, when the road network does not have a large confluence, intersection or branch point in the installation intervals of the vehicle detectors such as a highway or a highway in the suburbs, the above-mentioned simple arithmetic interpolation method is used. Even if the traffic volume information of the missing position is calculated by using it, the traffic volume information does not deviate significantly from the correct traffic volume information.

However, as shown in FIG. 6, in the center of a large city with heavy traffic, each vehicle flows from many side roads and many intersections to the installation position of the defective vehicle detector,
Discharge. Therefore, even if the traffic volume information of the corresponding position is simply obtained by using the arithmetic interpolation method that is composed of the traffic volume information detected by the vehicle detectors installed at the positions before and after the position, There may be a large deviation from the quantity information.

For example, in FIG. 6, a vehicle merges at a central position 2b of the road 1a from a road 1d intersecting with the road 1a and is discharged to the corresponding road 1d. Therefore, if the vehicle detector 4b at the central position 2b fails,
Even if the arithmetical interpolation is simply performed from the traffic information of the vehicle detectors 4a and 4b at the positions 2a and 2c on both sides of the corresponding road 1d, it is obvious that the correct traffic information of the corresponding position 2b cannot be obtained. is there. As a result, there arises a problem that the traffic situation cannot be analyzed correctly.

In particular, when the vehicle detector 4b arranged at an important position fails, there is a concern that it cannot be dealt with at all.
The present invention has been made in view of such circumstances,
By using the neural network method, even if an abnormality occurs in the vehicle detector arranged at an important position, the traffic volume information at the corresponding position is accurately calculated. It is an object of the present invention to provide a traffic condition monitoring device capable of performing accurate traffic condition analysis and improving the reliability of the entire device.

Further, by preparing only one moving vehicle detector, it is possible to accurately detect the traffic volume information at arbitrary positions and at a plurality of positions between the positions where the vehicle detector is not permanently installed, so that a more precise traffic situation can be obtained. It is an object of the present invention to provide a traffic condition monitoring device that can be analyzed.

[0014]

SUMMARY OF THE INVENTION The present invention is directed to a plurality of vehicle detectors arranged at a plurality of positions in a road network and detecting traffic information at each position, and to each vehicle detector detected by each vehicle detector. The present invention is applied to a traffic condition monitoring device including a traffic condition analysis unit that analyzes the traffic condition at each position in a road network based on traffic information and outputs it as road traffic information.

In order to solve the above-mentioned problems, according to the first aspect of the present invention, the traffic condition monitoring device has an abnormality in a vehicle detector disposed at a predetermined specific position among the plurality of vehicle detectors. When the traffic detection information detected by the vehicle detector at each position adjacent to the vehicle detector at the specific position is input to the input layer and the abnormality detection means for detecting the Based on the neural network that outputs traffic information and the actual traffic information that is output from the vehicle detector at the specific position and the traffic information that is output from the output layer during the period when the abnormality detection unit does not detect an abnormality The learning unit that learns each weighting coefficient between the middle layer, the input layer, and the output layer, and the traffic volume information output from the output layer is abnormally generated during the period when the abnormality detection unit detects the abnormality. Vehicle detector traffic And a switching means for delivering to the traffic status analyzing unit as broadcast.

Further, in the traffic condition monitoring apparatus according to the present invention, a plurality of abnormality detecting means for detecting abnormality of each vehicle detector, a corresponding position and a plurality of positions adjacent to the corresponding position are provided for each position. Grouping means for grouping and each traffic information detected by the vehicle detector at each position provided in each group grouped by this grouping means and adjacent to the center position of the corresponding group in the input layer A plurality of neural networks that output traffic information at the center position from the output layer via the middle layer, and an abnormality detection unit provided for each group, and abnormality detection means detects an abnormality in the vehicle detector at the center position of the corresponding group. Based on the actual traffic volume information output from the vehicle detector at the center position and the traffic volume information output from the output layer during the non-detection period, each weight between the middle layer, the input layer and the output layer is calculated. A plurality of learning section for learning the coefficients,
The traffic of the vehicle detector at the center position where the abnormality detection stage detects the abnormality of the vehicle detector at the center position of the corresponding group is provided for each group It is provided with a plurality of switching means for sending the traffic amount information to the traffic condition analysis unit.

