JPH0883392A - Method and device for detecting vehicle - Google Patents

Abstract

PURPOSE: To separately detect plural vehicles even when the respective vehicles are projected by a monitor camera while overlapping one part at night. CONSTITUTION: A feature extraction circuit 4 binarizes current image data for one screen picked up with a monitor camera 1 and preserves these data in an image memory 5. A CPU 2 calculates the X- and Y-coordinate values at the end point of the locus data of a tail lamp shown by the binarized data in the image memory 5 and preserves the values in a magnetic disk device 7. The CPU 2 decides whether the Y-coordinate value at the end point includes a coordinate on one edge of one screen or not, judges the locus of the original tail lamp extracting this end point as an appearance locus only when the Y- coordinate value includes it, and strores the end point of this appearance locus in an external memory 6 as data detecting one vehicle again. When the vehicle appears in the picture and even when plural vehicles adjacently appear later, the vehicles can be exactly separately detected one by one without overlapping one part of the locus data of the tail lamp.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle detection device for processing an image captured by a surveillance camera to detect an object,
Particularly, the present invention relates to a vehicle detection method when detecting a vehicle traveling on a road.

[0002]

2. Description of the Related Art Conventionally, a vehicle detection device of this type has been used to detect a number of vehicles passing through a predetermined range on a road by using a video camera which is a surveillance camera. In such a vehicle detection device, the background of the range (the state where there is no vehicle) is picked up by the monitoring camera for one field in advance and stored in an image memory or the like. Since the surveillance camera outputs the current image in the range in units of one field, the difference between the current image in the one field and the background image stored in the image memory is calculated, the difference data is evaluated, and Detects whether a vehicle has passed the range.

That is, when the vehicle is not shown in the current image, the current image and the background image are the same, so the difference between them is zero, and it is detected that the vehicle has not passed (no vehicle). On the other hand, when the vehicle is present in the current image, the current image and the background image are not the same, so the difference is not zero, and it is detected that the vehicle is passing (the vehicle is present).

However, in the above-described method of detecting the presence or absence of a vehicle based on the difference between the current image and the background image, for example, when two vehicles pass in succession, a part of the two vehicles is displayed in the current image in an overlapping manner. In such a case, it is not possible to distinguish between two vehicles and one vehicle only by the difference data between the current image and the background image as described above, and eventually one vehicle There is a drawback that an error occurs when a vehicle has been erroneously detected as having passed and the amount of traffic is determined based on the above detection result.

Further, since the above-mentioned surveillance camera outputs the image data in one field cycle, the subsequent image processing must be performed in this cycle, and a high-speed CPU is required to perform the image processing. There is a drawback that the device becomes expensive.

[0006]

In the conventional vehicle detection device as described above, it is determined whether the vehicle has passed in the range, based on the difference between the current image and the background image when the surveillance camera picks up a predetermined range, that is, The presence or absence of a vehicle is detected. For this reason,
When a plurality of vehicles pass in succession, a part of the plurality of vehicles may be superimposed on the current image. In such a case, it is not possible to distinguish between a plurality of vehicles and one vehicle only by the difference data between the current image and the background image as described above, and it is erroneous that one vehicle passes after all. There is a drawback that an error occurs when the detection is performed and the amount of traffic is determined based on the above detection result. In addition, since the above-mentioned surveillance camera outputs image data in one field cycle, the subsequent image processing must be performed in this cycle, and a high-speed CPU is required to perform the image processing. It had the drawback of being expensive.

Therefore, the present invention eliminates the above-mentioned drawbacks, and at night, even if a part of a plurality of vehicles are overlapped and displayed on a surveillance camera, each vehicle can be detected separately. It is an object to provide a vehicle detection device using this method.

