JP3734594B2 - Abnormal event detection device and traffic flow measurement device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
  The present invention relates to an abnormal event detection device that detects an abnormal event in traffic and a traffic flow measurement device that measures traffic parameters.
[0002]
[Prior art]
  In traffic flow monitoring using conventional abnormal event detection devices, sensors for traffic flow measurement are installed along the trajectory of traffic such as roads, and traffic volume, temporal or spatial occupancy, and average speed are measured from the measured values. , Traffic parameters such as traffic jams were calculated, and abnormal events that occurred between sensors were detected based on this. The traffic parameters as described above are called statistical parameters because they have statistical properties obtained by measuring a plurality of vehicles over a long period of time. On the other hand, traffic parameters such as speed obtained for each vehicle, measurement time, and occupation time for each vehicle are referred to as vehicle parameters.
[0003]
  If a vehicle failure, fire, fallen object, or any other abnormal event occurs at a point between the sensors, the traffic flow is obstructed and the traffic capacity at the point where the abnormality occurs decreases, so the upper limit of traffic volume on the downstream side decreases. To do. For this reason, an abnormal event can be indirectly detected by utilizing the following change in the statistical parameters obtained from one or a plurality of sensors.
  ・ Difference in traffic parameters between upstream and downstream sensors.
  ・ There is a difference in traffic parameters compared to normal operation with the same sensor..
  ・ Congestion is detected on the upstream side, but not on the downstream side..
  -Differences occur compared to parameters that are estimated and calculated sequentially from past measurement values, measurement times, upstream measurement values, etc..
[0004]
  For example, the abnormal traffic flow detection apparatus disclosed in Japanese Patent Laid-Open No. 3-209599 is configured as shown in FIG. That is, from the vehicle detection signal obtained by the vehicle detection means 1 such as an ultrasonic sensor, the traffic state quantity detection means 2 calculates a statistical parameter such as an average speed, and the traffic state quantity prediction means 3 or threshold setting means 4. An abnormal state of the traffic flow is detected by the traffic flow abnormality determining means 5 or the like, and the presence or absence of the traffic jam is detected by the traffic congestion determining means 7. Available traffic parameters include traffic density and traffic volume in addition to average speed. Furthermore, whether two or more devices shown in FIG. 8 are arranged upstream and downstream of the traffic flow and the abnormality judgment result is compared by comparing means 6 to determine whether the cause is a truly sudden abnormal event or due to natural congestion. Determine.
[0005]
  In addition to the devices in this example, statistical parameters such as using more sensors or using statistical analysis or neural networks for traffic flow abnormality judgment means are significant in statistical properties between parameters to be compared. Many methods have been proposed for detecting an abnormal event by using a difference. These are hereinafter referred to as statistical detection methods.
[0006]
  By the way, in order to level the fluctuations of individual measurement values to such an extent that they can be statistically ignored, the calculation of statistical parameters requires a certain time or a certain number of measurements. For example, conventionally, an ultrasonic sensor, a loop coil sensor, or the like was used for measurement for 5 minutes, and a 5-minute traffic volume, time occupation rate, average speed, or the like was used. Therefore, as the traffic volume increases, parameters can be obtained in a shorter time with higher accuracy, and abnormality detection accuracy is improved. Conversely, when the traffic volume is low, if the measurement time is extended to measure a certain number of cars, changes due to abnormal events are buried at the same time as the leveling, so the accuracy does not improve.
[0007]
  However, the above statistical detection method has two drawbacks. The first is that an abnormal event cannot be detected unless a sufficient time has elapsed after the occurrence of the abnormal event. This means that a predetermined time is required from the time when the first vehicle passes the sensor until a sufficient number of vehicles pass for parameter calculation, and detection in less time is impossible in principle. That's why.
