JPH0652485A - Dangerous traffic event detecting method - Google Patents

Abstract

PURPOSE:To intensively record the movement of vehicles by detecting one dangerous traffic event corresponding to one mixed area consisting of consecutive mix measurement points in the case there is a mix measurement point and discriminating the dangerous degree of the dangerous event to be assumed in the traffic flow. CONSTITUTION:An image pickup means obtains a video signal corresponding to a picture in an area including a road. With the use of vehicle binary data group binarized as a vehicle correspondence point by picture processing, quaternary data group (Fn) consisting of the difference of two vehicle binary data groups obtained by the constant time difference is obtained. Based on the Fn, the speed and direction of the vehicle correspondence point are obtained to calculate a vehicle movement vector (Mj) showing the predicted position in the lapse of the prescribed time of each vehicle correspondence point. Then, a mix measurement point (Nij) mixing mix inhibiting vector (M'j) such as stop lines including each vehicle movement vector (Mj) is calculated. Thus, the dangerous traffic event (Lm) can be detected.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for detecting a dangerous event in a traffic flow, and more specifically, it processes a video signal obtained by imaging a road or the like to obtain a brightness level corresponding to a specified position on the road surface. By using the image type vehicle detection device that detects vehicles existing on the road etc. from changes caused by vehicles, dangerous events of traffic flow such as contact between vehicles or possibility of contact with pedestrians are detected and And a device therefor.

[0002]

2. Description of the Related Art Needless to say, prevention of danger associated with road traffic is an important social issue. Taking Japan as an example, the number of road traffic fatalities is as high as 15,000 per year, which is becoming more serious. There is an opportunity for related public organizations to jointly collect data on traffic accidents, thoroughly elucidate the causes of traffic accidents that have occurred, and work to resolve the problems. For this purpose, it is necessary to collect more data on traffic accidents that unfortunately occurred and to obtain more detailed data themselves. However, in the past, an appropriate method for this has not been established,
Therefore, in the present situation, there is a shortage of data or the content of data, and there has been a demand for a better data collection means for preventing the danger associated with road traffic.

[0003]

SUMMARY OF THE INVENTION The present invention has been made in view of the above situation and is common to the image type vehicle detection device which has been remarkably developed in recent years and has already been installed in many places such as intersections. The purpose of the present invention is to propose a method for detecting a dangerous traffic event, which newly detects a dangerous event expected in a traffic flow and further judges the degree of danger of this event, by making extensive use of the technology described above. By using this method, it is possible to intensively record not only the actual accident but also the trend of the vehicle under the situation where the probability of accident occurrence is high. Therefore, it becomes possible to collect the traffic accident data more surely and more finely than before. Furthermore, by using this, the correlation between the degree of danger and the occurrence of an accident can be analyzed, which can be useful for safety measures and the like. The present application also proposes a method using a plurality of imaging means for handling a wider area and detecting a dangerous traffic event.

[0004]

In order to solve the above-mentioned problems, in the dangerous traffic event detecting method of the first invention of the present application, an area including a road is imaged by an imaging means to obtain a corresponding video signal,
Luminance data group (D) for each field, which consists of the luminance level (Cij) at each temporal position in the video signal corresponding to each of the plurality of measurement points (Pij) set on the road in the image
t) are sequentially obtained, and a luminance data group consisting of luminance levels (Cr ij) corresponding to the road surface at each measurement point (Pij) is extracted at a predetermined interval and stored as reference road surface data (Dr). (Dt) is compared with the reference road surface data (Dr), and data that has a sufficient difference is binarized as vehicle corresponding points to obtain a binarized vehicle data group (Dn), which are sequentially held and a fixed time difference is obtained. The four-valued data group (Fn) consisting of the difference between the two vehicle-two-valued data groups (Dn) obtained in step (4) is obtained, and the speed and direction of the vehicle corresponding point are calculated based on the four-valued data group (Fn). Then, a vehicle movement vector (Mj) indicating the predicted position of each vehicle corresponding point after a predetermined time has elapsed is calculated,
An intersection measurement point (Nij) at which intersection prohibition vectors (M'j) such as a stop line including each vehicle movement vector (Mj) intersect is calculated, and if there is an intersection measurement point (Nij), the intersection measurement points (Nij) are continuously calculated. The intersection area (L ') consisting of the intersection measurement points (Nij)
m) A danger detection signal (Sm) indicating that one dangerous traffic event (Lm) is detected corresponding to one m) is output.

Further, in the dangerous traffic event detection method of the second invention of the present application, in addition to the steps of the first invention described above, when the dangerous traffic event (Lm) is detected, the corresponding intersection area (L ') is detected. It also includes a process of obtaining and outputting risk data (Gm) quantifying the risk level of the dangerous traffic event (Lm) based on the distribution status of each intersecting measurement point (Nij) of m).

A method for detecting a dangerous traffic event according to the third invention of the present application is
An image including a road is picked up by a plurality of image pickup means to obtain corresponding video signals, and each video signal is set in association with an actual position coordinate (Xi, Yj) on the road in the image. Further, the coordinate brightness data group (Dtp) consisting of the brightness level (Cp ij) at each temporal position in the video signal corresponding to each of the plurality of measurement points (Pp ij) and the position coordinate information is sequentially obtained. For each video signal, each measurement point (Pp
ij), a luminance data group consisting of luminance levels (Crpij) corresponding to the road surface at (ij) is extracted at a predetermined interval and held as coordinate reference road surface data (Drp), and a coordinate luminance data group (Dtp) is stored for each video signal. The coordinate vehicle binarized data group (Dnp) obtained by binarizing the data for which a sufficient difference is recognized as compared with the coordinate reference road surface data (Drp) as vehicle corresponding points is obtained and sequentially held, and for each region , A coordinate quaternary data group (Fnp) which is a difference between the two coordinate vehicle binary data groups (Dnp) obtained at a fixed time difference, and the coordinate quaternary data group (Fnp) is obtained for each area. ), The velocity and direction of the vehicle corresponding point are obtained, and the coordinate vehicle movement vector (Mjp) indicating the predicted position on the road (Xi, Yj) of the vehicle corresponding point after a predetermined time elapses is calculated. Calculate and calculate each vehicle movement vector (Mjp) for each area. Including a stop line, etc. The intersection prohibition vector (M'jp) intersects with each other.
ij) and, if there is a coordinate intersection measurement point (Np ij), one dangerous traffic event corresponding to one coordinate intersection area (L'mp) consisting of consecutive coordinate intersection measurement points (Np ij) (L
The temporary coordinate danger prediction signal (S'mp) indicating that mp) has been detected is associated with the actual position coordinates (Xi, Yj) on the corresponding road, and the temporary coordinate danger obtained for each of all areas. If any of the predicted signals (S'mp) whose position coordinates (Xi, Yj) substantially match, they are combined into one, and then all the temporary coordinate danger predicted signals (S'mp) are converted to the danger predicted signal. Output as (Smp).