[0008]

According to a first aspect of the present invention, vehicle detection for detecting the presence or absence of a vehicle on the road by analyzing image data of a predetermined range on the road obtained by a surveillance camera at night is performed. In the vehicle detection method of the device, the monitoring camera is an image storage type monitoring camera, and a tail lamp of the vehicle extracted by binarizing image data for one screen obtained by a slow shutter of the monitoring camera. After obtaining the coordinates of the end point of the trajectory image in the vehicle traveling direction, it is determined whether or not the coordinates include the coordinate value of one side forming the one screen, and when the coordinate value of the one side is included and the end point is included. When the number of trails of the tail lamps and their intervals satisfy the vehicle condition, there is a method of detecting the vehicle by regarding the trail of the tail lamp having this end point as the trajectory of the vehicle appearing in the one screen. That.

According to a second aspect of the invention, after obtaining the coordinate values of the end points in the vehicle traveling direction of the trajectory image of the tail lamp of the vehicle,
When it is determined whether or not this coordinate value includes the coordinate value of one side that constitutes the one screen, the locus of the tail lamp having an end point that does not include the coordinate value of the one side is displayed in the one screen of the vehicle having the tail lamp. It is considered as a continuous locus after the appearance of the vehicle and the continuity with the locus of the tail lamp when the vehicle appears in the one screen is confirmed.

The invention according to claim 3 has a method for calculating the speed of a vehicle having the tail lamp from the length of the trajectory of the tail lamp having an end point not including the coordinate value of the one side and the slow shutter speed of the surveillance camera. .

According to a fourth aspect of the present invention, in a vehicle detection device for detecting the presence or absence of a vehicle on the road by analyzing image data of a predetermined range on the road obtained by a surveillance camera at night, an image storage type Of the surveillance camera, binarizing means for binarizing one screen of image data obtained by the slow shutter of the image storage type surveillance camera, and image data binarized by the binarizing means. Coordinate acquisition means for obtaining the coordinates of the end points in the vehicle traveling direction of the trail image of the tail lamp of the vehicle, and whether the coordinate values of the end points obtained by the coordinate acquisition means include the coordinate values of one side constituting the one screen. First determining means for determining, and when the first determining means determines that the coordinate value of the end point includes the coordinate value of the one side, the number of trails of the tail lamp having the end point and the interval thereof. The second determination means for determining whether or not the vehicle condition is satisfied, and if the second determination means determines that the number of trails of the tail lamps having the end points and the intervals thereof satisfy the vehicle condition, the second determination means has the end point. A vehicle detection unit that detects a vehicle by regarding the trajectory of the tail lamp as the trajectory of the vehicle appearing in the one screen is provided.

According to a fifth aspect of the present invention, when the first determining means determines that the coordinate value does not include the coordinate value of one side forming the one screen, the tail lamp locus having the end point is set to the tail lamp trajectory. Considering as a continuous locus of the vehicle with the appearance in the one screen, a confirmation means for confirming the continuity of the locus of the tail lamp and the locus of the tail lamp when the vehicle appears in the one screen. Is provided.

According to a sixth aspect of the present invention, the speed of the vehicle having the tail lamp is calculated from the length of the trajectory of the tail lamp having the end point indicated by the coordinate value not including the coordinate value of the one side and the slow shutter speed of the surveillance camera. It has a configuration provided with a speed calculation means for calculating.

[0014]

According to the vehicle detection method of the present invention, the surveillance camera is an image accumulation type surveillance camera, and the image data for one screen obtained by the slow shutter of the surveillance camera is binarized. After obtaining the coordinates of the end point in the vehicle traveling direction of the extracted trail image of the tail lamp of the vehicle, it is determined whether or not the coordinates include the coordinate value of one side forming the one screen, and the coordinate value of the one side is included. In this case, when the number of trails of the tail lamp having this end point and the interval thereof satisfy the vehicle condition, the trail of the tail lamp having this end point is regarded as the appearance trajectory of the vehicle in the one screen, and the vehicle is detected. Therefore, at night, even if a part of multiple vehicles overlap and are displayed on the surveillance camera, each vehicle is detected when it appears on the screen of each vehicle, so multiple vehicles are separated and accurately detected. To do Can.