Further, in the traffic condition monitoring apparatus according to the present invention, a moving vehicle detector which is movably provided at an arbitrary position in the road network and detects traffic information at a temporary position of the moving destination, and a temporary installation at the input layer. When each traffic volume information detected by the vehicle detector at each position adjacent to the position is input, the neural network that outputs the traffic volume information of the temporary position from the output layer to the traffic situation analysis unit via the intermediate layer, and the temporary network A learning unit that learns each weighting coefficient between the middle layer, the input layer, and the output layer based on the actual traffic volume information output from the moving vehicle detector at the position and the traffic volume information output from the output layer. Is equipped with.

In the traffic condition monitoring apparatus thus constructed, when the vehicle detector disposed at the specific position becomes abnormal, the vehicle detectors at the respective positions adjacent to the specific position detect each of them. The traffic volume information of the specific position is obtained from the traffic volume information using the neural network, and the traffic situation analysis is executed based on the calculated traffic volume information and the traffic volume information of each regular position.

The weighting factors between the intermediate layer and the input layer and the output layer in the neural network are learned by the learning section during a normal period of the vehicle detector arranged at the specific position.

In another invention, for each position,
The corresponding position and a plurality of positions adjacent to the corresponding position are grouped. A neural network is provided for each group. Therefore, even if an abnormality occurs in the vehicle detector installed at any position, the traffic volume information at the corresponding position is calculated by the neural network corresponding to the corresponding group, so that the correct traffic situation analysis is executed.

Further, in another invention, a moving vehicle detector that can move to an arbitrary position in the road network is provided. Then, the traffic information of the temporary position where the moving vehicle detector is located is calculated by the neural network.
Therefore, by learning the weighting factor at each temporary position in advance by the learning unit, the traffic volume information of a plurality of positions where the vehicle detector is not actually installed can be obtained. Therefore,
More detailed traffic situation analysis of the required location can be obtained.

[0022]

Embodiments of the present invention will be described below with reference to the drawings. (First Embodiment) FIG. 1 is a block diagram showing the schematic arrangement of a traffic condition monitoring apparatus according to the first embodiment. Main position 1 of each road 11a, 11b, 11c, 11d that constitutes the road network
Positions 12 corresponding to 2a, 12b, 12c, 12d, 12e
Vehicle detectors 14a, 14b, 14c, 14 that detect the vehicle traffic volume per unit time such as 1 minute as traffic volume information of the vehicle 13 passing a to 12e every 1 minute.
d and 14e are provided. And each vehicle detector 1
The vehicle traffic amounts Q _A , Q _B , Q _C , Q _D , and Q _E detected at 4a, 14b, 14c, 14d, and 14e are collected at the monitoring center 15. In this embodiment, the central position 12d is assumed to be the specific position.

The traffic volume of each vehicle Q _{A to} Q _E is the monitoring center 1
5 is input to the input port 16 in The vehicle traffic amount Q _D at the specific position 12d output from the input port 16 is input to a traffic jam occurrence detection unit 18 as a traffic condition analysis unit via one of the switching terminals and a common terminal of the switching device 17. Also,
Vehicle passing amounts Q _A , Q at positions 12a to 12c, 12e other than the specific position 12d output from the input port 16
_B, Q _C, with Q _E is directly input to the traffic congestion detection unit 18, is input to the neural network 19.

The traffic jam occurrence detecting section 18 detects each input position 1
With each vehicle traffic amount Q _A to Q _E of 2a~12e to determine the presence or absence of traffic congestion at each position 12a to 12e, when it is determined that congestion occurs, via the display output unit 20, each road 1a~ Each electronic display device 2 arranged at a plurality of positions 1e
The traffic jam occurrence and the occurrence position are displayed and output to 1.

The neural network 19 uses the input port 16 to pass each vehicle traffic Q _A , Q every one minute.
_{When B} , Q _C , and Q _E are input, the vehicle traffic amount Q _D ′ at the specific position 12d is learned from the output terminal by the learning unit 23 and the switching unit 17.
To the switching terminal.