Further, in the dangerous traffic event detection method of the fourth invention of the present application, in addition to the steps of the third invention described above, when the dangerous traffic event (Lmp) is detected, the corresponding intersection area (L ') is detected. mp), the risk level data (Gmp) quantifying the risk level of the dangerous traffic event (Lmp) is obtained based on the distribution of the coordinate intersection measurement points (Np ij) and is output.

[0008]

In order to solve the above-mentioned problems, in the dangerous traffic event detecting method according to the first aspect of the present invention, first, a video signal of an image of a road area is obtained by the image pickup means, and temporally measured at each measurement point (Pij) on the road. By obtaining the luminance level (Cij) of the corresponding video signal, that is, the luminance data group (Dt) in a unit of one field, and comparing this with the appropriate reference road surface level data (Dr) corresponding to only the road surface. Vehicle binary data group (Dn
) Are held in order, and processing similar to the conventional one is performed.

A 4-valued data group (Fn) is obtained by calculating the difference between the two vehicle-valued data groups (Dn) with a constant time difference thus obtained, and detection is performed based on the distribution of the 4-valued data group (Fn). The speed and direction of the vehicle corresponding point expected at that time are obtained, and the position of each vehicle corresponding point predicted after a predetermined time elapses (predicted vehicle corresponding point, that is, vehicle movement vector: Mj
Then, the cross prohibition vector (included in a kind of cross prohibition vector (M'j)) is calculated, and the cross prohibition vector (preventing crossing prohibition vector) between these predicted positions (cross prohibition vector) or other stop lines or pedestrian crossings ( Measurement point (Nij) intersecting with M'j)
To calculate. If there is such an intersection, it is considered as a dangerous traffic event that is expected to be dangerous, and there is one continuous intersection measurement point group, that is, one intersection area (L'm). It is assumed that one dangerous traffic event (Lm) is detected, and a dangerous detection signal (Sm) to that effect is output. Each process is repeated.

In the dangerous traffic event detection method of the second invention of the present application, in addition to the same processing as described above, when a dangerous traffic event (Lm) is detected, each intersection measurement point of the corresponding intersection area (L'm) is detected. Distribution of (Nij), for example, intersection point group (Lm)
The number of consecutive intersections that make up the position, the position in the intersection prohibition vector (Kj) of individual intersections (such as near the periphery or far from the periphery to the center), or the distribution status with the quaternary value added Quantify the risk of dangerous traffic event (Lm) based on Then, this result (value) or grades obtained by dividing this value into ranges is output as the risk data (Gm).

In the dangerous traffic event detecting method of the third invention of the present application, video signals corresponding to images of a plurality of regions are obtained by using a plurality of image pickup means. Each video signal is subjected to substantially the same processing as in the above-mentioned first invention, but each measurement point (Pp ij) is set in association with the coordinate value indicating the actual position coordinate (Xi, Yi) on the road. . Then, for each video signal, substantially the same processing as in the above-described first invention is performed, and a dangerous traffic event (Lmp) is detected. After that, the temporary coordinate danger prediction signal (S'mp) obtained for each video signal is based on the actual position coordinates (Xi, Yi) on the road to which each corresponding coordinate intersection measurement point (Np ij) corresponds. If there are multiple temporary coordinate danger prediction signals (S'mp) that have been verified and are associated with substantially the same coordinates, these are considered to correspond to dangerous traffic events that occurred at substantially the same point, and are combined into one. After that, all the temporary coordinate danger prediction signals (S'mp) are finally output (coordinate danger prediction signal (Smp)).

The method for detecting a dangerous traffic event according to the fourth aspect of the present invention is
Each process including the position coordinates (Xi, Yi) is performed in the same manner as the third invention described above, and each coordinate intersection is performed on the dangerous traffic event (Lmp) detected and determined based on the same idea as the second invention described above. The risk level of dangerous traffic events (Lmp) is quantified based on the distribution status of the measurement points (Np ij), and the risk grade obtained by classifying the result (value) or the range is used as the risk level data (G).
output as mp).

In each of the above inventions, 2
A four-valued data group is obtained from the difference between the two coordinate binary data groups, a vehicle movement vector is calculated based on the four-valued data group, and a dangerous traffic event is detected in response to detection of an intersection with the intersection prohibition vector. The technical idea of "is common" and is a characteristic part of the present application.

[0014]

The present invention will be described below with reference to the drawings based on the embodiments. First, an image processing apparatus suitable for carrying out the method of each invention of the present application will be described. Conventionally, image processing has been applied to grasp traffic conditions, and a method and apparatus for detecting a vehicle passing through the road by processing a video signal obtained by capturing an image of the road by an image capturing means have been put to practical use. There is. Such an image processing apparatus is disclosed in, for example, Japanese Patent Application Laid-Open No. 4-188005 (Japanese Patent Application No. 2-188005) by the present inventors.
No. 318992): Proposed as "imaging vehicle detector".

To carry out the method of detecting a dangerous traffic event according to the present invention, an apparatus having a similar structure to the image type vehicle detecting apparatus which has been conventionally used can be used. Then, under the control of the control unit, first, through the image processing similar to the above-described conventional invention, the luminance data group corresponding to the measurement point group is sequentially obtained at predetermined time intervals. However, unlike the case where only vehicle detection is performed, as the measurement point group in the present application, a large number of planar measurement point groups corresponding to the entire surveillance road surface are set. Figure 1
Shows a simplified representation of this situation.

By the way, a specific setting example suitable for practical use of each measuring point in the present application is 20 m × 20 m.
In this area, grid-like measurement points of 50 × 50 = 2500 points are set and monitored. This is a detection resolution equivalent to about 40 cm × 40 cm, and can sufficiently detect the target vehicle. On the basis of the brightness data group obtained corresponding to the planar measurement point group, the dangerous traffic event is extracted or the risk degree is further calculated by the predetermined process described later.

Further, a device for carrying out the present invention will be briefly described with reference to an example. FIG. 11 shows a block diagram of an image type vehicle detection device of this type, which is similar to the device disclosed in Japanese Patent Laid-Open No. 4-188005. This image-type vehicle detection device is generally, image pickup means 1
0, a brightness data conversion unit 20, an arithmetic processing unit 30, and a control unit 40.