According to the vehicle detection method of the invention of claim 2,
After obtaining the coordinate value of the end point in the vehicle traveling direction of the trajectory image of the tail lamp of the vehicle, it is determined whether or not the coordinate value includes the coordinate value of one side forming the one screen, the coordinate value of the one side. The trail of the tail lamp having an end point that does not include is regarded as a continuous trajectory after the appearance of the vehicle having this tail lamp in the one screen, and the trajectory of the tail lamp when the vehicle appears in the one screen. Since the continuity is confirmed, at night, even if a part of multiple vehicles are overlapped and displayed on the surveillance camera, and the trail of the tail lamp of the vehicle in front becomes one, it may be mistaken as a vehicle, Alternatively, it is possible to prevent the vehicle from being mistakenly recognized as a single vehicle and improve the vehicle detection accuracy.

In the vehicle detection method according to the invention of claim 3,
Since the speed of the vehicle having the tail lamp is detected from the length of the trajectory of the tail lamp having an end point that does not include the coordinate value of the one side and the slow shutter speed of the surveillance camera, the speed of the vehicle can be detected.

In the vehicle detection device of the invention of claim 4,
The image storage type surveillance camera and the binarizing means binarize one screen of image data obtained by the slow shutter of the image storage type surveillance camera. The coordinate acquisition means obtains the coordinates of the end point in the vehicle traveling direction of the trajectory image of the tail lamp of the vehicle in the image data binarized by the binarization means. The first determination means determines whether the coordinate value of the end point obtained by the coordinate acquisition means includes the coordinate value of one side forming the one screen. The second determination means, when the first determination means determines that the coordinate value of the end point includes the coordinate value of the one side, the number of trails of the tail lamp having the end point and the interval thereof satisfy the vehicle condition. Determine whether or not. When the vehicle determination means determines by the second determination means that the number of trails of the tail lamp having the end points and the intervals thereof satisfy vehicle conditions, the trail of the tail lamps having the end points is displayed within the one screen of the vehicle. The vehicle is detected by considering it as an appearance locus. As a result, at night, even if a part of multiple vehicles overlap and are projected on the surveillance camera,
Since each vehicle is detected when each vehicle appears on the screen, a plurality of vehicles can be separated and accurately detected.

In the vehicle detection device of the invention of claim 5,
The confirmation means determines that the coordinate value is 1
When it is determined that the coordinate value of one side forming the screen is not included, the trajectory of the tail lamp having this end point is regarded as a continuous trajectory after the appearance of the vehicle having this tail lamp in the one screen, and the tail lamp Of the tail lamp when the vehicle appears in the one screen is confirmed. As a result, at night, even if a part of multiple vehicles overlap and are displayed on the surveillance camera, and the trail of the tail lamp of the vehicle in front becomes one, it is mistakenly recognized as not a vehicle or as a single vehicle. It is possible to improve the vehicle detection accuracy by preventing this.

In the vehicle detection device of the invention of claim 6,
The speed calculation means calculates the speed of the vehicle having the tail lamp from the length of the trajectory of the tail lamp having the end point indicated by the coordinate value not including the coordinate value of the one side and the slow shutter speed of the surveillance camera. Thereby, the speed of the vehicle can be detected.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a vehicle detection device of the present invention using the vehicle detection method of the present invention. 1
Is a surveillance camera composed of a video camera or the like for picking up a predetermined area on the road, and is an image storage type surveillance camera capable of accumulating and outputting an image signal of about 10 fields by a slow shutter. 2 is a CPU for performing various image processing related to vehicle detection, 3 is an A / D converter for converting an analog image signal obtained from the surveillance camera 1 into a digital image signal, and 4 is a predetermined image of the current image captured by the surveillance camera 1. A feature extraction circuit for binarizing with a threshold value to extract a feature portion, 5
Is an image memory for storing the binarized image data extracted from the feature extraction circuit 4 or the image data processed by the CPU 2, 6 is an external memory for storing vehicle measurement data by the CPU 2, and 7 is an image memory 5 A D / A converter for converting the digital image data read from the device into analog image data, a display unit 8 for displaying various image data relating to vehicle detection, and an input unit 9 for inputting various commands from an operator to the CPU 2. Reference numeral 10 is an internal memory that stores XY coordinate values of line data with time as a parameter.