  Second, depending on conditions such as the traffic volume and the point of occurrence of an abnormality, such an abnormal event cannot be detected in the case of a mild abnormal event that allows the vehicle to pass slowly through the abnormality occurrence point. For example, if the traffic volume flowing in from the upstream side is less than the traffic capacity decreased at the abnormality occurrence point, there is almost no difference in the traffic volume between the upstream side and the downstream side. In addition, if the location where an abnormality has occurred or the end of a traffic jam is not close to the sensor that detects the running state of the vehicle, the vehicle may be before decelerating or after accelerating again at the sensor measurement point. Often, there is no difference in speed distribution.
[0008]
  In view of this, for example, the traffic flow measuring apparatus disclosed in Japanese Patent Application Laid-Open No. Hei 8-279903 proposes another abnormality detecting means using vehicle parameters such as vehicle speed and measurement time.
  In normal times, the time that the vehicle passes between the sensors can be roughly estimated by the vehicle speed measured upstream. Based on such a prediction principle, when an abnormality occurs, it is detected that the passing time or the vehicle speed does not become as predicted in normal times due to sudden deceleration or stop of the vehicle passing through the abnormality occurrence point.
[0009]
  FIG. 9 is a configuration diagram thereof. The speed of the vehicle 20 measured by the sensor 21 and the measuring unit 22 is stored in the buffer 23 together with the measurement time as a vehicle parameter for a predetermined time or more. The prediction means 24 calculates and stores vehicle parameters such as time and vehicle speed estimated to be measured downstream from vehicle parameters in the upstream buffer. The determination means 26 verifies whether or not a vehicle parameter that matches the estimated and calculated vehicle parameter is measured from the downstream side, and outputs an abnormality degree that is an abnormality degree depending on the degree of coincidence.
  This makes it possible to detect even minor abnormal events that do not cause a difference in normal statistical parameters, and to detect abnormal events quickly even at the early stage of the event before the difference occurs. . Hereinafter, such a method is referred to as a microscopic detection method.
[0010]
  Next, the prior art of the traffic flow measuring device will be described. In recent years, a device that uses a TV camera as a sensor and processes an image obtained from the TV camera to measure a traffic situation on a road has been developed. FIG. 10 is a block diagram showing a configuration of a traffic flow measuring apparatus using conventional image processing described in Nobuaki Inoue et al .: “A Trial of Vehicle Dynamics Measurement” (17th Image Engineering Conference 1986, 16-13). It is. In the figure, 31 is a TV camera as an image pickup device, 32 is a TV camera control circuit for controlling the TV camera 31, 33 is a frame memory for capturing a digitized image pickup signal, and 34 is a window setting circuit for setting a vehicle extraction area. , 35 is a vehicle detection circuit, 36 is a vehicle travel position detection circuit, 37 is a vehicle movement amount detection circuit, and 38 is a vehicle information measurement circuit. FIG. 11 shows an installation example of a general TV camera. The TV camera is installed with a certain depression angle downward so as to capture a wide range.
[0011]
  The operation will be described. The TV camera 31 images the traveling vehicle at a certain frame interval according to the camera control circuit 32. The frame memory 33 takes in data obtained by digitally converting the imaging signal of the TV camera 31. The window setting circuit 34 sets a window that is the size of the vehicle on the road on the screen. The vehicle detection circuit 35 detects the vehicle by extracting features of the vehicle such as a horizontal line and a vertical line from the region set by the window setting circuit 34. The vehicle travel position detection circuit 36 detects the travel position on the screen of the vehicle detected by the vehicle detection circuit 35, and the vehicle movement amount detection circuit 37 detects the travel position and the travel position of the front frame of the vehicle. The amount of movement on the vehicle screen is detected and transmitted to the window setting circuit 34. The window setting circuit 34 moves the window according to the movement amount. The above operation is repeated to track the vehicle. When the vehicle disappears from the screen, or when a measurement area is specifically set in the image, the tracking ends when the vehicle exits from the area, and the vehicle information measurement circuit 38 determines the vehicle speed and the distance between the vehicle and the preceding vehicle. Measure distances.