The image pickup means 10 is, for example, a CCD camera, is provided at one corner of the intersection as shown in FIG. 12, and picks up an image of the detection area of the intersection as shown in FIG. The video signal of is sent to the luminance data converter 20. Using this video signal, the brightness data conversion unit 20 causes a plurality of straight lines (detection lines: L ₁ and L ₁₎ which are assumed in the horizontal direction in the image as shown in FIG.
₂ , ...) on the plurality of measurement points Pij (P ₁₁ , P ₁₂ , ...,
P _21, P _22, ...) is set and the luminance data group at the certain time of the set luminance level respectively determined in temporally corresponding luminance signals to the position of each measurement point Pij for each field Request one by one.

That is, the brightness data conversion unit 20 corresponds to the sync signal extraction circuit 21 for obtaining the horizontal sync signal and the vertical sync signal from the video signal, and digitizing the current search position on the search line based on these sync signals. A horizontal address counter 22 and a vertical address counter 23 for obtaining the horizontal coordinate and the vertical coordinate, and outputs of these and the microprocessor 4
A digital comparator 24 that compares the coordinates designated by 1 (described later) and generates a capture signal at the moment when they coincide with each other, and obtains and outputs the luminance data by digitizing the luminance at the corresponding moment of the video signal according to the capture signal. The video A / D conversion circuit 25 is configured to operate. A video amplifier 27 for amplifying the transmitted video signal to an appropriate level is also provided in the input section, and a clamp level fixing circuit 28 for normalizing the video signal to a constant amplitude prior to A / D conversion is also provided. It is provided.

The arithmetic processing unit 30 is a circuit for receiving the luminance data from the video A / D conversion circuit 25 and averaging a plurality of luminance data and calculating the correlation, and is required to perform high-speed calculation. Therefore, the signal processor 31 is used and a storage unit necessary for storing data is also provided.

The control unit 40 controls each of the above-mentioned units, receives data from the arithmetic processing unit 30 and further performs arithmetic processing on the data to detect the presence of a vehicle and calculate the speed of the vehicle. In each of the inventions of the present application, detection of a dangerous traffic event and quantification of the degree of danger are also performed in the part of outputting to the device. The control unit 40 includes a microprocessor 41, a ROM 42 storing a control program, and an RA for data storage.
It comprises an M43 and an output I / O circuit 44. Reference numeral 45 denotes a CPU bus through which the microprocessor 41 exchanges data and addresses, control commands and response signals with other parts.

In addition to the illustrated apparatus, a DC power source 5 is also provided.
1. Clock circuit 52 supplied to each unit, image pickup means 10
D / A for sending out a signal for increasing the sensitivity of the image pickup means 10 when the level of the video signal from the camera is low and inappropriate for processing.
The conversion circuit 53 and the like are provided. Also, monitor TV6
0 and the personal computer 61 are auxiliary equipments required when setting the respective measurement points so that the image device fits in the actual installation site. The description is omitted because it is described in detail in the specification of the above-mentioned application. Various information such as the detection of a dangerous traffic event, its risk level data, and the position of occurrence is sent to a recording device, a totaling device, or the like (not shown) to start image recording, data recording, or analysis.

Next, the dangerous traffic event detecting process of each invention of the present application will be described. Note that FIG. 10 is a simplified flowchart showing an embodiment of each invention of the present application. In the dangerous traffic event detection method of the first invention of the present application, for example, by using the apparatus shown in FIG. 11 described above, a video signal is first generated by each of the above-mentioned parts controlled by the control unit 40 in substantially the same manner as general image processing. A plurality of measurement points Pij (P ₁₁ , P ₁₁ ,
_{P 12, ..., P 21,} P 22, ...) the brightness level (Cij) in temporally corresponding luminance signals respectively obtained by obtaining sequentially set the luminance data group a (Dt) for each field. Then, when the vehicle does not exist (corresponding to only the road surface), the data corresponding to the vehicle corresponding point is obtained by comparing with the brightness data group (reference road surface level data, Dr) including the brightness data of each measurement point. The presence of a vehicle at each measurement point at a certain time is detected.

That is, as shown in FIG. 1, the image pickup means 10 (CCD camera) picks up an image of a detection area such as a predetermined intersection and sends a corresponding NTSC video signal to the luminance data converter 20. It should be noted that a plurality of measurement lines such as L ₁ , L ₂ , ..., L ₄₀ are set corresponding to this detection area as shown in FIG. 1, and a plurality of measurement lines are set on the respective measurement lines L _{1 to} L ₄₀ at predetermined intervals. Individual measurement points (P ₁₁ , P ₁₂ , ..., P
₂₁ , P ₂₂ , ...) are set (assumed).

In the brightness data converter 20, the sync signal extraction circuit 21 obtains a horizontal sync signal and a vertical sync signal from the video signal. The horizontal address counter 22 and the vertical address counter 23 operate based on these synchronization signals,
The current position of the investigation on the investigation line is shown one after another. These outputs are compared with the coordinates designated by the microprocessor 41, and the digital comparator 24 generates a capture signal at the moment when they coincide with each other. Video A according to this capture signal
The / D conversion circuit 25 digitizes the luminance at the corresponding moment of the video signal and outputs it as luminance data together with the corresponding horizontal and vertical coordinates. In the input section, the transmitted video signal is amplified by the video amplifier 27 to an appropriate level, and the clamp level fixing circuit 28 following this amplifies the video signal to a constant amplitude prior to A / D conversion.

In this way, a plurality of measurement points Pij (P ₁₁ , P ₁₁₎ set in substantially the entire area of the image of FIG.
_12, ..., the luminance level Cij at P _21, P _22, ...) temporal position in the luminance signal corresponding to each of the positions of the resulting.
From these luminance levels, a vehicle binarized data group Dn in which one screen (one frame) is binarized is obtained. That is,
When there is no vehicle that is constantly updated at an appropriate interval (corresponding to the road surface only), a sufficient difference is recognized by comparison with the brightness data group (reference road surface data, Dr) consisting of the brightness data at each measurement point. The presence of the vehicle at each measurement point at a certain time is detected by using the obtained data as the vehicle corresponding point.

For this purpose, the arithmetic processing unit 30 equipped with the signal processor 31 receives the luminance data from the video A / D conversion circuit 25 and rapidly averages a plurality of luminance data and calculates the correlation. After that, the brightness data for each measurement point is binarized by comparing with the reference road surface data Dr which is composed of the brightness level of the road surface extracted at appropriate time intervals.
A set of binarized data group Dn corresponding to the field is determined.