Further, the vehicle traffic amount Q _D at the specific position 12 d output from the input port 16 is input to the abnormality detecting section 22. The abnormality detection unit 22 uses the input vehicle traffic amount Q
_{If D} is normal, if it is abnormal, the switch 1
7 and the abnormality detection signal a to the learning unit 23.

The abnormality detecting unit 22 may be used, for example, when the detected vehicle traffic amount Q _D exceeds a predetermined limit amount, or when the other vehicle detectors installed in the vicinity detect the value. it is judged that an abnormal the appropriate vehicle traffic amount Q _D in the case of extremely different values for the detected value to be inferred from the vehicle traffic volume Q.

During a period in which the abnormality detection signal a is not input from the abnormality detection unit 22, the switching unit 17 determines the traffic amount Q _D output from the input port 16 as shown in FIG. Send to. Further, when the abnormality detection signal a is input from the abnormality detection unit 22, the vehicle traffic amount Q of the specific position 12d output from the neural network 19 is output.
_D ′ is sent to the traffic jam occurrence detector 18.

Accordingly, the traffic jam occurrence detector 18 determines the occurrence of traffic jam by using the vehicle traffic amounts Q _{A to} Q _E of the vehicle detectors 14a to 14e while the vehicle detector 14d at the specific position 12d is normal. Car detector 1 at specific position 12d
During the period when 4d is abnormal, the vehicle detector 14 at the specific position 12d
Instead of vehicle traffic amount Q _D of d, performs determination of traffic congestion using the vehicle traffic amount Q _D 'output from the neural network 19.

FIG. 2 is a schematic diagram showing a schematic configuration of the neural network 19. The neural network 19 is roughly divided into four vehicle traffic volumes Q _A. Q
_B, Q _C, an input layer 24 consisting of four elements that are input or Q _E, an intermediate layer 25 composed of N elements, the specific position 1
An output layer 26 including one element that outputs a vehicle traffic amount Q _D ′ of 2d, a plurality of input layer weighting factors Wij that connect each element of the input layer 24 and each element of the intermediate layer 25, and an intermediate layer Two
5 and each element of the output layer 26 are connected to the output layer weight coefficient Wxj. However, i = A, B. C.
E, and j = 1.2.3. …, N.

Every time one minute elapses, the vehicle traffic Q _D ′ at the specific position 12 d output from the output layer 26 is input to the learning unit 23. Further, the actual vehicle traffic amount Q _D output from the vehicle detector 14d at the specific position 12d is input to the learning unit 23 at the same timing.

Then, the learning section 23 uses the abnormality detection signal a
Learning operation is performed without inputting, that is,
Every time one minute elapses, the vehicle traffic amount Q _D ′ output from the output layer 26 and the actual vehicle detector 14 d at the specific position 12 d
The actual vehicle traffic amount Q _D output from the above is compared, and the above-mentioned input layer weighting factors Wij and output layer weighting factors Wxj are corrected so that they match.

Therefore, when the traffic condition monitoring device is operated for a certain period or longer, the weighting factors Wij and Wxj are learned by the learning unit 23 and finally set to the optimum value. When the abnormality detection signal a is input, since is not inputted normal vehicle traffic amount Q _D from the vehicle detector 14d of a specific position 12d, and stops the learning operation.

In the traffic condition monitoring apparatus thus configured, when the vehicle detector 14d arranged at the specific position 12d, which is relatively important for traffic monitoring on the road network, becomes abnormal, the vehicle position is detected at the specific position 12d. Each adjacent position 1
2a, 12b, 12c, 12e vehicle detectors 14a,
Each vehicle traffic amount Q detected in 14b, 14c, 14e
Neural network 19 from _A , Q _B , Q _C , Q _E
Is used to obtain the vehicle traffic amount Q _D ′ at the specific position 12 d, and the congestion occurrence detection unit 18 calculates the calculated vehicle traffic amount Q _D ′ and the remaining normal positions 12 a to 12 c, 1
Each vehicle traffic amount Q _A to Q _C of 2e, determination of traffic congestion based on the Q _E is executed.

