JPS5843797B2 - Vehicle shadow identification method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for distinguishing signals caused by vehicle shadows, which are one of disturbance factors in traffic flow measurement, from signals caused by vehicles.

BACKGROUND ART One type of traffic flow measuring device that uses an optical system is used to collect traffic flow information such as the speed of vehicles traveling on a road, the number of vehicles passing by, the distance between vehicles, the degree of congestion, and so on.

This usually collects various types of traffic information based on the output video signal of a light-receiving element placed on the imaging plane of an optical system that covers a desired point or range of a road within its field of view.

This traffic flow measuring device has a problem in that it detects the shadow of the vehicle in addition to the vehicle image.

The shadow of a vehicle includes a shadow of a vehicle to be detected and a shadow of a vehicle traveling in another lane and not within the field of view of the optical system.

The present invention provides a vehicle shadow identification method that can distinguish between a signal caused by a vehicle and a signal caused by a shadow, as well as distinguish between the two types of shadows described above.

Hereinafter, with reference to the drawings, a case will be described in detail about a case where a traffic flow measuring device installed at a position that is a multiple of the required range of the road and having a large number of light receiving elements is used.

This measuring device has the advantage of being able to collect vehicle information from many points on the road at one location and being easy to install.

FIG. 1 shows how the camera of the traffic flow measuring device is installed.

The camera 1 is mounted at a required height position (for example, 6 m) above the road using a support 3 or the like, and within a required range (for example, 6 m) in the length direction of the road.
For example, it is set up to be an enemy of Loom).

As shown in FIG. 4, the camera 1 includes an optical system 4 and a large number of light receiving elements.

In this example, there are six detection points P1 within the field of view of camera 1.
~ There is P6.

On the imaging plane of the optical system 4 in the camera 1 and corresponding to these detection points, one pair of light receiving elements ds1.
aR1-ds6. dR6 has been placed.

These light-receiving elements detect a vehicle traveling on one lane of the road.

If you feel safe, place the light receiving elements ds1 to ds on the image forming surface of the optical system 4.
Another row of light receiving element groups may be arranged in parallel to the arrangement direction of R6 to measure traffic flow in two lanes.

The light receiving element consists of a photo diode, for example.

At each detection point (represented by P), as shown in Figure 2,
A set area S and a reset area R are set at a required interval l.

A pair of light receiving elements (represented by dsdFt) correspond to each of these regions S and R, respectively.

When the vehicle OA passes the detection point P from the set area S to the IJt area R, the light receiving element d
A vehicle detection signal (video signal) shifted by a time t from S and dR is output.

Here, the rising edge of the output of each light receiving element dS, dR is detected, and the difference between the rising edges of both signals is defined as the detection time t.

Since this detection time t is the time required for the vehicle to travel between the set area S and the reset area R (distance l), the traveling speed V of this vehicle is determined by the following equation.

However, K is a constant. The distance that the vehicle actually moves during the detection time t is influenced by the color of the vehicle, the height of the vehicle, etc., and differs slightly from vehicle to vehicle, and is actually not equal to the above-mentioned distance l.

However, by imagining a sensing surface at a position above the road surface by the required height, measuring the distance between both areas S and R on this sensing surface, and statistically correcting this distance, it is possible to increase the vehicle's running speed. It is possible to measure with high precision.

Furthermore, if necessary, the average value of the traveling speeds of the same vehicle measured at the six detection points P1 to P6 may be determined.

Light receiving element ds that detects the set area S and reset area R
, dR may be used to calculate the traveling speed of the vehicle, not on the difference in time between the rises of the output signals of dR, but on the basis of the time when these rises reach a required level.

The distance between vehicles and the degree of congestion can also be calculated based on this traveling speed.

For example, when a vehicle at the detection point P1 is detected, the headway distance can be determined from the product of the speed of the preceding vehicle detected at the detection point P1 and the elapsed time.

The degree of congestion can be determined, for example, by the number of points among the six points P1 to P6 where the measured travel speed is less than or equal to a predetermined speed (for example, 20 Km/h).

Now, the processing device 2 performs the various processes described above, as well as the waveform pattern detection process of the video signal and the vehicle shadow identification process, which will be described later.