That is, the method of each invention of the present application first includes the following series of steps (see FIG. 10). 1. An image signal such as a CCD camera is used to obtain a video signal corresponding to an image of a region including a road to be sensed. 2. Measurement point Pij (P ₁₁ , P ₁₁ ,
_{P 12, ..., P 21,} P 22, ...) to advance multiple sets the corresponding position (time position in the video signal), husband corresponding to these positions i.e. the respective measurement points Pij s corresponding luminance level ( C
The luminance data group (Dt) consisting of ij) is sequentially obtained. 3. The luminance data group corresponding to the case where there is no vehicle only on the road surface is extracted at a predetermined interval to obtain the reference road surface level data (D
hold as r). This reference road surface level data (Dr
) Will be appropriately corrected over time depending on the situation such as road illuminance. 4. The luminance data group (Dt) is used as the reference road surface level (Dr).
) And binarized vehicle data group (Dn) are sequentially obtained and held. That is, when a sufficient difference is recognized (both high and low), it is regarded as the vehicle corresponding point where the vehicle exists, and the value “1” is associated (or, if necessary, appropriately processed. To detect the vehicle and send out a sensing signal). It should be noted that each of the above processes is a technique similar to the conventional image type vehicle detection method, and for example, the above-mentioned Japanese Patent Application No. 2-3189.
No. 92 also describes in detail the vehicle detection principle based on the pattern correlation comparison between the road brightness pattern and the vehicle brightness pattern.

Control unit 4 equipped with microprocessor 41
0 controls each of the above-mentioned units, and the arithmetic processing unit 30
It receives the data from and performs further arithmetic processing on it. That is, in the present invention, the vehicle binary data D thus obtained
Based on n, the control unit 4 executes a predetermined process described in detail below.
Dangerous traffic event (Lm)
And the risk level of this dangerous traffic event is determined. Next, the process for this will be described.

The microprocessor (host CPU) of the control unit 40 has the vehicle binarized data (Dn
) Is transmitted, and vehicle binary data corresponding to each address is sequentially stored in the RAM 43. The vehicle binarization data (D ₁ ) for the previous frame (T ₁ frame) corresponding to the screen of FIG. 1 is shown in FIG. 2, and the vehicle binarization data (D ₂ ) for the current frame (T ₂ frame) is shown. ) Is shown in FIG.
In the figure, “0” means “without vehicle”, and “1” means “with vehicle”.

Subsequently, a quaternarization process is performed. That is, first, based on the image binarized data of the previous frame of FIG. 2 and the image binarized data of the current frame of FIG. 3, a differential four-valued data group of (previous frame−current frame) is obtained and the same. It is stored in the RAM 43. For example, both 2 above
From the digitized data, a four-valued data group (Fn) derived by a process of (current frame vehicle binarized data)-(previous frame vehicle binarized data) is obtained. In FIG. 4, as an example, “0” − “0” = “F”, “0”
The results of quaternarization are illustrated by setting “−1” = “− 1”, “1” − “1” = “+ 2”, and “1” − “0” = “+ 1”.

Therefore, in FIG. 4, "F"("0"-
“0”) is the measurement point where the vehicle did not exist both last time and this time. “-1” (“0”-“1”) was the measurement point where the vehicle existed in the previous frame but left in this frame. "+2"("1"-"1") is the measurement point where the vehicle was present in both the current frame and the previous frame, and "+1"("1"-"0") is the new vehicle in the current frame. Will indicate the measurement points at which the existence of the.

The speeds and directions of a group of vehicle corresponding points (corresponding to individual vehicles) are obtained based on the above-mentioned vehicle quaternary data, and predicted positions (corresponding measurement point group) of the respective vehicle corresponding points after a predetermined time has elapsed. ) Is defined as a vehicle movement vector (Mj) and is obtained as follows. In (c) of FIG. 4, the movement width ΔL ₁ of “+1”,
[Delta] L _2, obtains successively all ... etc., the speed of each vehicle V ₁ (= delta
L ₁ / ΔT), V ₂ (= ΔL ₂ / ΔT), ... Are sequentially calculated (V ₁ etc. are converted into km / h).

Next, it is known that a vehicle running at a normal V (km / h) requires a braking distance of 1/2 V (m) before the braking is stopped.
Each value of ₁ (m), 1/2 V ₂ (m), ... Is calculated in order.

Further, a window of a (m) × a (m) is provided around the measurement point of "+2" of the vehicle quaternary data, and a "-1" area and a "+1" area are provided in front and behind. Then, the movement vector direction is obtained. The direction from “−1” to “+1” is the direction of the moving vector of the vehicle.

After all, the movement vector of one vehicle is represented by the following four parameters. 1) Starting point of vehicle movement vector ... A measurement point group in which “+1” exists. 2) Size of vehicle movement vector ... Vehicle speed V (km /
For h), 1/2 V ₁ (m) + vehicle length (region length of “+1”, “+2”) 3) Direction of vehicle movement vector ... “−1” to “+2” or “+2” to “+1” Same as direction to. 4) Width of vehicle movement vector ... Width W of “+1” measurement point group.

From the above process, the movement vector (Mj) of each vehicle moving in the monitoring screen can be obtained in real time based on the detected vector for each vehicle, that is, the moving direction and the size (speed). . This vehicle movement vector is the position of the measurement point on the screen that each vehicle occupies at an arbitrary time after a predetermined time has passed, so to speak, the “expected occupation range”.
Is represented.

Subsequently, the control unit 40 detects a dangerous traffic event based on each vehicle movement vector (Mj). That is, the vehicle movement vector (Mj) and the intersection prohibition vector (M'j:
Intersection with vehicles (including the vehicle movement vector Mj itself) By verifying the presence or absence of measurement points (interaction / collision)
Alternatively, a measurement point group corresponding to a stop line or a pedestrian crossing is set as the intersection prohibited area (M'j) to set the vehicle movement vector (M
By detecting whether there is an intersection with j), a dangerous traffic event such as a signal-ignoring vehicle or a contact between a vehicle and a pedestrian can be detected.