In this case, the vehicle traffic Q _D ′ at the specific position 12 d calculated using the neural network 19
In addition, since the inflow and outflow contribution rates of the vehicle 13 from each direction, which cannot be analytically obtained, are statistically added, as in the conventional method shown in FIG. The accuracy is remarkably high as compared with the vehicle traffic amount obtained by the arithmetic interpolation method from the vehicle detectors 4a and 4c at the position.

Therefore, even if the vehicle detector at an important position fails, the road condition can be analyzed with high accuracy, and the driver of each vehicle 13 can be provided with a suitable road condition.

(Second Embodiment) FIG. 3 is a diagram showing a road network to which a traffic condition monitoring apparatus according to a second embodiment of the present invention is applied.

In this road network, a vehicle detector 31 is provided at each intersection position 32 on each road. Then, as shown in FIG. 4, for each vehicle detector 31 at each position 32, each group 33 is formed by each vehicle detector 31 at four positions 32 adjacent to this position 32. Therefore, when N vehicle detectors 31 are arranged, N groups 33 are formed.

A total of five vehicle detections, that is, the vehicle traffic of the vehicle detector 31 at the central position of the corresponding group 33 forming each groove 33 and the vehicle detection amounts of the vehicle detectors 31 at four adjacent surrounding positions are detected. The amount is input to each signal processing unit 35 in the monitoring center 34.

In each signal processor 35, shown in FIG.
The input port, the abnormality detection unit 22, the switching unit 17, the neural network 19, and the learning unit 23 are housed. Then, in the signal processing unit 35, the vehicle detection amount at the central position is calculated from the vehicle detection amounts at the four surrounding positions input by the neural network. When the vehicle detector at the central position is normal, the detected vehicle amount at the central position is sent to the next traffic jam occurrence detection unit 36 as it is. On the other hand, when the vehicle detector at the central position is abnormal, the vehicle detection amount at the central position output from the neural network is sent to the next traffic jam occurrence detection unit 36.

In this way, the detected traffic amount at the central position of each group 33 is input from each signal processing unit 35 to the congestion occurrence detection unit 36. Therefore, the congestion occurrence detection unit 36
Determines the occurrence of traffic congestion at each position based on the vehicle detection amount at each of the N positions 32, and outputs the determination result to the electronic display device 38 via the display output unit 37.

In the traffic condition monitoring apparatus thus constructed, the neural network is provided for all the positions in the road network. When the vehicle detector at the corresponding position fails, the neural network detects the vehicle at the corresponding position. The detected amount is output. Therefore, even if an abnormality occurs in the vehicle detector installed at any position, the monitoring center 34 can correctly perform the traffic condition analysis, and the reliability of the entire traffic condition monitoring device can be improved.

(Third Embodiment) FIG. 5 is a block diagram showing the schematic arrangement of a traffic condition monitoring apparatus according to the third embodiment of the present invention. The same parts as those in the first embodiment shown in FIG. 1 are designated by the same reference numerals. Therefore, the detailed description of the overlapping part is omitted.

In this embodiment, a movable vehicle detector 14m is prepared. The position where the moving vehicle detector 14m is installed is defined as the temporary position 12m.

The moving vehicle detector 14 at the temporary position 12 m
m and each position 12a, 1 adjacent to this temporary position 12m
2b, 12c, 12e vehicle detectors 14a, 14b, 1
4c, the detected number of vehicles Q _M of _{14e, Q A, Q B,} Q C Q
_E is input to the input port 16 of the monitoring center 15.

Temporary position 1 output from the input port 16
The vehicle traffic amount Q _M of 2 m is input to the learning unit 23. Further, each position 12a~12c other than temporary position 12m outputted from the input port 16, the vehicle traffic amount Q _A of 12e,
Q _B, Q _C, with Q _E is directly input to the traffic congestion detection unit 18, is input to the neural network 19.

The neural network 19 passes each vehicle traffic Q _A , Q from the input port 16 every one minute.
_{When B} , Q _C , and Q _E are input, the vehicle traffic amount Q _M ′ at the temporary position 12 m is output from the output terminal to the learning unit 23 and the congestion occurrence detection unit 18.