[0012]
  In speed measurement that converts the amount of movement on the screen into speed on the road, it can not be solved mathematically if it is as it is, so usually the movement of the vehicle is restricted by the movement parallel to the road, for example In many cases, the vehicle speed is calculated assuming that the lower end of the extracted vehicle is on a measurement reference plane that is a plane parallel to the road. For example, when it is known that features such as bumpers and lights are easily extracted as the lower end of the vehicle, the above-mentioned calculation measurement reference plane is calculated as the height of the average vehicle bumper and lights.
  In addition, it can be obtained directly without recognizing individual vehicles from the proportion of the vehicle part extracted on the screen, such as spatial and temporal occupation rate, average speed field, traffic jam, etc. Many statistical parameters are possible.
[0013]
  Note that general sensors other than image processing are mainly sensors that obtain statistical parameters such as occupancy rate and traffic jam using a vehicle detector using ultrasonic waves, infrared rays, radio waves, and the like. In order to measure the speed of each individual vehicle, for example, a Doppler effect type radio wave sensor or the like can be measured alone. However, as other means, a vehicle detector 50 as shown in FIG. 12 (a) or (b). There is also a complex sensor configured to calculate the speed from the time difference between the sensing signals and the interval between measurement points.
[0014]
[Problems to be solved by the invention]
  In the above-described microscopic detection means for abnormal events, the vehicle can slowly pass through the abnormality occurrence point except when vehicle parameters that can identify a specific vehicle such as the land transportation number information of an automobile are measured. In some cases, an erroneous response that the estimated value of the vehicle parameter for the vehicle that has passed the upstream sensor coincides with the vehicle parameter of another vehicle that has passed the downstream sensor happens occasionally, and the vehicle does not slow down It seems to have reached the downstream sensor as usual. Usually, since an abnormality is detected also by a delay of a subsequent vehicle, occurrence of an erroneous response up to a certain rate does not matter. However, if there is a large amount of traffic flowing in from the upstream side, the proportion of false correspondence increases. If the ratio is close to 100%, that is, the state in which it is erroneously determined that most vehicles have reached the downstream side without delay, the difference in the number of units measured by both sensors will continue. Even if it is obvious, since it includes a measurement error, it is not reliable to determine that there is an abnormal event, and no abnormal event could be detected during that time. In addition, when there is traffic jam, the sensor side can measure that the traffic jam has occurred, but the measurement error of the vehicle parameter increases due to reasons such as the distance between the vehicles being too narrow. Met.
[0015]
  On the sensor side, for example, when the vehicle is large or has a complicated shape, the sensor has a characteristic portion of the vehicle to be measured and a portion similar in appearance to this, and one vehicle However, there is a problem that may be measured twice or more. Furthermore, in the case of a measurement algorithm in which it is assumed that the characteristic part of the vehicle is at a certain measurement reference plane height, a significant speed measurement error occurs when a vehicle with a high vehicle height is erroneously measured. There was a problem. Especially in places where the sensor height is restricted low, such as in tunnels, the error may be more than twice the true value. In abnormal event detection by the microscopic detection method, Since it greatly affects the calculation, it has been a cause of erroneously outputting an abnormality judgment during normal times. Further, when such a speed error frequently occurs due to an increase in the mixing ratio of large vehicles, the statistical parameter is also affected, so that an erroneous abnormality determination may be output even by the statistical detection method.
  In addition, especially when the angle of depression of the camera is small, the measurement time difference between the erroneously measured part and the original measurement target part becomes extremely large, and the problem is that the distance between the vehicles is measured as if they were separate vehicles. There was also.
[0016]
  The above problem exists in addition to the image processing sensor. Even in a complex sensor composed of two vehicle sensors, if the installation direction is set at an angle as shown in FIG. Similar vehicle speed errors and measurement time errors occur. Further, it goes without saying that even if the vehicle detector is a side-fired type installed in the horizontal direction from the road side, there is a similar problem as long as the installation direction forms an angle.