First, if there is an intersection between vehicles, each vehicle movement vector (Mj) is regarded as an intersection prohibition vector (M'j), and it is verified whether or not there is an intersection between the vehicle movement vectors (Mj). To do. For example, a value of, for example, "1" indicating vehicle occupancy at the data storage address in the memory corresponding to the measurement point group corresponding to the assumed occupancy range of each vehicle (that is, the vehicle movement vector Mj) after time t ₁ is sequentially set. Add up. As a result, the value of "1" or more is held in the data of the storage address corresponding to the measurement point where the vehicle movement vectors Mj in the screen match each other (the actual event corresponds to the contact or the collision). Therefore, the measurement point with a value of "1" or more (hereinafter,
If there is an intersection measurement point Nij), it can be determined that a dangerous traffic event (Lm) such as a collision is present and a danger detection signal (Sm) can be output. Further, the position of the intersection measurement point represents the occurrence position of the dangerous traffic event. In addition, usually 1
Multiple intersection measurement points (Nij) corresponding to one dangerous traffic event
Therefore, the intersection area (L'm) consisting of a set of intersection measurement points (Nij) that are positionally continuous (adjacent) is regarded as one dangerous traffic event (Lm). Correspond to dangerous traffic events. The number of dangerous traffic events can be known from the number of intersection areas (L'm).

As described above, the method for detecting a dangerous traffic event according to the first aspect of the present invention follows each step for obtaining the binarized data described above, and the following steps (see FIG. 10): A four-valued data group (Fn) consisting of the difference between the two vehicle binary data groups (Dn) obtained with a constant time difference is obtained. 6. A vehicle movement vector (Mj) indicating each position of the vehicle corresponding point prediction point after the lapse of a predetermined time of each vehicle corresponding point by obtaining the speed and direction of the vehicle corresponding point based on the four-valued data group (Fn)
Is calculated. 7. An intersection measurement point (Nij) at which intersection prohibition vectors (M'j) intersect each other, such as a stop line including each vehicle movement vector (Mj), is calculated. One dangerous traffic event (Lm) corresponding to one intersection region (L'm) consisting of a series of intersection points (Nij)
Outputs a danger detection signal (Sm) indicating that the The dangerous traffic event is detected through the processes of and.

The binarization corresponding to the measurement points of 50 × 50 = 2500 points in the embodiment is the device configuration shown in FIG.
It can be executed at high speed in real time in a field unit (1/60 seconds) by a DSP (digital signal processor). The details of this process are also described in Japanese Patent Application No. 2-31899. The vehicle quaternization is also completed in 2500 points × 10 microseconds = 2.5 msec, and real-time processing is possible within one field (1/60 second) including the calculation of the vehicle movement vector. For example, as shown in FIG.
If the host CPU performs the quaternarization process on the vehicle in the even field, the frame unit (1/30 seconds)
Thus, a dangerous traffic event can be detected by obtaining each processing movement vector in the monitoring screen.

If an image is recorded for a predetermined time in response to the detection of a dangerous traffic event, recording is performed only for a certain time having valid information and unnecessary video recording is not performed, so that the recording medium can be saved and the recording can be performed. Maintainability is improved by reducing the frequency of medium replacement. In addition, since unnecessary parts are appropriately removed and recorded, there is no need for the labor required for later analysis. In addition, interlocking the dangerous event detection output with the video signal recording device,
In response to the detection of a dangerous traffic event, the pre-trigger mechanism can record a video signal for a certain time before and after the detection time. For this purpose, for example, video signal output (image),
A signal delay means for a predetermined time is provided between the input for actually recording this and the actual recording is started in response to the occurrence of the danger detection signal (Sm) and the recording is performed for the predetermined time. It may be configured.

Although the case of detecting a dangerous traffic event occurring between vehicles has been described above, in addition to this, the presence of a signal-ignoring vehicle can be detected as a dangerous traffic event in substantially the same process. In order to detect this signal-ignoring vehicle, as shown in FIG. 5B, a measurement point corresponding to the position of the stop line SL is grasped in advance, and the current traffic signal facing the stop line SL is detected. The indicated state is configured to be able to be input to the control unit by using an appropriate means, and the area of the stop line SL is virtually crossover prohibition vector (M ′) only during the period when the facing signal is the red signal.
j)) and verifying the intersection with the above-mentioned vehicle movement vector (Mj) which is another intersection prohibition vector (M'j) by calculation, the intersection between the vehicle and the stop line, that is, the presence of a signal-ignoring vehicle Is detected as a dangerous traffic event (Lm).

For this purpose, as in the case of vehicles, "1" is held only in the period during which the red traffic light is being displayed at the data storage address corresponding to the plurality of measurement points corresponding to the stop line area, and each of them is held. For example, a value "1" indicating the occupancy of the vehicle is added to the data storage addresses corresponding to the plurality of measurement points corresponding to the vehicle movement vector (Mj) of the vehicle. Therefore, if there is a vehicle ignoring the signal, the data of the storage address corresponding to the stop line is
The above values are retained, and the presence of the (anticipatory) signal-ignoring vehicle is detected as a dangerous traffic event.

Furthermore, when a pedestrian is present at a pedestrian crossing in the same manner, the possibility of intersection with a vehicle can be detected as a dangerous traffic event. To this end, the same FIG.
As shown in (b), the measurement points corresponding to the pedestrian crossing WL are grasped in advance, and when the pedestrian is detected by an appropriate means such as detecting the operation of a push button, the pedestrian crossing area is maintained for a certain period. Is virtually set as an intersection prohibition vector (M'j), and the intersection prohibition vector (M'j) corresponding to the intersection prohibition vector (M'j) and the vehicle movement vector (Mj) is set.
The intersection between the vehicle and the pedestrian crossing WL is detected as a dangerous traffic event by verifying the presence / absence of the intersection with the vehicle) by the same calculation as described above.

The first invention of the present application for detecting the fact itself of the occurrence of a dangerous traffic event has been described above. The risk level of each dangerous traffic event (Lm) detected further is quantified or the risk level is classified by grade. It can be classified and output. The second invention of the present application for this purpose will be described below.
The method for detecting a dangerous traffic event according to the second invention of the present application is the same as the above-described process (1.
7. ), The intersection measurement point Nij (equivalent to a dangerous traffic event) is first obtained, and then the risk of the dangerous event is quantitatively determined based on the distribution of the intersection, and the risk is also used as information. It is what is output. As a result, the dangerous traffic event can be classified and detected more finely, and thus more detailed analysis and countermeasures can be performed.

To obtain the above quantified risk,
When there are measurement points that intersect with each other, the degree of risk is quantified by performing weighted product-sum operation based on the distribution for each of these intersecting measurement points (Nij), and the value is classified by a threshold value to determine the danger grade. The risk levels are classified as 1, risk grade 2, ... In order to quantify this risk, for example, as shown in FIG. 5, among the intersection measurement points of the vehicle movement vector, a large weighting coefficient is set at the intersection measurement point close to the vehicle body. The product-sum calculation data of the measurement points (intersection measurement points) to be superposed is the cumulative value of (ai · bi).

