JP2940296B2 - Parked vehicle detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of measuring a traffic flow of a traveling vehicle (including a stopped vehicle) from image information captured by a camera and detecting a parked vehicle.

[0002]

2. Description of the Related Art In a traffic control system for a general road or an expressway, a number of vehicle sensors are arranged on a road side to measure a parked vehicle. One of the more advanced measurement functions is a parked vehicle detection system using an ITV camera, which has been studied in the past.

The system for detecting a parked vehicle using an ITV camera uses a television camera as a sensor, and determines the presence of a vehicle by analyzing in real time an image taken while looking down on a road obliquely. FIG.
FIG. 7A is a diagram for explaining an outline of conventional processing. FIG. 7A shows a measurement area 51 on the screen, FIG. 7B shows measurement sample points set in each lane, and FIG. Fig. 7 (4) shows the measurement sample points and the vehicle existence region (represented by reference numeral 1) converted into rectangular coordinates, and Fig. 7 (4) shows the vehicle existence region (represented by reference numeral 1) viewed from the cross direction of the road. Are shown).

[0004] A parked vehicle can be detected based on the information on the area where the vehicle is detected (the portion indicated by reference numeral 1) (see "Sumitomo Electric" No. 127, pp. 58-62). , September 1985). According to the processing method,
The process of assigning a sign to the measurement sample point is performed by taking the difference between the luminance data of each point and the road surface reference luminance. The road surface reference luminance is determined as follows.

That is, after creating an initial background image, a value obtained by subtracting the background image luminance value from the image luminance value is integrated,
When the integrated value exceeds a preset threshold value, the area where the integrated value exceeds the predetermined threshold value is regarded as the existence area of the parked vehicle.

[0006]

In the above-described method, it usually takes a considerable time (for example, 5 minutes) for the integrated value to exceed the threshold value, so that a short-time parked vehicle cannot be detected. Therefore, it was not possible to provide a warning before the driver who stopped the vehicle left the vehicle. actually,
If a vehicle attempts to stop in a no-parking zone, it is preferable to give a warning as soon as possible.

Further, since the integral value depends on the external environment,
For example, when it is dusk or cloudy, the detection time is longer than that in fine weather. The detection time also varies depending on the lightness difference (dark or bright color) of the vehicle body. If the threshold value is lowered in order to shorten the detection time, a vehicle other than the parked vehicle is also detected, and the detection accuracy is reduced.

Further, since there is only a judgment element only for the integral value, it is not possible to judge including an element such as the presence or absence of a nearby parked vehicle. Because a traffic flow usually exists in a parking prohibition area, it is preferable to make a distinction between a case where the vehicle is stopped due to traffic congestion or a signal waiting and a case where the vehicle is parked. Therefore, an object of the present invention is to determine whether or not a vehicle parked in a predetermined area such as a parking prohibited area is due to traffic congestion or a signal waiting from the surrounding situation, It is an object of the present invention to provide a parked vehicle detection method capable of giving a warning before the driver leaves the vehicle (for example, in the case of illegal parking).

[0009]

Means and Action for Solving the Problems

(1) The method for detecting a parked vehicle according to claim 1 for achieving the above object captures a road with a camera installed on the side of the road and, based on a luminance distribution included in the video information, detects a headway of the vehicle. The position is determined, and the speed of the vehicle is calculated from the change between the position of the head obtained in the previous video information and the current position of the head. For a vehicle having a speed of 0, the stop time T1 of the vehicle is calculated. Is determined to be longer than the threshold value, and if the stop time T1 of the vehicle is shorter than the threshold value, if there is a vehicle stopped ahead of the vehicle, the stop time T2 of the vehicle stopped ahead of the vehicle is determined. If the stop time T1 is longer than the threshold value or if the stop time T1 is longer than the stop time T2 of the vehicle stopped ahead, this method determines that the vehicle is a parked vehicle.