[0017]
  The present invention has been made to solve the above-described problems. Even if it is a mild abnormal event, the abnormal event detection device has a high detection accuracy regardless of the level of traffic congestion (volume of traffic). The purpose is to obtain. It is another object of the present invention to provide a traffic flow measuring device capable of removing erroneous measurement values including large errors in speed and measurement time, and to prevent erroneous abnormal event determination by an abnormal event detection device.
[0018]
[Means for Solving the Problems]
  The abnormal event detection apparatus according to the present invention includes a first traffic flow measurement device installed on the upstream side and a downstream side, a second traffic flow measurement based on vehicle parameters measured by the first traffic flow measurement device. A first anomaly detector for calculating a first anomaly by calculating vehicle parameters estimated to be measured by the device and comparing the calculated values with actual values measured by the second traffic flow measuring device; A second anomaly detector that calculates a second anomaly from a temporal or spatial distribution of statistical parameters by a flow measuring device, and the occurrence of an anomalous event is determined from the first anomaly and the second anomaly Equipped with a judgment unitThe determination unit selects the first abnormality degree when the congestion degree is low, selects the second abnormality degree when the congestion degree is high, and the first and second abnormality when the congestion degree is in between. Select the higher of the degrees, and judge by the selected degree of abnormalityIs.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
Embodiment 1 FIG.
  FIG. 1 is a block diagram showing the configuration of an abnormal event detection apparatus according to the present invention. In FIG. 1, reference numeral 61 denotes a vehicle parameter such as a vehicle speed and a measurement time from a measured value of a sensor (not shown), a congestion degree, and a predetermined time. A traffic flow measuring device that measures traffic parameters including traffic parameters (number of vehicles), average speed, occupancy, and other statistical parameters. A plurality of sensors are installed along a track at intervals of 100 to 200 m, Detects driving. It is assumed that the track is an automobile-only road or a tunnel, and there are no branches in the measurement target section.
[0020]
  62 is a first abnormality detection unit that inputs vehicle parameters such as speed and measurement time from two traffic flow measuring devices 61 adjacent to each other, and calculates a first abnormality degree as a traffic flow abnormality degree by a microscopic detection method. , 63 is a second anomaly detector that inputs statistical parameters from the traffic flow measuring device 61 and calculates a second anomaly as an anomaly of traffic flow by a statistical detection method, and 64 is a first and second anomaly detector. It is a determination unit that determines whether or not an abnormal event has occurred from the first and second abnormalities that are the outputs of the abnormality detection units 62 and 63.
  Here, the traffic flow measuring device 61 is provided with a multi-function sensor that can obtain all necessary traffic parameters, like a measuring device based on image processing. Instead, a plurality of types of single-function sensors are provided for each parameter. You may use the apparatus provided with.
[0021]
  In the first abnormality detection unit 62, whether or not a vehicle that matches the parameters of the vehicle that has passed the upstream sensor of the traffic flow in the normal traffic state has passed the downstream sensor by the microscopic detection method. Verify whether. Then, the first abnormality degree of the traffic flow is calculated and output according to the degree of the match.
[0022]
  That is, two adjacent traffic flow measuring devices 61 are connected to the first abnormality detection unit 62. Among them, vehicle parameters such as vehicle speed and measurement time measured by the first traffic flow measuring device 611 installed on the upstream side are stored for a predetermined time or more. Based on the vehicle parameters, vehicle parameters such as measurement time and vehicle speed estimated to be measured for the same vehicle by the second traffic flow measuring device 612 installed on the downstream side are calculated according to a certain calculation method. Then, it is verified whether or not the vehicle parameter that matches the estimated vehicle parameter is measured by the second traffic flow measuring device 612 installed on the downstream side, and the degree of abnormality that is the degree of abnormality is determined by the degree of matching. Output as the first degree of abnormality.
[0023]
  In addition to the first and second traffic flow measurement devices 611 and 612, the second abnormality detection unit 63 is also connected to some traffic flow measurement devices 61 on the upstream and downstream sides thereof.