As the above-mentioned weighting factor, for example, a weighting factor of +10 is set for the vehicle body portion of the vehicle movement vector (Mj), and for the corresponding portion of V / 2, in order from the column closer to the vehicle body, +.
5, +4, +3, +2, +1 are set as weighting factors. Then, after obtaining the product of the weighting factors for each intersection measurement point (Nij), all the results are accumulated and the dangerous traffic event (Lm) is concerned.
The risk data (Gm) of As shown in FIG. 7, the risk level data (Gm) of each dangerous event is compared with preset thresholds DL _{1 to} DL _4, and the risk grades G _{1 to} G ₄ associated with the intervals between the thresholds are compared. Each of the danger levels is determined depending on which danger grade it belongs to, and is sent to the subsequent device as the danger signal (Sm).

Besides, even if the number of intersection measurement points belonging to a certain group of intersection areas is excessive, it is possible to know the risk degree of the corresponding dangerous traffic event. That is, the number of intersecting measurement points belonging to one group of intersecting regions (L'm) increases as the occupied positions of both vehicles coincide, which is both directional and temporally related to speed. It corresponds to the case where the collision is unavoidable and the actual damage is large. On the contrary, when the number of intersection measurement points corresponding to the dangerous traffic event (Lm) is small, a little direction change operation or speed operation by the brake may be performed.
It means that the collision is more likely to be avoided. Therefore, the greater the number of intersection measurement points (Nij) belonging to one group of intersection areas (L'm), the higher the risk may be associated.

In the above description, the case of vehicles is taken as an example to quantify or classify the degree of danger of a dangerous traffic event. However, detection of a signal-ignoring vehicle described above (danger due to intersection of stop line and vehicle) Event) or the detection of a dangerous event due to the intersection of a pedestrian and a vehicle, it is of course possible to quantify and classify the risk level of the dangerous event, and based on the distribution situation of the intersection points exactly as described above. It is possible to output and collect information that takes into account the degree of risk.

As described above, in the dangerous traffic event detecting method of the second invention of the present application, the individual intersection measurement points (Nij) are first obtained by the same process as the first invention (1 to 7 described above). Further, 8. When a dangerous traffic event (L'm) is detected, the risk level of the dangerous traffic event (L'm) is quantified based on the distribution status of each intersection point (Nj) of the corresponding intersection point group (Lm). It is configured to include the process of obtaining the risk data (Gm) and outputting it (see FIG. 10).

As described above, the dangerous traffic event detection method of each of the above-described inventions using image processing is used to reliably detect a dangerous traffic event and, if necessary, quantify the risk grade of each event. You can This makes it possible for the first time to improve the following traffic measures that were difficult to implement in the past. 1) The shape and actual traffic flow is complicated, and it is applied to intersections where the frequency and risk grade of dangerous events occurring due to traffic are difficult to grasp, and the number of dangerous events in actual traffic flow and the quantification of the dangerous grade are applied. Then, the cause of the dangerous event occurrence can be considered by statistically analyzing the results, and appropriate measures such as improvement of the control program of the traffic light and extension of the traffic light can be taken. 2) Corresponding to the detection of a signal-ignored vehicle, the vehicle number of the signal-ignored vehicle can be recorded and can be used for traffic control. In this case, if only the necessary time is recorded by driving the image recording device with a pre-trigger as described above, it is possible to record without omission each time a signal is ignored, and at the same time, unnecessary portions are not recorded, which is convenient for later use. is there. 3) When installing a new traffic light, the current situation is mainly dependent on the requests from the residents. Therefore, the traffic light may not be installed depending on the actual traffic conditions. Unfortunately, on the other hand, there is the inconvenience that it is installed even though the need is not so great. But,
By quantifying dangerous events and quantitatively measuring their dangerous grades with the device of the present application, and installing traffic lights at intersections with a certain degree of risk or more, it is possible to install traffic lights that are objective and therefore expected to have actual effects. Yes, efficient traffic administration can be implemented.

Next, the third and fourth inventions of the present application will be described. In any of the above-described inventions, one camera for obtaining an image is used, and a dangerous traffic event is detected based on only a video signal obtained by the camera, and the degree of danger is calculated. Only using a camera can handle only a certain monitoring range due to the limit of resolution, and the monitoring range may be insufficient. For example, it is difficult to detect a dangerous traffic event by monitoring a wide intersection or a road with a single camera in terms of resolution.

The following invention corresponds to such a case. A plurality of video signals are obtained by using a plurality of cameras, and the same image processing as that described above is applied to apply a wider range of surveillance area or surveillance. Dangerous traffic events can be detected as the target range up to the extended area of the area. In this case, the same dangerous event may be detected by a plurality of cameras in duplicate, so it is necessary to eliminate it as described later. First, a dangerous traffic event detecting method that uses a plurality of video signals (cameras) in cooperation with each other according to the third invention of the present application will be described in detail (see FIG. 10).

In FIG. 8, the intersections are shown by a plurality of cameras C1 to C5.
2 shows an example of the arrangement of cameras when monitoring and detecting by. On the road surface such as an intersection to be monitored, the ground surface coordinates (Xi, Yj) are set in advance in a grid pattern, and the respective regions RE1 to RE are imaged corresponding to the respective cameras C1 to C5.
5, a plurality of coordinate measurement points (Ppij) of appropriate coordinates are determined for each camera in conformity with the appropriate ground surface coordinate (Xi, Yj) position, and are recognized by the control unit together with the ground surface coordinate information. It

That is, as shown in FIG. 9, a large number of coordinate measurement points (Ppij) associated with the ground surface coordinates are set on the monitor screen (DIS ₃ ) of the camera C3, for example, as shown in FIG. . Then, the brightness level (Cpij) is obtained for each coordinate measurement point (Ppij), and subsequently, the same processing as that described above is performed with the coordinate value. That is, each brightness level (Cp
ij), the coordinate luminance data group (Dtp) is sequentially obtained, and this is compared with the coordinate reference road surface data (Drp) to obtain the coordinate vehicle 2
A group of digitized data (Dnp) is sequentially obtained.

A coordinate quaternarized data group (Fnp) is obtained by difference calculation based on these, and from this, a vehicle movement vector (Mjp) (coordinate) indicating the predicted position of the vehicle corresponding point at the position coordinate on the road is calculated. To be done. Then, a dangerous traffic event (corresponding to the intersection point of the (coordinates) vehicle movement vectors (Mjp) or the intersection point with another (coordinates) intersection prohibition vector (M'jp), that is, the coordinate intersection measurement point (Npij) Lmp) is detected and at the same time the ground surface coordinates (Xi, Yj) of the occurrence location are obtained. In this way, a dangerous traffic event is detected in association with the coordinates set on the ground surface.