Next, the operation of this embodiment will be described. As shown in FIG. 2, the storage-type monitoring camera 1 is arranged downward so that the vehicles 22 and 23 that are detection target objects traveling on a road 21 can be imaged. However, the vehicles 22 and 23 are assumed to travel in the direction of the arrow in the figure. FIG. 3 is an example of a screen in which a predetermined range on the road is imaged by the surveillance camera 1 arranged as described above. Reference numerals 31 and 33 denote left and right shoulders on both sides of the road. A left lane A and a right lane B are provided between these shoulders, and both lanes are separated by a lane boundary line 32. The surveillance camera 1 is used to detect a vehicle traveling in both lanes.

The image picked up by the storage type surveillance camera 1 is stored in a slow shutter for, for example, about 10 fields, and then converted into an analog image signal (luminance signal) for one screen and output to the A / D converter 3. To be done. The operator issues a vehicle detection operation command from the input unit 19 to the CPU 2.
As a result, the CPU 2 causes the feature extraction circuit 4 to capture the stored analog image signal output from the A / D converter 3 and starts the vehicle detection operation.

The accumulated image signal digitized by the A / D converter 3 is controlled by the CPU 2 to be monitored by the surveillance camera 1.
In accordance with the output timing of the image signal (same as the slow shutter timing), the accumulated image data taken into the feature extraction circuit 4 is binarized by a predetermined threshold value, and
The binarized image data is stored in the image memory 5. The binarized image data for one screen stored in the image memory 5 is subsequently processed by the CPU 2 for vehicle detection.

FIG. 4 is a flow chart showing a process for detecting a vehicle by the CPU 2 described above. The CPU 2 fetches one screen of the binarized image signal stored in the image memory 5 in step 401. Here, since this example is used at night, the binarized image data stored in the image memory 5 becomes an image as shown in FIG. 5, and the locus of the tail lamp of the vehicle as shown by 51 and 52. An image (typically equivalent to a streamed image) is stored. Next, in step 402, the CPU 2 reads the trajectory images 51, 5 of the tail lamps read from the image memory 5.
The XY coordinate values of the end points a, b, c, h, d, h, h, g, and d in the Y coordinate direction of 2 are obtained, and the obtained XY coordinate values are stored in the internal memory 10 as the tail lamp trajectory.

After that, the CPU 2 proceeds to step 403 to determine whether the Y coordinate value of the end point of the locus of each tail lamp stored in the internal memory 10 is the end portion ya of the screen as shown in FIG. 5 indicates that the locus of the corresponding tail lamp is the locus appearing on the screen as shown in 52 of FIG. 5, or 5 of FIG. 5 after appearing on the screen.
It is determined whether or not the continuous locus is as shown in 1. However, the appearance locus is defined as a locus image of the tail lamp whose end point is the Y coordinate value ya of the screen edge as described above, and the other locus images are defined as continuous loci.

For example, in the example of FIG. 5, since the Y coordinate value of the rear end point of the tail lamp locus 52 is ya, it is judged that the tail lamp locus 52 is the appearance locus.
It is distributed to the appearance locus side. On the other hand, since the Y coordinate value of the front or rear end point of the tail lamp locus 51 is not ya, it is determined to be a continuous locus and is distributed to the continuous locus side. If the CPU 2 determines that the trail of the tail lamp of interest is the appearance trajectory by the above determination processing, the CPU 2 proceeds to step 405 to perform vehicle determination processing, and if not, the step 406.
Proceed to and perform vehicle tracking processing.

The CPU 2 proceeds to step 405 and considers whether or not the trail of the tail lamps distributed to the appearance locus side is really the trail of the tail lamps of the vehicle, and whether or not the intervals of the trajectory images and the number thereof satisfy the vehicle condition. If, as a result, the vehicle conditions are satisfied, the end point coordinate value of the trail of the tail lamp is detected as one vehicle, the processing is stored in the internal memory 10, and the vehicle detection processing is ended. To do.