  The second abnormality detection unit 63 uses a statistical detection method to obtain a temporal distribution of statistical parameters such as average speed and traffic volume obtained over a predetermined time, for example, 5 minutes, and a space between sensors. The second abnormality degree is calculated and output as the abnormality degree of the traffic flow from the general distribution. For example, when the traffic volume on the downstream side is smaller than that on the upstream side, the degree of abnormality is high because there is an abnormality between them.
[0024]
  In the determination unit, for example, the first and second abnormality levels output from both abnormality detection units 62 and 63 are compared, the larger value is selected, and whether or not an abnormal event has occurred based on the value of the abnormality level Is output.
  In the first abnormality detection unit 62, an unnatural state in which the first abnormality degree is temporarily reduced every time a mishandling occurs while the abnormal state continues (however, it may be practically acceptable). However, the second abnormality degree by the second abnormality detection unit 63 is output instead, and the determination output “abnormal” is continued.
[0025]
  As a determination method different from the above, a first threshold value and a second threshold value higher than this are set for the degree of congestion (or traffic volume) measured by the traffic flow measuring device 61. , Information on the degree of traffic congestion and traffic volume is input to the determination unit 64, and if the traffic congestion level is less than or equal to the first threshold, the first abnormality level is selected to determine whether or not there is an abnormal event. If the degree is greater than or equal to the second threshold value, it is determined based on the second degree of abnormality. If the degree of congestion is between the first threshold value and the second threshold value, the first degree of abnormality is determined. The second abnormality degree may be compared and the determination may be made based on the higher abnormality value.
[0026]
  As another determination method, the weighting of the first abnormality degree is increased in advance when the degree of congestion is low, and conversely, the weighting of the second abnormality degree is increased when the degree of congestion is high. It is possible to increase the certainty of the determination by comparing the magnitudes of the two abnormalities and selecting the larger one so that one of the abnormalities is easily selected depending on the case. . Moreover, you may make it calculate the abnormality degree used by determination by calculating from both abnormality degree.
  As yet another determination method, the first degree of abnormality may be preferentially selected and used. That is, when the degree of traffic congestion is below a certain threshold, the determination is made based on the first degree of abnormality. This is because in such a case, the first degree of abnormality immediately increases. When the degree of traffic congestion is higher than the threshold value, when the first abnormality degree becomes large, the abnormality degree is used. If the first degree of abnormality does not increase, an abnormal event is determined based on the magnitude of the second degree of abnormality.
  Instead of the threshold value (including the first and second threshold values), a fuzzy value may be used instead of a fixed value.
[0027]
  Further, in the next section, 612 and 613 are given the roles of the first and second traffic flow measuring devices, respectively. The 1st, 2nd abnormality detection parts 62 and 63 and the determination part 64 are also provided separately corresponding to the area.
  In this way, measurement and determination similar to the above are performed in all sections.
  As described above, the two types of anomaly detection units are capable of handling a wider range of traffic than the conventional ones, and when the first anomaly detection unit 62 is activated, it takes a short time, and many times otherwise. In this case, there is an effect that an abnormal event can be detected in an appropriate time, that is, in a time required for accumulating and calculating statistical data and operating the second abnormality detecting unit 63.
[0028]
Embodiment 2. FIG.
  FIG. 2 is a block diagram showing an example of a traffic flow measurement device suitable for use in the abnormal event detection device as shown in the first embodiment, where a part of the vehicle is separated and erroneously measured. Is a configuration example of a traffic flow measuring device having an erroneous measurement value removing means that can determine and remove it. In the figure, reference numeral 71 denotes a sensor for detecting the traveling of the vehicle. As shown in FIG. 3, a sensor such as a camera is installed so as to face the vehicle toward the upstream side of the traffic flow. Reference numeral 72 denotes a measurement unit that outputs a vehicle speed and a measurement time as vehicle parameters based on data obtained from the sensor 71.