With reference to FIG. 4, each light receiving element ds, dR in the camera 1
After the output signal is amplified by an amplifier 5 equipped with an automatic gain control function, it is sent to a multiplexer channel device 6, where the outputs of each of the 12 light receiving elements are sequentially switched and A-D converted. Sent to vessel I.

The AD converter 7 performs AD conversion on the input output of the light receiving element at a predetermined sampling period (4.8 m5 in this example), and sends the result to a central processing unit (referred to as CPU) 11.

The processing device 2 is equipped with two 0PU11.12.

Based on the data sent from the A-D converter 8, the 0PU 11 executes pattern detection processing of the output signal waveform of each light receiving element, running speed calculation processing at each of the above-mentioned detection points P1 to P6, etc. Multiplexer
Controls the channel device 6 and the A-D converter 1.

The 0PU12 executes vehicle shadow identification processing based on the information obtained by the 0PU11.

Each 0PU 11.12 is equipped with a program memory (not shown) that stores an execution program, and local memo IJ 13 and 14 that store data, respectively.
The processing device 2 also includes a shared memory 15 that can be accessed by both CPUs 11 and 12.

The processing device 2 is provided with a filter circuit and an amplifier for removing low frequency components of the signal input from the camera 1, but illustration thereof is omitted for the sake of simplification.

First, the waveform pattern detection process of a video signal will be explained.

FIG. 5 shows an example of the waveform of the video signal output from the light receiving element of the camera 1.

This human input signal is AD converted by an AD converter γ every sampling period.

The current AD conversion result is defined as current data Dt.

In order to determine whether the current data Dt is increasing, decreasing, or in equilibrium compared to the previous data, it is necessary to find the deviation Jω between the current data Di and the previous data. .

The previous data that is the object of determining this deviation Jω is assumed to be preprocessed data DO.

Further, for the above determination, the reference amount to be compared with the deviation Jω between the current data Dt and the pre-processed data DO is defined as a reference amount ω0.

Furthermore, the concept of a prescribed period T is introduced, and this prescribed period T is set to be a time multiple times (in this example, four times) the sampling period.

Then, for each sampling period, the deviation Jω and the reference amount ω0
When the deviation Jω is greater than or equal to the reference amount 00,
Alternatively, when the prescribed period T has elapsed, the current data Di is set as the preprocessed data DO and the preprocessed data DO is updated.

If the rise or fall of the signal is slow, the amount of increase or decrease may not reach the reference amount ω0 during the sampling period.

A prescribed period T is introduced to detect such slow changes in the signal, and it is checked whether the amount of change reaches the reference amount ω0 within the range of the specified period T.

The states of signal change include increase, decrease, and equilibrium state with neither increase nor decrease.

If Dt-Do≧ωO within the specified period T, it is regarded as an increasing state.

If D t-DO≦-ωO within the specified period T, it is regarded as a decreasing state.

Then, when these determinations are made, the current data Dt is adopted as the preprocessed data DO.

Also, even if the specified period T has passed, 1Dt-D. <If ω0, it is regarded as an equilibrium state, and the current data Di is adopted as the preprocessing data DO.

FIG. 5a shows a typical example of a vehicle video signal.

Referring to this figure, in a steady state where no vehicle is detected, the preprocessing data is 0 (steady state level) (time t1
).

Let the preprocessed data at this time be DOl.

The signal hardly rises during each subsequent sampling period, and even when the specified period T has passed (time t2), the deviation (Jω = Dt - DOI) has not reached the reference amount ω0, so the time Between t1 and t2 is an equilibrium state.

Therefore, the current data at this time is preprocessed data DO2
Update as.

Even in the next sampling period, the deviation Jω does not reach the reference amount ω0.

Therefore, when one cycle of the sampling knob has elapsed (time t3), the deviation Aω is determined and compared with the reference amount ω0.

When two sampling periods have passed since time t2 (time t3), Jω≧ωO, so at time t
2 to t3 is an increasing state.

Therefore, at time t3, the current data at that time is updated as preprocessed data DO3.

Even when one sampling period has passed since time t3 (
At time t4), Jω≧ωO and it is determined that there is an increasing state,
Preprocessing data is updated as DO4.

Period of 4 sampling cycles after time t4 (regular cycle T
) is in an equilibrium state with almost no signal change, and the preprocessed data is changed to DO5 at time t5.