According to the above method, first, the position of the head of the vehicle is determined based on the luminance distribution of the sample points in the measurement area, and the speed of the vehicle is calculated from the change in the position of the head of the vehicle. Then, for a vehicle whose speed is 0, the parked vehicle is determined as usual by comparing the stop time T1 of the vehicle with a threshold value. If the stop time T1 is shorter than the threshold value, it has not been possible to determine that the vehicle is a parked vehicle in the past.
According to the present invention, when the stop time T1 of the vehicle is shorter than the threshold value, if there is a vehicle stopped in front of the vehicle, the stop time T2 of the vehicle stopped ahead is compared with the stop time T2 of the vehicle.
If the stop time T1 is longer than the stop time T2 of the vehicle stopped ahead, the preceding vehicle has stopped later than the vehicle to be determined, and thus it can be determined that this vehicle is a parked vehicle. (2) The method for detecting a parked vehicle according to claim 2 determines whether or not there is a vehicle parked ahead of the vehicle, and determines the stop time T
When 1 is longer than the threshold value or when there is a parked vehicle ahead, this method determines that this vehicle is a parked vehicle.

According to this method, if the vehicle in front is a parked vehicle, the vehicle stopped behind can be determined to be parked by the driver's intention without waiting for traffic lights or congestion. The vehicle can be determined to be a parked vehicle. (3) The parking vehicle detection method according to claim 3 has a plurality of lanes.
One road, it is determined whether there is a vehicle that is stopped or slowing in the lane adjacent to the vehicle, if the stop time T1 is longer than the threshold value, or if there is no vehicle that is stopped or slowing in the adjacent lane Is a method for determining that this vehicle is a parked vehicle.

[0012] According to this method, when a vehicle is stopped without a stopped vehicle or the like in an adjacent lane, it is unlikely that the vehicle is stopped due to a signal or a traffic jam. Can be instantaneously determined. (4) The parked vehicle detection method according to claim 4 is a method for immediately giving a warning to a driver of a vehicle determined to be the parked vehicle.

According to the method for detecting a parked vehicle, it is possible to determine that the vehicle is a vehicle to be parked even if the stop time of the vehicle is short, so that a warning is given before the driver leaves the vehicle. Can be.

[0014]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. FIG. 2 is a conceptual view of the installation of the ITV camera 2.
An ITV camera 2 is provided above a pole installed on the side of the road, and a control unit 1 is provided below the pole.

The field of view of the ITV camera 2 covers all lanes of a road having four lanes. FIG. 1 shows a device configuration in the control unit 1. The control unit includes an image input unit 3 for inputting an image signal acquired from the ITV camera 2, a vehicle candidate point detection unit 4, and a measurement processing unit 5. It has a main body, a transmission unit 6 for transmitting information such as a traffic flow measurement output calculated by the control unit main body to a traffic control center through a communication line, an input / output unit 7 for outputting a warning command signal, and a power supply unit 8.

The outline of the measurement processing performed by the control unit body will be described. As shown in FIG. 3, the image input unit 3 includes M sample points along the cross direction of the road (referred to as the ξ direction) and N sample points along the traveling direction of the vehicle (referred to as the η direction). Corresponding to M × N coordinates (i, j) determined from the point
Each luminance value P (i, i) of the image signal output from the TV camera 2
j) is stored. The intervals between the sample points are Δη and Δξ.

The vehicle candidate point detector 4 performs a spatial differentiation process in the η direction. Specifically, the following Sobel operator is applied to each (i, j).

[0018]

(Equation 1)

That is, the derivative P ′ (i, j) of the luminance value P (i, j)
j) by the equation P '(i, j) = P (i-1, j-1) + 2P (i-1, j) + P (i-1, j + 1) -P (i, j-1) −2P (i, j) −P (i, j + 1) However, in the special case where the calculation area exceeds the measurement area, P '(0, j) = 0 P' (i, 0) = 2P (i-1,0) + P (i-1,1) -2P (i, 0) −P (i, 1) P ′ (i, N−1) = P (i−1, N−2) + 2P (i−1, N−1) −P (i, N−2) ) -2P (i, N-1) is adopted.