[0029]
  73 is a filter for removing vehicle parameters determined to be erroneously measured by the determination means, 74 is a buffer for storing past vehicle parameters determined to be true through the filter 73, and 75 is stored one time before. The distance conversion means for calculating the distance between the vehicle parameters and the last measured vehicle parameter, 76 is a determination means for determining whether or not there is an erroneous measurement, and is determined in advance with reference to the average vehicle length etc. When the inter-vehicle distance is shorter than the threshold value, it is determined that the last measured vehicle parameter is an erroneous measurement. The determination means 76, the filter 73, and the distance conversion means 75 constitute an erroneous measurement value removal means 77.
[0030]
  Next, the operation of the distance conversion means 75 will be described with reference to FIG. The measurement reference plane 80 is a plane parallel to the road surface, and the calculation is performed on the assumption that the characteristic part of the vehicle (for example, a bumper or a light) extracted from the image for calculation is at the height. In addition, a point 81 where measurement for one vehicle is completed and a measurement time is determined, that is, a vehicle position at the time of measurement is defined as a measurement point. For example, in an image processing sensor, a point corresponding to the downstream end of the measurement region is usually equivalent to a measurement point of a downstream sensor in a composite sensor as shown in FIG. The straight line 83 is a set of points on a space that cannot be distinguished from the measurement point from the sensor 71, and this is called a measurement line. 84 is a point directly below the sensor as the installation point of the sensor 71. More precisely, it corresponds to the focal point of the TV camera, or directly below the intersection of two straight lines passing through the vehicle detector and the measurement point in FIG. It is a point to do. The distance between the two points 81 and 84 is L. Furthermore, the measurement target portion when the measurement is normal is the vehicle lower portion 86, and the roof portion 87 is a portion in which the image shape is similar to that of the vehicle lower portion 86 and is erroneously measured.
[0031]
  Taking speed measurement by image processing as an example, if the difference in height between the sensor 71 and the roof portion 87 is g, and the difference in height between the sensor 71 and the measurement reference plane 80 is h, the roof portion 87 on the image is displayed. Since the apparent speed is h / g times, there is a problem that the vehicle speed is also calculated to be h / g times the true value. This is also the case with a complex sensor as shown in FIG. 12B, and the speed is h / g times because the difference in sensing time by the vehicle detector is g / h times lower than the lower part of the vehicle.
[0032]
  Further, since the roof portion 87 is measured not at the position 82 traveled by the distance ΔL from the lower portion of the vehicle after the vehicle lower portion is correctly measured at the position 81 at the time T1, the roof portion 87 is measured at the position 85 on the measurement line. The measurement time T2 is greatly deviated from the passage time T3 that actually passed the measurement point 81. For this reason, even if the time difference between the two measurement times is simply multiplied by the speed, the distance ΔL cannot be calculated accurately, and in some cases, the value is apparently larger than the average vehicle length. It was not possible to determine whether or not the measurement was incorrect by thresholding the time difference or the distance. In other words, if the distance between the two measurement parts is smaller than the average vehicle length, it can be determined that different parts of one vehicle have been measured twice, and it can be determined that there is an erroneous measurement. In some cases, the value may be larger than the average vehicle length, so such a determination cannot always be made.
[0033]
  FIG. 4 is a diagram illustrating the movement 91 of the vehicle lower portion 86 measured normally, the apparent movement 92 and the true movement 93 of the roof portion 87 measured in error, and the horizontal axis represents time such as measurement time and passage time. And the vertical axis represents distance. Therefore, tilt is speed. Theoretically, the apparent position and the true position of the roof portion 87 must always coincide with each other at a point immediately below the sensor as shown in the figure. Therefore, each of the vehicle lower portion 86 and the roof portion 87 is drawn or calculated as indicated by the solid line in the drawing from the measurement speed, time, and distance L at the measurement point, and each passes under the sensor 84. When the expected times 94 and 95 are obtained, the time is almost true regardless of whether or not they are erroneous measurements. If this is used, first, the passage time difference Δt between the vehicle lower portion 86 and the roof portion 87 at the position directly below the sensor 84 is calculated, and then Δt is calculated by multiplying the passage time difference Δt by the speed measured earlier. It can be converted into a distance (or distance between vehicles) ΔL. This is because, from the geometric relationship between the measurement line 83 and the traveling direction of the vehicle, the portion where the height is low and the speed is close to the true value is always measured first. In this way, it is determined whether the vehicle parameter measured later is not an erroneous measurement.