Similarly, the dangerous event detected by other cameras is detected in association with the ground surface coordinates (Xi, Yj). In this way, since the dangerous traffic event (Lmp) in the arbitrarily set area is detected using each of the plurality of cameras, it is possible to detect the dangerous traffic event in a wider area by integrating these. It will be possible.

When the image pickup areas overlap,
Although the same event is detected as a separate dangerous traffic event for each channel, the dangerous traffic event detected by each camera is represented by the ground surface coordinates (Xi, Yj). A temporary coordinate danger prediction signal (S'mp) is temporarily output corresponding to the detected dangerous traffic event,
After comparing the ground surface coordinate coordinates (Xi, Yj) corresponding to each of these provisional coordinate danger prediction signals (S'mp), and if there are a plurality of ones showing the same point, after correcting them collectively as one case , All remaining temporary coordinate danger prediction signals (S'm
By outputting p) as the final danger prediction signal (Smp), duplicate detection of the same dangerous traffic event is prevented.

The ground surface coordinates associated with the temporary coordinate dangerous traffic event are not strictly single but specified with a certain range, or a representative value is used for the ground surface coordinates and there is a difference in comparison. There is ambiguity such that the coordinates within a predetermined range are considered to be the same. Further, in associating the actual ground surface coordinates with the measurement points, the coordinates do not have to be associated with all the measurement points. For example, as shown in FIG. 9, the ground surface coordinates of P _{1 to} P ₄ are known. Only the four measurement points are associated with each other on the screen and the four measurement points are associated with the surface coordinates, and the remaining measurement points can be set in association with the surface coordinates by performing simple geometrical correction and calculation.

Also in the present invention, just as mentioned above, as the dangerous traffic event (Lmp), in addition to the intersection between vehicles, the signal is ignored due to the intersection between the stop line and the vehicle in the red signal indicating period. The contact with the pedestrian due to the intersection of the pedestrian crossing and the vehicle may be detected. Of course, it is possible to quantify the degree of danger for each dangerous traffic event (Lmp), obtain the degree of danger data (Gmp), and output this, according to the same technical idea as described above (the fourth aspect of the present application). Invention, FIG.
0).

According to the third and fourth inventions of the present application described above, the same effect as that of the previous invention can be obtained by targeting a wider area, that is, a dangerous traffic event can be reliably detected and, if necessary, the danger of each event. The grade can be quantified. Therefore, it becomes possible to improve the above-mentioned transportation policy, which was difficult in the past.

In addition, when a plurality of cameras are used, the respective imaging ranges can be used without overlapping, and the outside of the imaging ranges can be virtually calculated to increase the monitoring range of the dangerous traffic event. It can be expanded further. For example, only the inflow portion of the intersection of the four-way intersection is imaged to obtain the intersection at the intersection where no image is actually obtained, or the intersection with the stop line that is not in the imaging range is virtually detected. In this case as well, it is possible to obtain dangerous traffic events and materials for estimating the degree of danger. However, it is not possible to record an image outside the imaging range, and the application is limited in this respect.

[0064]

As described above in detail, in the dangerous traffic event detecting method of the present invention, a vehicle in which a video signal corresponding to an image of a region including a road is obtained by the image pickup means and binarized as a vehicle corresponding point by image processing is used. Using the binarized data group (Dn), a four-valued data group (Fn) consisting of the difference between the two vehicle binarized data groups (Dn) obtained at a constant time difference is obtained, and the four-valued data is obtained. The velocity and direction of the vehicle corresponding points are obtained based on the data group (Fn) to calculate the vehicle movement vector (Mj) indicating the predicted position of each vehicle corresponding point after a predetermined time has passed, and each vehicle movement vector (Mj) is calculated. The intersection measurement point (N ') at which intersection prohibition vectors (M'j) such as stop lines intersect
By calculating ij), a dangerous traffic event (Lm) can be detected. This makes it possible for the first time to make the following improvements in traffic measures that were difficult in the past. The detection output of a dangerous traffic event (Lm) can be used together with the pre-trigger function of the image recording device to eliminate unnecessary times by the image recording device and record only the required time (necessary traffic event) without fail. Can be extended equivalently, and unnecessary portions are not recorded at the same time, which is convenient for later use. Also,
If you record the vehicle number of the signal-ignoring vehicle, you can use it for traffic control. In addition, by applying to intersections and counting dangerous events in the actual traffic flow and statistically analyzing the results, it is possible to take objective and appropriate measures such as the addition of traffic lights.

In the second invention of the present application, when the dangerous traffic event (Lm) is further detected, the corresponding intersection area (L'm) is detected.
By performing the above-described predetermined processing based on the distribution of the intersection measurement points (Nij) in (1), the risk data (Gm) quantifying the risk can be obtained. Therefore, by using this risk data (Gm), in addition to the above-mentioned effects, it is applied to intersections where actual traffic flow is complicated and difficult to grasp, and statistics of dangerous events and risk grades of actual traffic flow are collected. Through analysis, it is possible to consider the causes of dangerous events and improve the control program for traffic signals.

Further, in the third and fourth inventions of the present application, the actual position coordinates (Xi, Y) on the road are particularly used by using a plurality of image pickup means.
i)), the dangerous traffic event is detected by performing the same processing as that of the above-described invention with respect to the plurality of measurement points respectively determined to be associated with each other, or the risk degree is subsequently obtained. Therefore, it is possible to obtain the same effect on the traffic policy as described above in a wider monitoring area.

[Brief description of drawings]

FIG. 1 is an explanatory view virtually showing measurement points on an image surface in a dangerous traffic event detection method according to each invention of the present application.

FIG. 2 is an explanatory view virtually showing an example of a vehicle binary data group according to the dangerous traffic event detection method of each invention of the present application in association with an image plane.

FIG. 3 relates to a method for detecting a dangerous traffic event according to each invention of the present application, 2
FIG. 9 is an explanatory view virtually showing an example of a vehicle binary data group at different times corresponding to the figure in association with an image surface.

FIG. 4 relates to a method for detecting a dangerous traffic event according to each invention of the present application, 4
It is explanatory drawing which showed an example of a value-ized data group and a vehicle vector virtually corresponding to an image surface.

FIG. 5 is an explanatory view (a) virtually showing an example of a case where vehicle movement vectors intersect with each other according to the dangerous traffic event detection method of the invention of the present application and an image plane, and a vehicle movement vector and intersection prohibited area. It is explanatory drawing (b) which showed an example when (a stop line, a pedestrian crossing) cross | intersects virtually on the image surface.