The vehicle candidate point detection unit 4 applies a threshold value Th1 given as a constant in advance, and binarizes all pixels subjected to the spatial differentiation processing. That is, if P ′ (i, j) ≧ Th1, P ′ (i, j) = 1, and if P ′ (i, j) <Th1, P ′ (i, j) = 0.

The vehicle candidate point detecting section 4 includes a small car, a normal car,
Masks are prepared according to the classification of large vehicles. In this embodiment, the mask to be prepared is M1 as shown in FIG.
To M8. M1 to M4 are for ordinary vehicles, and M5 to M8 are for large vehicles. M1,2,5,6
Denotes a mask of two rows, and M3, 4, 7, and 8 denote masks of three rows. The pixel of interest is at the lower left of M1, 3, 5, and 7, and is at the upper left of M2, 4, 6, and 8.

The mask is applied by raster-scanning the measurement area, and matching the pixel to the "pixel of interest" of the mask when the pixel of the code 1 first appears. However, if 1 is continuous, the second and subsequent pixels are not masked. Then, the number of 1 pixels present in the mask is counted. The counted number is called a mask score.

For example, FIG. 5A shows an example in which a mask M1 is applied in accordance with the second pixel of interest (i, j) from the left and the second from the bottom. The score at this time is 9. FIG. 5B shows an example in which a mask M2 is applied to the second pixel of interest (i, j) from the left and the second from the bottom. The score at this time is 7. Then, for the pixel of interest, the mask number and the score are stored as a set. For example, in the case of FIG. 5A, (i, j, M1,
Store in the form as in 9). In the case of FIG. 5B, (i,
j, M2, 7).

As a result of multiplying the target pixel by eight masks, the mask having the highest score is selected. If the score of the mask for the large car and the score of the mask for the small car are the same, the mask for the small car is selected. Then, for the pixel of interest, only one set of the mask number and the maximum value of the score is obtained. If this is a mask that takes a score equal to or higher than a certain threshold, the mask is multiplied again. The center of gravity is determined based on the distribution of pixels that are 1. This center of gravity is called a vehicle head candidate point.

As a result, the coordinates of the head candidate point, the mask number, and the maximum score are stored as a set. For example, in the case of FIG. 5A, assuming that the coordinates of the center of gravity are (i, j + 5), the form is (i, j + 5, M1, 9). Hereinafter, the image data and the binarized data are not used, and the process proceeds based only on the information of the head candidate points.

In some cases, the information of the candidate head point includes information of a position different from the head, such as a plurality of head positions detected, a boundary between a windshield and a roof, a sunroof, and the like. From such a situation, the most probable head position (head effective point) must be extracted. Therefore, it is assumed that the measurement processing unit 5 sequentially examines the vehicle head candidate points, and that there are n vehicle head candidate points in an adjacent area (for example, an area where almost one vehicle exists). First, the first (n = 1) head candidate point is registered as a head effective point. Next, the score of the head candidate point after n = 2 is compared with the score of the head effective point, and the higher score is newly set as the head effective point, or a point closer to the traveling direction of the vehicle is newly set as the head effective point. And And
The head candidate points that have not become head valid points are deleted. In this way, the headway effective point is determined from a plurality of nearby headway candidate points.

If there are a plurality of valid head points in the measurement area, a point indicating the head position (vehicle fixed point) is determined from them. The procedure is as follows.
The positions of the vehicle head effective points are checked in order, and if there are m vehicle head effective points, the first vehicle head effective point is first registered as a vehicle head fixed point. Next, the next vehicle head effective point is compared with the registered vehicle head effective point. From the positional relationship between the two, if both points are within the range of the length and width of the vehicle corresponding to the mask, such as a large car, small car, etc., the vehicle heading direction between the head front fixed point and the next head front effective point Is determined as a head-fixed point, and the other points are deleted from candidates for the head-fixed point. In this way, each head effective point is checked, and the remaining head effective points are determined as head fixed points. If there are a plurality of head-fixed points, it is considered that a plurality of vehicles have entered the measurement area.