  The true movement 93 of the roof portion 87 is obtained as a straight line passing through a point that is parallel to the movement 91 of the vehicle lower part 86 in the drawing and that indicates the passage of the point 84 directly below the sensor at time 95.
[0034]
  Even if there is a large speed error in the erroneously measured vehicle parameters by the method as described above, or even if there is a large difference between the time when the vehicle actually passed the measurement point and the measurement time, an accurate inter-vehicle distance can be obtained, It is possible to make an erroneous measurement determination by inspecting whether ΔL is not unduly short by threshold processing, that is, comparison with an average vehicle length or the like.
  Finally, a sensor whose measurement point does not change every time a vehicle is measured is found especially in a part of a two-dimensional measurement sensor such as an image processing sensor. When such a sensor is used, If the fluctuation range seems to be large, the information on the measurement points is updated sequentially and reflected in the calculation.
[0035]
Embodiment 3 FIG.
  FIG. 5 is a block diagram showing a configuration example of a traffic flow measuring device used when a sensor 71 such as a camera is installed toward the downstream side of the traffic flow as shown in FIG. Reference numeral 78 denotes an erasing unit for erasing the vehicle parameter determined to be erroneous measurement by the determining unit 76, and the other units have the same functions as those in the second embodiment, so that the description thereof is omitted. . Here, the determination means 76, the erasure means 78, and the distance conversion means 75 constitute an erroneous measurement value removal means 77. The distance conversion means 75 calculates the inter-vehicle distance from the last measured vehicle parameter to a plurality of vehicle parameters previously stored in the buffer 74. The erasing unit 78 removes the parameter that is erroneously measured based on the determination result from the buffer 74 or sets a flag indicating erroneous measurement.
[0036]
  FIG. 7 is a diagram illustrating the movement of each part of the vehicle. A method similar to that described in the second embodiment is used, but contrary to the case of FIG. 4, since it is known that the true value is measured later, it passes the velocity measured later. By multiplying the time difference Δt, Δt is converted to ΔL.
[0037]
  As the traffic flow measurement device of the abnormal event detection device described in the first embodiment, the traffic flow measurement device shown in the second embodiment or the third embodiment can be used. The first and second abnormality detection units 62 and 63 in FIG. 1 take out the vehicle parameters from which the erroneous measurement values have been removed from the buffer 74 and use them or calculate statistical parameters.
[0038]
【The invention's effect】
  Vehicle parameters estimated to be measured by the second traffic flow measuring device from the first and second traffic flow measuring devices installed on the upstream side and the downstream side, and vehicle parameters measured by the first traffic flow measuring device. A first anomaly detection unit that calculates the first anomaly by calculating and comparing the calculated value with an actual measurement value of the second traffic flow measuring device; Alternatively, a second abnormality detection unit that calculates the second degree of abnormality from the spatial distribution and a determination unit that determines occurrence of an abnormal event from the first degree of abnormality and the second degree of abnormality are provided.The determination unit selects the first abnormality degree when the congestion degree is low, selects the second abnormality degree when the congestion degree is high, and the first and second abnormality when the congestion degree is in between. Select the higher of the degrees, and judge by the selected degree of abnormalityIs.