Between times t5 and t7, the signal falls sharply, and is determined to be decreasing at each sampling period, and the preprocessed data is updated (indicated by DO6 and DO7).

Then, after time t1, the equilibrium state is reached again, and the preprocessing data is updated every prescribed period T (such as DO8).
.

In FIG. 5a, the signal increases twice in a row between times t2 and t4.

Further, the decreasing state continues between times t5 and t7.

Assuming that when the same signal change state continues in this way, the waveform pattern is the same, and when the signal changes to another state, the signal waveform pattern changes, then the signal shown in FIG. It consists of a pattern of decrease and equilibrium.

A series of signal waveform patterns for such l signals are referred to as a waveform pattern group.

FIG. 5b shows an example of an output video signal waveform when the light receiving element of the camera 1 detects the shadow of a vehicle traveling in another lane adjacent to the traffic flow measurement lane.

Since the shaded area is dark, the signal falls below the steady state level.

This signal waveform pattern group is composed of patterns of balance, decrease, balance, increase, decrease, and balance.

The signal component A above the steady state level at the rear end of the signal is due to the characteristics of the circuit shown in FIG. 4, in particular the effect of the amplifier 5 with automatic gain control.

The level of this signal component A is not so high.

FIG. 5c shows a vehicle traveling on a traffic flow measurement lane,
An example of a video signal is shown when the vehicle casts a shadow in front of it.

The signal component B lower than the steady state level at the front end of this signal is due to the shadow of the host vehicle.

Since a vehicle is detected following this shadow, there is a large rise after signal component B.

The waveform pattern group of this signal is composed of a waveform pattern group of balance, decrease, increase, balance, decrease, and balance.

In order to execute such signal waveform pattern detection processing at 0PUII, the memory 13 is provided with areas M1 to M9 for storing various data.

Area M1 has the current data I) t, Area M2 has the preprocessed data DO, Area M3 has the deviation Jω, Area M4 has the previous signal state, and Area M5 has the reference amount ω.
0 is stored respectively.

The prescribed period T is stored in the area M6, the equilibrium continuation determination amount OO is stored in the area M7, and the area M8. M9 is used as a period counter C1 and a balance counter 02, respectively.

The period counter 01 counts whether or not the specified period T has been reached, and its contents are incremented by 1 every sampling period, and when the specified period T is reached, it is reset.

The balance force/Kunta C2 counts the continuous rotation in the balanced state, and is used to distinguish between a steady state where no vehicle is detected and a balanced state during vehicle detection such as from time t4 to t5. It will be done.

The equilibrium continuation determination amount CO represents the number of times the equilibrium state continues, which is the basis for this distinction, and a value larger than the maximum possible number of times the equilibrium state can be continued in vehicle detection is adopted.

Then, when the content of the equilibrium counter 02 is equal to or greater than this determination amount 00, it is determined that the steady state is present.

The waveform pattern detected by 0PUI 1 is stored in the shared memory 15.

The shared memory 15 has an area M that stores waveform patterns and signal levels (current data) prepared with a steady state level (O in this example) when the waveform pattern changes.
10 is provided, and each time the waveform pattern changes, the waveform pattern and the signal level at that time are sequentially stored in each storage location of this area M10.

FIG. 6 shows the procedure of signal waveform pattern detection processing by the 0PU11.

First, check whether the sampling period has elapsed (step 20), and if it is the sampling period, the input signal is
AD conversion is performed by the D converter 7, and this AD conversion result is stored in the area M1 as current data Dt (step 2).
1).

Then, the deviation Jω is calculated by subtracting the preprocessed data DO from the current data Dt (step 22), and the absolute value of this deviation jω is compared with the reference amount ω0 to check the magnitude relationship between them (step 23).

If the absolute value of the deviation Aω is smaller than the reference amount ω0, the contents of the period counter C1 are incremented by 1 (step 24), and it is checked whether the contents of this counter 01 have reached the specified period T (step 25).

If the specified period T has not been reached, the process ends.

When the prescribed period T has elapsed, the period counter C1 is reset to clear its contents (step 26), and an equilibrium state is determined (step 21).