As described above, the measurement processing unit 5 can find the head-of-vehicle fixed point in the measurement area of one frame. Then, the positional relationship with the head-of-vehicle fixed point found in the measurement area one frame before is checked, and the speed of the vehicle is calculated.
Specifically, it reads the information of the vehicle head fixed point one frame before,
If there is no head-fixed point one frame before, the current head-fixed point is output as it is, and the speed is an average speed learning value calculated for each lane. When the head-fixed point exists one frame before, the area obtained by adding (the predicted speed range of the vehicle) x (one frame time) to the position of the head-fixed point is defined as the area where the vehicle proceeds to the next. Check if there is a current head fixed point in. Here, (the range of the predicted speed of the vehicle) ranges from a negative value to a constant positive value. The reason why negative values are included is that an attempt is made to detect a parked vehicle or a congested vehicle.

If a vehicle is present in the area, the actual speed of the vehicle is calculated based on the difference in distance between the determined head point and the determined head point one frame before. If the calculated speed has a negative value, the speed is replaced with 0. If the vehicle does not exist, the vehicle is considered as a vehicle that has newly entered the measurement area, and is output as a head-end fixed point. As described above, referring to the position information of the head-fixed point of the previous frame,
It is possible to calculate the predicted position in the current frame, extract the vehicle head fixed point closest to the predicted position, and obtain the vehicle speed. Further, since 0 or a negative value is included in the range of the predicted speed of the vehicle, even a stopped vehicle can be detected.

Next, it is determined by the following algorithm (see FIG. 6) whether or not the stopped vehicle is a parked vehicle. First, a vehicle whose speed is 0 (hereinafter, referred to as a "stop vehicle") is detected (step S1). If it is a stopped vehicle, it is determined whether the position of the stopped vehicle is near the center line (step S
2). If the vehicle is near the center line, the vehicle is not regarded as a parked vehicle, and it is determined that the traveling vehicle is stopped at a traffic light or an accident (step S11), and the search for the next vehicle is started (step S9).

If the position of the stopped vehicle is near the road shoulder, there is a possibility that the vehicle is a parked vehicle. Therefore, the stop time of the vehicle is measured from the past measurement results in the same measurement area, and it is determined whether the stop time is longer than a threshold (step S3). Even if it is shorter than the threshold value, if there is a vehicle stopped ahead, it is determined whether it is longer than the stop time of the vehicle stopped ahead (step S4). If the vehicle stops longer than the vehicle that stops ahead, the vehicle ahead comes and stops before the vehicle that has stopped, and the vehicle behind can be regarded as being parked.

If the vehicle stops after the vehicle stopped ahead, it is determined whether the vehicle stopped ahead is a parked vehicle (step S).
5). If the vehicle stopped ahead is a parked vehicle, the vehicle behind can be immediately regarded as having intention to park. If there is no vehicle stopped ahead, it is checked whether there is a stopped vehicle or a low-speed vehicle (in the case of slowing down due to traffic congestion) in the adjacent lane (step S6). If not, the vehicle can be considered to be willing to park.

If YES is selected in any of the steps S3 to S6, the vehicle having the speed 0 is regarded as a parked vehicle and registered (steps S7 and S8), and the next vehicle is examined. (Step S
9). If there is no vehicle, the process ends (step S1).
0). As described in the above processing, the difference in the stop time from the preceding vehicle is checked in step S4, so that the determination can be made more reliably and promptly as compared with the conventional case where only the stop time of the vehicle is used.

If the vehicle stops behind the stopped vehicle determined to be a parked vehicle in step S5, the vehicle can be determined to be a vehicle to be parked without measuring the stop time. It can be determined that the vehicle is parked. In step S6, if there is no stopped vehicle or low-speed vehicle in the adjacent lane and the vehicle stops, it is difficult to imagine that the vehicle is stopped due to a signal waiting or traffic congestion. Can be determined.

In steps S5 and S6, it is preferable to give a warning to the vehicle by voice or the like. The warning can be given by, for example, a speaker built in the control unit 1.