  As described above, according to the abnormal event detection device of the present invention,Vehicle parameters estimated to be measured by the second traffic flow measuring device from the first and second traffic flow measuring devices installed on the upstream side and the downstream side, and vehicle parameters measured by the first traffic flow measuring device. A first anomaly detection unit that calculates the first anomaly by calculating and comparing the calculated value with an actual measurement value of the second traffic flow measuring device; Alternatively, the second abnormality detection unit that calculates the second abnormality degree from the spatial distribution, and a determination unit that determines the occurrence of the abnormal event from the first abnormality degree and the second abnormality degree, When the degree of traffic congestion is low, the first abnormality degree is selected, when the degree of congestion is high, the second abnormality degree is selected, and when the traffic congestion degree is in between, the first and second abnormality degrees are high. Select the method, and judge by the selected degree of abnormalityTherefore, an appropriate abnormal event can be determined in accordance with the degree of traffic congestion, and the detection accuracy is improved.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an abnormal event detection apparatus according to Embodiment 1 of the present invention.
FIG. 2 is a block diagram showing a configuration of a traffic flow measuring apparatus according to Embodiment 2 of the present invention.
FIG. 3 is an explanatory diagram for explaining the operation of a distance conversion means in Embodiment 2 of the present invention.
FIG. 4 is a diagram for explaining the operation of distance conversion means in Embodiment 2 of the present invention.
FIG. 5 is a block diagram showing a configuration of a traffic flow measuring apparatus according to Embodiment 3 of the present invention.
FIG. 6 is an explanatory diagram for explaining the operation of a distance conversion means in Embodiment 3 of the present invention.
FIG. 7 is a diagram for explaining the operation of a distance conversion means in Embodiment 3 of the present invention.
FIG. 8 is a block diagram showing a configuration of a conventional abnormal event detection apparatus.
FIG. 9 is a block diagram showing a configuration of another conventional abnormal event detection apparatus.
FIG. 10 is a block diagram showing a configuration of a conventional traffic flow measuring device.
FIG. 11 is an explanatory diagram showing an installation example of a TV camera.
FIG. 12 is an explanatory diagram showing an installation example of a vehicle detector.
[Explanation of symbols]
  61 traffic flow measuring device, 62 1st abnormality detection part, 63 2nd abnormality detection part,
64 determination units, 71 sensors, 72 measurement units, 73 filters, 75 distance calculation means,
76 determining means, 77 erroneous measurement value removing means, 78 erasing means,
611 1st traffic flow measuring device, 612 2nd traffic flow measuring device.

Claims (1)

A plurality of sensors for detecting the traveling of the vehicle are installed at a distance along the track, and the vehicle parameters including the vehicle speed and the measurement time of the vehicle, and the vehicle that has passed within a certain time from the measured values of the sensors. Installed upstream of traffic flow in an abnormal event detection device that detects traffic abnormal events by a traffic flow measurement device that measures traffic parameters consisting of statistical parameters including any of the average speed, number of vehicles, and degree of congestion The first traffic flow measuring device, the second traffic flow measuring device installed on the downstream side of the traffic flow, and the second traffic flow measuring device based on the vehicle parameters measured by the first traffic flow measuring device. The vehicle parameter estimated to be measured in step 1 is calculated, and the calculated value is compared with the actually measured value measured by the second traffic flow measuring device. An anomaly detector, a second anomaly detector that calculates a second anomaly of traffic conditions from a temporal or spatial distribution of statistical parameters by one or more of the traffic flow measuring devices, and the first anomaly A determination unit that determines the occurrence of an abnormal event from the second degree of abnormality and the second abnormality degree, and measures the degree of congestion by the first or second traffic flow measuring device, and the determination unit has a low degree of congestion The first abnormality degree is selected, the second abnormality degree is selected when the congestion degree is high, and the first abnormality degree and the second abnormality are selected when the congestion degree is between the high and low levels. An abnormal event detection apparatus characterized by comparing the degree of abnormality of each other, selecting a higher degree of abnormality, and determining occurrence of an abnormal event based on the value of the selected degree of abnormality .

Images