Then, it is checked from the contents of area M4 whether the previous time was in an equilibrium state (step 28), and if the previous time was not in equilibrium but an increase or decrease, the waveform pattern has changed, so the area M10 of the memory 15 The next memory location (this, the equilibrium pattern and the current data Di (5th
b, designated as Lll, L21, etc. in FIG. 5c) (step 29.30).

After this, store the current state in area M4 and step 3
Go to 1.

If the previous time was in an equilibrium state, proceed directly to step 31.
Proceed to.

In step 31, +1t is added to the contents of the equilibrium counter C2.
Then, the magnitude relationship between the contents of the equilibrium counter C2 and the equilibrium continuation determination amount CO is checked (step 32).

If the content of the equilibrium counter C2 is equal to or greater than the determination amount 00, the steady state is reached, and a predetermined process is performed to indicate that the steady state has been returned (step 33).

Thereafter, the process moves to step 46, where the current data Di is stored in the area M2 as preprocessed data DO, and the process ends.

If the answer in step 32 is No, the process also moves to step 46.

If the absolute value of the deviation dω is greater than or equal to the reference amount 00 (step 2
3), it is an increase or a decrease, so the period counter C1 and the balance counter C2 should be reset, respectively.Steps 35 and 36) determine whether the deviation jω is positive or negative (
Step 37).

If the deviation jω is positive, it is determined that the state is increasing (step 3
8) Check whether the previous state was an increase (step 39).

If the previous data is not in an increasing state, the increasing pattern and current data Dt are stored in the area M10 (step 40.41).
).

Then, store the increased state in area M4 and step 4
Proceeding to step 6, the current data Dt is stored in the area M2 as preprocessed data DO.

If the previous time was an increase, the process directly advances to step 46.

If the deviation Jω is negative, it is determined that it is in a decreasing state.Step 4
2) As in the case of increase, check whether the previous state was in a decreasing state (step 43), and if the previous state is not in a decreasing state, store the decreasing pattern and current data Dt in the area MIO (steps 44 and 45). , the decreasing state is stored in area M4, and the process proceeds to step 46, where the current data D
t is stored in area M2 as preprocessed data DO.

If there was a decrease last time as well, the process continues at step 46, and all processes are completed.

By the way, although there are various types of signals caused by the shadow of a vehicle traveling in another lane depending on the shape, speed, etc. of the vehicle, it has been found from observation that the signals are limited to the six types shown in FIG. 1.

If we exclude the equilibrium patterns before and after the signal waveform and represent the increasing pattern by increasing, the decreasing pattern by decreasing, and the balanced pattern by flat, the waveform pattern group of the signal waveform shown in Figure 1a is decreasing and increasing, and the number of patterns is is 2.

Similarly, the signal waveform in Figure 7b has 3 patterns with decrease, flat, and increase, and the waveform in Figure 7c has 3 patterns with decrease, increase, and decrease.
The waveform in Figure 7d has a decreasing, increasing, flat, decreasing pattern number of 4, rth
The waveform in figure e has 4 patterns of decreasing, flat, increasing, and decreasing, and the waveform in figure 7f has 5 patterns of decreasing, flat, increasing, flat, and decreasing.

The shadow patterns shown in Figure 1d match the vehicle patterns shown in Figure 5c.

however. Both waveforms differ significantly in the magnitude of the signal component levels above the steady state level.

Focusing on this point, it is higher than component A of the signal due to the shadow,
By setting a predetermined level LO (shadow determination reference level) lower than the signal from the vehicle and comparing the levels of both signals with this reference level LO, it is possible to distinguish between the two waveforms.

0PU12 uses the data stored in the memory 15,
Based on the characteristics of the signal waveform caused by the shadow shown in Fig. 7, it is determined whether the detected video signal is caused by a shadow traveling in the other lane or by a vehicle as shown in Figs. 5a and 5C. There is.

The shadow determination reference level LO used for this process is set in advance in area Mll of the memory 14.

FIG. 8 shows the procedure of shadow identification processing by the 0PU12.

First, among the patterns stored in area MIO of the memory 15, the first waveform pattern following the steady state equilibrium pattern is read out to see if it is a decreasing pattern (step 51).

As can be seen from Fig. 7, the first pattern of the shadow pattern group is always a decreasing pattern, so if the first pattern read out is not decreasing, it is determined to be a signal waveform caused by a vehicle, and various processing is performed on the vehicle signal. (Step 59).