The traffic congestion detector 18, the positions 12a~12c entered, 12e, each vehicle traffic amount Q _A ~ of 12m
Q _C, Q _E, using Q _M each position 12 a to 12 c, 12
If it is determined that a traffic jam has occurred at e and 12 m, and if it is determined that a traffic jam has occurred, the occurrence of traffic jam and the occurrence position on each of the electronic display devices 21 arranged at a plurality of positions on each road via the display output unit 20. Is displayed and output.

The position data calculation control section 40 controls the start / stop of the learning section 23. That is, the moving vehicle detector 14
Learning section 23 while m is installed at a temporary position of 4 m
Is maintained in the operating state, and the learning unit 23 is maintained in the stopped state during the period in which the moving vehicle detector 14m is removed.

During the period in which the learning unit 23 is controlled to be in the operating state by the position data calculation control unit 40, the vehicle traffic amount Q _M output from the output layer 26 is output every one minute.
′ ′ Is compared with the actual vehicle traffic amount Q _M from the moving vehicle detector 14 m at the actual temporary position 12 m, and the above-mentioned input layer weighting factors Wij and output layer weighting factors Wxj are calculated so that they match. I will fix it.

Each weighting factor Wi to the extent that they substantially match
When j and Wxj are learned, the operator operates the position data collection unit 4
1 is started, and the learned weighting factors Wij and Wxj set in the neural network 19 are set to the temporary position 12m.
It is written in the weighting coefficient storage unit 39 together with the installation position D of.

In this way, while moving the installation position D (temporary position 12m) of the moving vehicle detector 14m sequentially, the optimum learned weighting factors Wij, W at each installation position D are obtained.
xj is sequentially written in the weighting coefficient storage unit 39.

In the traffic condition monitoring apparatus having such a configuration, when the moving vehicle detector 14m is removed from the road network and, for example, the arbitrary position D is designated to analyze the traffic condition, the weighting coefficient storage unit 39 Corresponding stored position D
The weighting factors Wij and Wxj of the position D closest to are set in the neural network 19 and the operation of the learning unit 23 is stopped.

As a result, the vehicle traffic Q _M ′ at a position near the arbitrary position D designated by the neural network 19
Is sent to the traffic jam occurrence detector 18. Traffic jam detector 1
8 performs, other positions 12a, 12b, 12c, each vehicle traffic amount of 12e Q _A to Q _C, the determination of the traffic congestion at each position based on the vehicle traffic volume Q _M 'from Q _E and neural networks 19 .

In this way, the traffic information of each temporary position 12m where the moving vehicle detector 14m is located is calculated by the neural network 19. Therefore, by learning the weighting factors Wij and Wxj at the temporary positions 12m in advance by the learning unit 23, the vehicle detectors 14a to 1a are actually used.
Traffic volume information Q _M ′ at a plurality of positions where 4e is not installed is obtained. Therefore, a more detailed traffic situation analysis of the required position D can be obtained, so that the reliability of the entire device can be further improved.

[0056]

As described above, in the traffic condition monitoring apparatus of the present invention, the traffic amount information to be output by the vehicle detector in which an abnormality has occurred is calculated by using the neural network method. Therefore, even if an abnormality occurs in the vehicle detector installed at an important position, the traffic volume information at the relevant position is accurately calculated, and by using this traffic volume information, highly accurate traffic condition analysis can be performed. It can be implemented and the reliability of the entire device can be improved.

Furthermore, each weighting coefficient to be set in the neural network is obtained in advance using one moving vehicle detector. Therefore, by using the neural network, it is possible to accurately detect the traffic volume information between arbitrary positions and plural positions between the positions where the vehicle detector is not permanently installed, and it is possible to perform a more detailed traffic situation analysis.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a traffic condition monitoring device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a neural network incorporated in the traffic condition monitoring device.

FIG. 3 is a schematic diagram showing a road network to which a traffic condition monitoring device according to a second embodiment of the present invention is applied.

FIG. 4 is a block diagram showing a schematic configuration of the traffic condition monitoring apparatus.