[0036]

As described above, according to the first aspect of the present invention, even if the stop time T1 is shorter than the threshold value, if there is a vehicle that is stopped ahead of the vehicle, the vehicle stopped ahead of the vehicle will be Compared to the stop time T2, if the stop time T1 is longer than the stop time T2 of the vehicle stopped ahead, the preceding vehicle has stopped later than the vehicle to be determined, and when the vehicle enters ahead, This vehicle can be determined to be a parked vehicle. Therefore, the detection rate of the parked vehicle can be increased.

According to the second aspect of the present invention, if the preceding vehicle is a parked vehicle, it is determined that the vehicle stopped behind is parked at the will of the driver without waiting for traffic lights or congestion. Therefore, it is possible to immediately determine that this vehicle is a parked vehicle, and it is possible to increase the detection rate of the parked vehicle. According to the third aspect of the present invention, if a vehicle stops without a stopped vehicle or a slow vehicle in the adjacent lane, it is unlikely that the vehicle stops due to a traffic light or the like. It can be determined instantly that there is, and the detection rate of the parked vehicle can be increased.

According to the invention described in claim 4, in the invention described in claim 2 or 3, even if the stop time of the vehicle is short, it can be immediately determined that the vehicle is a vehicle to be parked. , Can give a warning before the driver leaves the vehicle.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a device configuration in a control unit 1 of a parked / stopped vehicle detection device.

FIG. 2 is a diagram showing an installation conceptual diagram of an ITV camera 2.

FIG. 3 is an arrangement diagram of sample points of the ITV camera 2.

FIG. 4 is a diagram showing eight types of masks prepared according to a vehicle type and the like.

FIG. 5A is a diagram illustrating an example in which a mask M1 is applied in accordance with a pixel of interest (i, j). (b) is a diagram showing an example in which a mask M2 is applied in accordance with a pixel of interest (i, j).

FIG. 6 is a flowchart showing a parked vehicle detection procedure.

FIG. 7 is a diagram illustrating an outline of detection of a parked / stopped vehicle in the related art.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Control part 2 ITV camera 3 Image input part 4 Vehicle head candidate point detection part 5 Measurement processing part

Continuation of front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G08G 1/00-1/16

Claims (4)

(57) [Claims] An image of a road is photographed by a camera installed on the side of the road, and the position of the head of the vehicle is determined based on the distribution of luminance contained in the image information. , The speed of the vehicle is calculated from the change from the current position of the head of the vehicle.
Is determined to be longer than the threshold value. If the stop time T1 of the vehicle is shorter than the threshold value,
If there is a vehicle stopped in front of the vehicle, it is compared with the stop time T2 of the vehicle stopped ahead, and if the stop time T1 is longer than the threshold value, or the stop time T
If 1 is longer than the stop time T2 of the vehicle stopped ahead, the method determines that the vehicle is a parked vehicle. 2. A method of photographing a road with a camera installed on the side of the road, determining the position of the head of the vehicle based on the distribution of luminance included in the image information, and determining the position of the head of the vehicle obtained in the previous image information. , The speed of the vehicle is calculated from the change from the current position of the head of the vehicle.
Is determined to be longer than the threshold value. If the stop time T1 of the vehicle is shorter than the threshold value,
It is determined whether there is a vehicle parked ahead of the vehicle. If the stop time T1 is longer than the threshold value, or if there is a parked vehicle ahead, this vehicle is determined to be a parked vehicle. A method for detecting a parked vehicle. 3. An image of a road taken by a camera installed on a side of a road having a plurality of lanes, a position of a head of the vehicle is determined based on a distribution of luminance included in the image information, and the position of the head is determined in the preceding image information. The vehicle speed is calculated from the change between the position of the head of the vehicle and the position of the current head of the vehicle.
Is determined to be longer than the threshold value. If the stop time T1 of the vehicle is shorter than the threshold value,
It is determined whether there is a vehicle that is stopped or slowing in the adjacent lane of the vehicle. If the stop time T1 is longer than the threshold value, or if there is no vehicle that is stopping or slowing in the adjacent lane, A method for detecting a parked vehicle, comprising determining that the vehicle is a parked vehicle. 4. The method for detecting a parked vehicle according to claim 2, wherein a warning is immediately given to a driver of the vehicle determined to be a parked vehicle.