FIG. 6 is a diagram illustrating an example of a time relationship of each processing process according to the dangerous traffic event detection method of each invention of the present application.

FIG. 7 is a diagram illustrating an example of a process of calculating risk data of a dangerous traffic event according to the dangerous traffic event detection method of the second and fourth inventions of the present application.

FIG. 8 is a diagram illustrating an installation example of a plurality of cameras according to the dangerous traffic event detection method of the third and fourth inventions of the present application.

FIG. 9 is a diagram for explaining the relationship between measurement points and ground surface coordinates according to the dangerous traffic event detection method of the third and fourth inventions of the present application.

FIG. 10 is a flowchart showing an embodiment of the method for detecting a dangerous traffic event according to each invention of the present application.

FIG. 11 is a block diagram of an electric circuit and the like showing an example of an image processing apparatus suitable for carrying out the dangerous traffic event detection method of the present invention.

FIG. 12 is a diagram showing an arrangement of image pickup means and the like at an intersection according to the present application.

FIG. 13 is a diagram showing an image of an intersection according to the present application.

[Explanation of symbols]

10 ... Imaging means, 20 ... Luminance data converter, 30
... arithmetic processing unit, 40 ... control unit, (Pij), (Ppi)
j) ... measurement point, (Cij), (Cp ij) ... luminance level,
(Xi, Yj) ... position coordinates, (Dt) ... luminance data group,
(Dtp) ... Coordinate luminance data group, (Dr) ... Reference road surface data, (Drp) ... Coordinate reference road surface data, (Dn)
... Vehicle binary data, (Dnp) ... coordinate vehicle binary data, (Fn) ... quaternary data group, (Fnp) ... coordinates 4
Quantized data group, (Mj) ... Vehicle movement vector, (Mj
p) ... Coordinate vehicle movement vector, (M'j), (M'jp)
… Intersection prohibition vector, (Nij)… Intersection measurement point,
(Np ij) ... coordinate intersection measurement point, (L'm) ... intersection region,
(L'mp) ... Coordinate intersection area, (S'mp) ... Temporary coordinate danger prediction signal, (Smp) ... Danger prediction signal, (Gm) ...
Risk data (Lm), (Lmp) ... Dangerous traffic event.

Claims (4)

[Claims] 1. An image pickup unit images an area including a road to obtain a corresponding video signal, and temporal images in the video signal corresponding to each of a plurality of measurement points (Pij) set on the road in the image. Luminance data group (D per field) consisting of luminance levels (Cij) at each position
t) are sequentially obtained, and the luminance level (Cr) corresponding to the road surface at each measurement point (Pij) is obtained.
ij) is extracted at a predetermined interval and stored as reference road surface data (Dr), and the brightness data group (Dt) is compared with the reference road surface data (Dr). A binarized vehicle binarized data group (Dn) is obtained as a corresponding point, and the binarized data group is sequentially held, and the binarization is made of the difference between the two vehicle binarized data groups (Dn) obtained with a constant time difference. The data group (Fn) is obtained, the speed and direction of the vehicle corresponding point are obtained based on the four-valued data group (Fn), and the vehicle movement vector (Mj) indicating the predicted position of each vehicle corresponding point after a predetermined time has elapsed. The intersection measurement point (Nij) at which intersection prohibition vectors (M'j) such as stop lines including each vehicle movement vector (Mj) intersect is calculated, and the intersection measurement point (Nij) is found. In some cases, an intersection consisting of consecutive intersection measurement points (Nij) Frequency (L '
m) A dangerous traffic event detection method comprising the steps of outputting a danger detection signal (Sm) indicating that one dangerous traffic event (Lm) is detected corresponding to one m). 2. When the dangerous traffic event (Lm) is detected, the degree of danger of the dangerous traffic event (Lm) is determined based on the distribution of the intersection measurement points (Nij) in the corresponding intersection area (L'm). The method for detecting a dangerous traffic event according to claim 1, further comprising the step of obtaining and outputting the quantified risk data (Gm). 3. An image of a region including a road is picked up by a plurality of image pickup means to obtain corresponding video signals, and for each video signal, an actual position coordinate (X
i, Yj) and a plurality of measurement points (P
p ij) The coordinate brightness data group (Dtp) consisting of the brightness level (C p ij) at each temporal position in the video signal corresponding to each p ij) and the position coordinate information is sequentially obtained, and for each video signal In addition, a luminance data group consisting of luminance levels (Crpij) corresponding to the road surface at each measurement point (Pp ij) is extracted at a predetermined interval and stored as coordinate reference road surface data (Drp). The brightness data group (Dtp) is compared with the coordinate reference road surface data (Drp), and the data in which a sufficient difference is recognized is binarized as a vehicle corresponding point to obtain a coordinate vehicle binarized data group (Dnp), which is sequentially obtained. The coordinate quaternary data group (Fnp) which is held and obtained from the difference between the two coordinate vehicle binary data groups (Dnp) obtained with a constant time difference is obtained, and the coordinate is calculated for each area. Velocity of vehicle corresponding points based on four-valued data group (Fnp) The direction is calculated, and the coordinate vehicle movement vector (Mjp) indicating the predicted position on the road (Xi, Yj) on the road after a predetermined time elapses for each vehicle corresponding point is calculated. A coordinate intersection measurement point (Np ij) at which intersection prohibition vectors (M'jp) such as a stop line including a vehicle movement vector (Mjp) intersect is calculated, and a coordinate intersection measurement point (Np) is calculated.
ij), the continuous coordinate intersection measurement points (Np i
1 corresponding to one coordinate intersection region (L'mp) consisting of j)
A temporary coordinate danger prediction signal (S'mp) indicating that one dangerous traffic event (Lmp) has been detected is obtained by associating it with the actual position coordinate (Xi, Yj) on the corresponding road, and is obtained for every area. Temporary coordinate danger prediction signal (S'mp)
If any of the position coordinates (Xi, Yj) substantially coincide with each other, they are combined into one, and all temporary coordinate danger prediction signals (S'mp) are output as danger prediction signals (Smp). A method for detecting a dangerous traffic event characterized by comprising each process. 4. When the dangerous traffic event (Lmp) is detected, the dangerous traffic event (Lmp) is calculated based on the distribution status of the coordinate intersection measurement points (Np ij) of the corresponding intersection area (L'mp).
4. The method for detecting a dangerous traffic event according to claim 3, further comprising a step of obtaining and outputting risk data (Gmp) quantifying the risk level of GMP.