FIG. 5 is a block diagram showing a schematic configuration of a traffic condition monitoring device according to a third embodiment of the present invention.

FIG. 6 is a schematic diagram of a conventional traffic condition monitoring device.

[Explanation of symbols]

11a-11e ... Road, 12a-12e ... Position, 14a
-14e ... Vehicle detector, 14m ... Moving vehicle detector, 15
... Monitoring center, 16 ... Input port, 18, 36 ... Congestion detector, 19 ... Neural network, 20, 37 ...
Display output unit 21, 38 ... Electronic display device, 23 ... Learning unit, 24 ... Input layer, 25 ... Intermediate layer, 26 ... Output layer, 33
Group, 35 signal processing unit, 39 weighting factor storage unit.

Claims (3)

[Claims] 1. A plurality of vehicle detectors which are arranged at a plurality of positions in a road network and detect traffic amount information at each position, and the road based on each traffic amount information detected by each vehicle detector. In a traffic condition monitoring device including a traffic condition analysis unit that analyzes the traffic condition at each position in the network and outputs it as road traffic information, the traffic condition monitoring device being arranged at a predetermined specific position among the plurality of vehicle detectors. Anomaly detection means for detecting anomalies in the vehicle detector, and when the traffic information detected by the vehicle detector at each position adjacent to the vehicle detector at the specific position is input to the input layer, the intermediate layer is entered. A neural network that outputs the traffic information of the specific position from the output layer, and the actual traffic information and the output that are output from the vehicle detector of the specific position during a period in which the abnormality detecting unit does not detect an abnormality. Layers A learning unit that learns each weighting coefficient between the intermediate layer, the input layer, and the output layer based on the traffic volume information output from the output unit, and outputs the output during a period in which the abnormality detection unit detects an abnormality. A traffic condition monitoring device comprising: switching means for transmitting the traffic amount information output from the layer to the traffic condition analyzing unit as traffic amount information of a vehicle detector in which an abnormality has occurred. 2. A plurality of vehicle detectors which are arranged at a plurality of positions in a road network and detect traffic amount information at each position, and the road based on each traffic amount information detected by each vehicle detector. In a traffic condition monitoring device including a traffic condition analysis unit that analyzes the traffic condition of each position in the network and outputs it as road traffic information, a plurality of abnormality detection means for detecting an abnormality of each vehicle detector, and For each position, a grouping means for grouping the corresponding position and a plurality of positions adjacent to the corresponding position, and a grouping means provided for each group grouped by this grouping means are provided at the center position of the corresponding group in the input layer. A plurality of neural networks that output the traffic information at the central position from the output layer via the intermediate layer when the traffic information detected by the vehicle detectors at adjacent positions is input, Provided for each loop, while the abnormality detecting means does not detect an abnormality of the vehicle detector at the central position of the corresponding group, the actual traffic volume information output from the vehicle detector at the central position and the output layer A plurality of learning units that learn each weighting coefficient between each of the intermediate layer, the input layer, and the output layer based on the traffic amount information output from, and the abnormality detection stage is provided for each of the groups. During the period in which the abnormality of the vehicle detector at the central position of the group is detected, the traffic volume information output from the output layer is sent to the traffic situation analysis unit as the traffic volume information of the vehicle detector at the central position where the abnormality has occurred. Traffic condition monitoring device having a plurality of switching means. 3. A plurality of vehicle detectors arranged at a plurality of positions in a road network to detect traffic volume information at each position, and the road based on the traffic volume information detected by each vehicle detector. In a traffic condition monitoring device including a traffic condition analysis unit that analyzes the traffic condition at each position in the network and outputs it as road traffic information, a movable position is provided at any position in the road network, and When the moving vehicle detector for detecting traffic information and each traffic information detected by the vehicle detector at each position adjacent to the temporary position are input to the input layer, the temporary layer is output from the output layer via the intermediate layer. A neural network that outputs position traffic information to the traffic condition analysis unit, based on actual traffic information output from the moving vehicle detector at the temporary position and traffic information output from the output layer A traffic condition monitoring device, comprising: a learning unit that learns each weighting coefficient between the intermediate layer and each of the input layer and the output layer.