Here, in the case of a four-wheeled vehicle, there are one pair of tail lamp locus images, and the intervals are within a predetermined range. Further, in the case of a single vehicle, since the image of the trail of the tail lamp is one, whether the vehicle is a vehicle or not based on the interval and the number of the trails of the tail lamp described above, and if it is a vehicle, what is the vehicle type (whether a four-wheel vehicle or a single vehicle), etc. You can know. In addition, since the size of the tail lamp interval and the size of the vehicle are almost proportional to each other, the size of the vehicle can be determined from the interval of the trail of the tail lamp. May be linked to the end point coordinates stored as having detected the vehicle and stored in the external memory 6.

On the other hand, in step 406, the CPU 2 searches the continuous loci stored in the internal memory 10 for a locus previously determined to be a vehicle, and the loci of both are overlapping or continuous. It is checked whether or not there is such a condition, and if there is one satisfying such a condition, these loci are recognized as the same vehicle locus, and this recognition information is linked to the corresponding end point coordinates of the internal memory 10 and stored.

Here, the above-described method of searching for a continuous continuous trajectory on a trajectory previously determined as a vehicle is a continuous trajectory with a straight line connecting the end points of the trajectory previously determined as a vehicle in the Y coordinate direction. It is performed by checking whether or not a part of the straight line connecting the end points of the locus determined to be the same overlaps.If they do, the continuous locus continues to the appearance locus previously determined to be the vehicle. It is determined that or,
The above-described method for recognizing the trajectory of the same vehicle is also performed by verifying whether or not a part of the straight lines connecting the end points of the trajectory determined to be the vehicle as described above overlap or are continuous.

FIG. 6 shows a situation in which two vehicles are displayed in an overlapping manner on the screen of the surveillance camera 1. The following vehicle 63 is traveling close to the vehicle 66 ahead. Therefore, when viewed from the position of the surveillance camera 1, the vehicle 66 and the vehicle 63
Of the tail lamp of the vehicle 66 can be seen only on the right side of 64. In such a case, the binarized image data stored in the image memory 5 shown in FIG. 1 is as shown in FIG. 7, and it is impossible to know what 71 is as it is.

However, as described above, in this example, from the stage of FIG.
Since the continuity of the trail of the tail lamp is determined as described in the processing of 1., the trail 71 of the tail lamp of the binarized image data as shown in FIG. 7 is also mistakenly recognized as a single vehicle, Or, without misidentifying that it is not a vehicle, 2
It can be correctly separated and judged as a single vehicle.

According to this embodiment, the image data obtained by the slow shutter of the storage-type surveillance camera 1 is binarized by the feature extraction circuit 4 to obtain the locus image of the tail lamp of the vehicle, and then the locus image of this locus image is obtained. Only when the trajectory image obtained by analyzing the end points is the appearance image and when the trajectory satisfies the vehicle condition, the trajectory image is detected as the tail lamp of the vehicle, and the XY of the end points is detected. Since the coordinate values are stored in the internal memory 10, even when two vehicles appear to overlap with each other as shown in FIG. 6, it is possible to accurately detect the number of vehicles passing through without erroneously detecting this as one vehicle. can do.

Further, since the storage-type surveillance camera 1 outputs image data for one screen of, for example, 10 fields by the slow shutter, the CPU 2 can also perform the subsequent image processing at the shutter interval of the slow shutter described above. Well, CP
Since the processing speed of U does not have to be high, a high-speed CPU
It is not necessary to use, and the device can be made inexpensive.

The CPU 2 detects the speed of this vehicle by measuring the length of the tail lamp trajectory from the coordinate values of the front and rear end points of the tail lamp continuous trajectory 51 shown in FIG. The speed detected by linking to the coordinate values of the end points of the vehicle can be stored in the external memory 6 and used for various statistical processes.

[0036]

As described above, according to the invention of claim 1 or 4 of the present invention, even if a part of a plurality of vehicles are overlapped and displayed on a surveillance camera at night, the vehicles are separated from each other. Can be detected and the device can be made inexpensive. According to the invention of claim 2 or 5, it is possible to further improve the detection accuracy by separating the vehicles.
According to the invention of claim 3 or 6, the speed of the vehicle can be detected.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a vehicle detection device of the present invention.

FIG. 2 is a diagram showing an arrangement example of a surveillance camera 1 shown in FIG. 1 on a road.