If the first pattern is a decrease, there is a possibility that it is a shadow, so next it is checked whether the number of patterns is 5 or less (step 52).

The maximum number of signal waveform patterns due to shadows is five.

Some vehicle signals have six or more patterns (
(not shown in Figures 5a and 5C).

Therefore, if the number of patterns is not 5 or less, it is determined that it is a vehicle signal (step 59).

If the number of patterns is 5 or less, the number of patterns is 2.
or 3 (step 5)
3.54).

In the case of a signal caused by a vehicle, it is impossible for the first pattern to be a decreasing pattern (YES in step 51) and the number of patterns to be two. Therefore, if the number of patterns is two, it is determined that the signal is caused by a shadow ( Step 58).

A signal with two patterns is shown in FIG. 7a.

If the number of patterns is not 2, then it is checked whether the number of patterns is 3, and if the number of patterns is 3, it is checked whether the second pattern is balanced (step 55).

A signal where the first is a decreasing pattern and the second is an equilibrium pattern, and the number of patterns is 3 cannot exist other than the one shown in Figure 7b, so if the second pattern is an equilibrium, it is determined to be shadow data. (step 58).

If the number of patterns is not 3 (NO in step 54), the number of patterns is 4 or 5.

In this case, and if the number of patterns is 3 and the second pattern is not balanced (No in step 55), the process proceeds to step 56 to check whether the last pattern is a decreasing pattern.

Among the six types of shadow pattern groups shown in FIG. 7, those shown in FIGS. 7a and 7b have already been removed in step 53.55, and those shown in FIGS. 7c to 7f have not yet been determined. .

The last patterns of these four types of shadow pattern groups are all decreasing patterns, so if the last pattern is decreasing, it is assumed that there is a possibility of a shadow pattern group and step 5 is performed.
Proceed to step 7 and determine that the others are vehicles (step 59).
.

In step 57, the level L14 of the last decreasing pattern is
．． L24 and the like are read out from the memory 15 and compared with the shadow determination reference level LO.

As mentioned above, in the case of a shadow, the level of signal component A has not reached the standard L/H/L/LO, so if the last level is within the standard level LO, it is determined to be shadow data, and the last level If it exceeds the LO, the vehicle will signal it.

As explained in detail above, according to the present invention, it is possible to clearly distinguish between signals caused by vehicles and signals caused by shadows, especially shadows caused by vehicles traveling in other lanes. can be collected.

[Brief explanation of drawings]

Figure 1 is a diagram showing the installation status of the camera of the traffic flow measurement device.
Fig. 2 is an enlarged view of the detection point, Fig. 3 is a waveform diagram showing the vehicle detection signal, Fig. 4 is a configuration and block diagram showing the inside of the camera and processing device, and Fig. 5 shows detection of the waveform pattern. Explanatory diagram, FIG. 6 is a flow chart showing the processing procedure when waveform pattern detection processing is executed by the CPU, FIG. 1 is a waveform diagram showing various signals with shadows, and FIG.
The figure is a flow chart showing a procedure for identifying signals based on shadows. ds1 to dR6... Light receiving element, Dt...
・This time data, DO...Preprocessed data, jω・
...Deviation, ω0 ...Reference amount, LO...
...Shadow judgment standard level.

Claims (1)

[Claims] 1. The output video signal of a light-receiving element that captures the image of a vehicle running on a road is sampled at a predetermined sampling period, and the data is used as current data, and for each sampling, the current data and the previous data are By determining the deviation from predetermined reference data and comparing this deviation with a predetermined reference amount, it is possible to detect whether the signal is an increasing, decreasing, or balanced pattern, and to generate a series of waveform patterns for that signal. This group of waveform patterns is compared with a preset group of waveform patterns of video signals due to the shadow of a vehicle, and if they match, the output video signal of the light receiving element is determined to be a signal due to the shadow of a vehicle. ,Vehicle shadow identification method. 2. If the video signal due to the shadow of the vehicle includes a signal component with a level higher than the steady state level, the output video signal of the light receiving element is A vehicle shadow identification method according to claim 1, wherein the signal is determined to be a signal due to a vehicle shadow.