FIG. 3 is a diagram showing an imaging range of the surveillance camera shown in FIG.

FIG. 4 is a flowchart showing a vehicle detection operation of the device shown in FIG.

5 is a diagram showing an example of binarized image data stored in the image memory shown in FIG.

FIG. 6 is a diagram showing an arrangement example of two vehicles running in close proximity to the surveillance camera 1.

7 is a diagram showing another example of binarized image data stored in the image memory shown in FIG.

8 is a diagram showing still another example of binarized image data stored in the image memory shown in FIG. 1. FIG.

[Explanation of symbols]

1 ... Surveillance camera 2 ... CPU 3 ... A / D converter 4 ... Feature extraction circuit 5 ... Image memory 6 ... External memory 7 ... D / A converter 8 ... Display unit 9 ... Input unit 10 ... Internal memory

Claims (6)

[Claims] 1. A vehicle detection method of a vehicle detection device for detecting the presence or absence of a vehicle on the road by analyzing image data of a predetermined range on the road captured by a surveillance camera at night. The surveillance camera is an image storage type surveillance camera, and is obtained by the slow shutter of this surveillance camera.
After obtaining the coordinates of the end points in the vehicle traveling direction of the trajectory image of the tail lamp of the vehicle extracted by binarizing the image data for the screen, whether these coordinates include the coordinate values of one side forming one screen. If it is determined that the coordinate value of one side is included and the number of loci of tail lamps having this end point and the intervals thereof satisfy vehicle conditions, the locus of tail lamps having this end point is displayed within the one screen of the vehicle. A vehicle detection method characterized by detecting a vehicle as an appearance locus. 2. After obtaining the coordinate values of the end points in the vehicle traveling direction of the trail image of the tail lamp of the vehicle, it is determined whether or not the coordinate values include the coordinate values of one side constituting the one screen, The trajectory of the tail lamp having an end point that does not include the coordinate value of the one side is regarded as a continuous trajectory of the vehicle having this tail lamp after appearing in the one screen, and when the vehicle appears in the one screen. The vehicle detection method according to claim 1, wherein continuity with a locus of the tail lamp is confirmed. 3. The speed of the vehicle having the tail lamp is calculated from the length of the trajectory of the tail lamp having an end point that does not include the coordinate value of the one side and the slow shutter speed of the surveillance camera. The vehicle detection method described. 4. A vehicle detection device for detecting the presence or absence of a vehicle on the road by analyzing image data of a predetermined range on the road captured by a surveillance camera at night time,
An image storage type surveillance camera, a binarizing means for binarizing image data for one screen obtained by a slow shutter of the image storage type surveillance camera, and an image binarized by the binarizing means. Coordinate acquisition means for obtaining the coordinates of the end points in the vehicle traveling direction of the image of the tail lamp trajectory of the vehicle in the data, and the coordinate values of the end points obtained by the coordinate acquisition means include the coordinate values of one side forming the one screen. First determining means for determining whether or not it is determined that the coordinate value of the end point includes the coordinate value of the one side by the first determining means,
Second judging means for judging whether or not the number of trails of tail lamps having the end points and their intervals satisfy vehicle conditions, and the number of trails of the tail lamps having the end points and their intervals by the second judging means When it is determined that the condition is satisfied, the vehicle detection means further includes vehicle detection means for detecting the vehicle by regarding the trajectory of the tail lamp having the endpoint as the trajectory of the vehicle appearing in the one screen. apparatus. 5. When the first determining means determines that the coordinate value does not include the coordinate value of one side forming the one screen, the locus of the tail lamp having the end point is determined by the vehicle having the tail lamp. Considering it as a continuous trajectory after appearing in the one screen, a confirmation means for confirming continuity between the trajectory of the tail lamp and the trajectory of the tail lamp when the vehicle appears in the one screen is provided. The vehicle detection device according to claim 4, which is characterized in that. 6. A speed calculation means for calculating the speed of the vehicle having the tail lamp from the length of the trajectory of the tail lamp having an end point indicated by coordinate values not including the coordinate value of the one side and the slow shutter speed of the surveillance camera. The vehicle detection device according to claim 5, further comprising: