JPH11232465A - Mobile object detecting device and program recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect a mobile object at high speed and with high precision in a mobile object detecting device which is mounted on a mobile object and to detects the external mobile object by processing an external picture which is read by means of a picture reading device. SOLUTION: This device 1 is provided with a deriving means 14 for deriving the moving vector of a partial picture area by calculating correlation between the picture to be segmented from the partial picture area which divides an input picture and the input picture to be sampled before or after the input picture, a specifying means 15 for specifying the partial picture area formed by the mobile object from the mobile vector which is derived by the deriving means 14, and a detecting means 17 for detecting the position of the mobile object by calculating correlation between the picture of the mobile object to be segmented from the partial picture area which is specified by the specifying means and the input picture.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used for a moving object detecting device mounted on a moving object to detect an external moving object by processing an external image read by an image reading device, and to realize the moving object detecting device. The present invention relates to a program recording medium on which a program is stored, and more particularly, to a moving object detection device capable of detecting a moving object with high speed and high accuracy, and a program recording medium on which a program used for realizing the moving object is stored.

[0002] In order to realize safe driving of a car, there is a demand for the development of a device for automatically detecting a car running ahead of a course from an image taken by a television camera mounted on the car. In order to put this device to practical use, it is necessary to be able to automatically detect a car traveling ahead of the route with high speed and high accuracy.

[0003]

2. Description of the Related Art An apparatus for automatically detecting a car traveling ahead of a route from an image taken by a television camera or the like mounted on the car has conventionally adopted a configuration in which template images of various vehicles are registered in advance. In addition, a configuration is adopted in which a vehicle traveling ahead of the path is detected by pattern matching.

FIG. 27 shows the configuration of a conventional apparatus employing this configuration. In addition, an image captured by a TV camera mounted on a car is differentiated to generate an edge image, and a region where vertical and horizontal edge patterns are densely extracted is detected to detect a car traveling ahead of the path. Some of them adopt a configuration that performs the following.

FIG. 28 shows the configuration of a prior art device employing this configuration.

[0006]

However, according to the prior art in which a template image of various vehicles is registered in advance and a vehicle traveling ahead of the path is detected by pattern matching, the registered vehicle template Since the number of images becomes enormous, there is a problem that the device scale increases and the processing time increases.

Further, according to the conventional technique of detecting a car traveling ahead of a course by generating an edge image by differentiating the image and extracting an area where edge patterns in a vertical direction and a horizontal direction are dense, a road is considered. Since the above various notations form an edge pattern, there is a problem that misrecognition often occurs.

The present invention has been made in view of the above circumstances, and has a configuration in which an external moving object is detected by processing an external image read by an image reading device mounted on a moving body. It is therefore an object of the present invention to provide a new moving object detecting device capable of detecting a moving object with high speed and high accuracy, and to provide a new program recording medium storing a program used for realizing the moving object detecting device.

[0009]

FIG. 1 and FIG. 2 show the principle configuration of the present invention. In the figure, reference numeral 1 denotes a moving object detection device equipped with the present invention, which is mounted on a moving body such as a car,
By processing external images read by the image reading device,
It detects an external moving object.

A moving object detecting device 1 according to the present invention, whose principle configuration is shown in FIG. 1, comprises an image reading device 10 and an image memory 11.
-1,2, switching means 12, deriving means 14, and specifying means 1
5, a tracking image memory 16, a detecting unit 17, a setting unit 18, a determining unit 19, and an updating unit 20.

On the other hand, a moving object detecting apparatus 1 according to the present invention, whose principle configuration is shown in FIG. 2, comprises an image reading apparatus 10, image memories 11-1 and 2, switching means 12, correlation calculating means 13,
Deriving means 14, specifying means 15, and tracking image memory 1
6, detecting means 17, setting means 18, and determining means 19
And updating means 20.

The image reading apparatus 10 reads an image of an external moving object. The image memories 11-1 and 2 store an image of a moving object read by the image reading device 10. The switching unit 12 alternately selects the image memories 11-1 and 2 and stores the image of the moving object read by the image reading device 10 in the selected image memories 11-1 and 2.

The correlation calculating means 13 provided in the moving object detecting apparatus 1 of the present invention whose principle configuration is shown in FIG. 2 calculates a correlation between two images designated as processing targets. The deriving unit 14 derives a movement vector of a partial image area that partitions the input image. The specifying unit 15 specifies a partial image area formed by the moving object from the moving vector derived by the deriving unit 14. The tracking image memory 16 includes the identifying unit 1
5 stores the image of the moving object to be cut out of the specified partial image area.

The detecting means 17 detects the position of the moving object. When the partial image region specified by the specifying unit 15 does not overlap the position of the moving object detected by the detecting unit 17, the setting unit 18 newly detects the image of the moving object cut out from the partial image region by the detecting unit 17. Set as target.

When the position of the moving object detected by the detecting means 17 and the partial image area specified by the specifying means 15 indicate a specified number of overlaps, the determining means 19 determines the form of movement of the moving object. When the position of the moving object detected by the detecting unit 17 and the partial image region specified by the specifying unit 15 indicate an overlap, the updating unit 20 sets the detection target of the detecting unit 17 according to the partial image region. The image of the moving object is updated, and the position of the moving object detected by the detecting unit 17 is updated.

Here, the functions of the moving object detecting device 1 of the present invention whose principle configuration is shown in FIGS. 1 and 2 can be realized by a program. In this case, the program is stored in a floppy disk or the like. The present invention is realized by being stored in a disk or the like of a server or the like, and then being installed in the moving object detection device 1 and operating on a memory.

In the moving object detecting apparatus 1 according to the present invention, which illustrates the principle configuration of FIG.
Alternately selects two image memories 11-1 and 2 and stores the latest image read by the image reading device 10 in the selected image memory 11-1 and 2.

In response to the input images stored in the image memories 11-1 and 11-2, the deriving means 14 samples an image from which a partial image area that divides the input image is cut out and samples the input image before or after the input image. By calculating a correlation with the input image, a motion vector of a partial image area that divides the input image is derived, and in response to this, the specifying unit 15 specifies a motion vector indicating a motion different from the background, A partial image area formed by the moving object is specified.

In response to the partial image area specified by the specifying means 15, when the specified partial image area does not overlap with the position of the moving object detected by the detecting means 17, the setting means 18 sets The image of the moving object to be cut out is stored in the tracking image memory 16.

Upon receiving the image of the moving object stored in the tracking image memory 16, the detecting means 17 synchronizes with, for example, the time when the specifying means 15 specifies the moving vector, and stores it in the tracking image memory 16. By calculating the correlation between the stored image of the moving object and the latest input image stored in the image memories 11-1 and 2 (the calculation may not be the same as the correlation calculation performed by the deriving unit 14), The position of the moving object is detected.

On the other hand, in the moving object detecting device 1 of the present invention, which illustrates the principle configuration of FIG. 2 configured as described above, the switching means 12 alternately selects two image memories 11-1, 2 provided. Then, the latest image read by the image reading device 10 is stored in the selected image memory 11-1,2.

In response to the input images stored in the image memories 11-1 and 11-2, the deriving means 14 samples an image from which a partial image area that divides the input image is cut out, and samples the input image before or after the input image. By activating the correlation calculation unit 13 with the input image as a processing target, a movement vector of a partial image region that divides the input image is derived. By specifying the vector, the partial image area formed by the moving object is specified.

In response to the partial image area specified by the specifying means 15, when the specified partial image area does not overlap the position of the moving object detected by the detecting means 17, the setting means 18 The image of the moving object to be cut out is stored in the tracking image memory 16.

Upon receiving the image of the moving object stored in the tracking image memory 16, the detecting means 17 synchronizes with, for example, the time when the specifying means 15 specifies the moving vector, and stores it in the tracking image memory 16. The position of the moving object is detected by activating the correlation calculating means 13 with the stored moving object image and the latest input image stored in the image memories 11-1 and 11 being processed.

In the moving object detecting device 1 of the present invention, which illustrates the principle configuration of FIGS. 1 and 2 configured as described above, when this processing configuration is adopted, the specifying unit 15 By extracting a motion vector that does not show a motion from among the motion vectors that move, a partial image area formed by a moving object that moves in the front can be specified. By extracting an object that indicates a movement in the direction, a partial image area formed by a moving object that moves in a manner interrupting forward is specified.

The determining means 19 determines the form of movement of the moving object when the position of the moving object detected by the detecting means 17 and the partial image area specified by the specifying means 15 overlap by a specified number of times. . For example, it is determined that the moving object is a moving object moving ahead.

Further, when the position of the moving object detected by the detecting means 17 and the partial image area specified by the specifying means 15 indicate an overlap, the updating means 20 detects the position of the moving object in accordance with the partial image area. The image of the moving object to be detected is updated, and the position of the moving object detected by the detecting means 17 is updated. By this updating process, it is possible to cope with a case where the image of the moving object stored in the tracking image memory 16 becomes inappropriate due to a change in the distance to the moving object, for example.

As described above, the moving object detecting apparatus 1 of the present invention
Then, the moving vector in the image read by the image reading device 10 is detected, and a moving vector indicating a motion different from the background moving vector is specified from the moving vector. By adopting a configuration for detecting the position of the moving object from the correlation between the image of the moving image and the image read by the image reading device 10, the position of the moving object can be detected without using a template image, and with small hardware and The moving object can be detected at a high speed, and the position of the moving object can be detected without using an edge pattern, so that the position of the moving object can be detected with high accuracy.

[0029]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to an embodiment applied to a front vehicle detecting device which is mounted on a vehicle and executes a process of detecting a vehicle running ahead.

FIG. 3 shows an embodiment of a forward vehicle detecting device 30 provided with the present invention. The front vehicle detection device 30 shown in FIG. 3 is mounted on a vehicle and executes a process of detecting a vehicle traveling ahead, and includes a TV camera 31 and
A / D converter 32, search image memory 33, first reference image memory 34, second reference image memory 35, input switch 36, first output switch 37, tracking image A memory 38, a second output switch 39, and a correlation calculator 4
0, the correlation peak position detector 41, and the address generator 4
2 and a control computer 43.

The TV camera 31 reads an image in front of the vehicle. The A / D converter 32 converts an input image read by the TV camera 31 from an analog signal to a digital signal. The search image memory 33 stores the input image converted into a digital signal by the A / D converter 32.
The first and second reference image memories 34 and 35 store the input images converted into digital signals by the A / D converter 32.

The input switch 36 selects one of the first reference image memory 34 and the second reference image memory 35 as a storage destination of the input image. First output switch 37
Selects one of an input image read from the first reference image memory 34 and an input image read from the second reference image memory 35 and outputs the selected image. The tracking image memory 38 stores the detected image of the preceding vehicle. Second
The output switch 39 selects and outputs either the input image output from the first output switch 37 or the image of the preceding vehicle read from the tracking image memory 38.

The correlation calculator 40 is composed of hardware and calculates and outputs a correlation value between the input image read from the search image memory 33 and the image output from the second output switch 39. The correlation peak position detector 41 is configured by hardware, and detects and outputs the peak position of the correlation value output from the correlation calculator 40. Address generator 4
2 is a search image memory 33 / first reference image memory 34
An address signal to be applied to the second reference image memory 35 / tracking image memory 38 is generated.

The control computer 43 receives the peak position detected by the correlation peak position detector 41 as an input, and according to software, an input switch 36 / first output switch 37 / second output switch 39 / address generator. The control unit 42 performs a process of detecting a vehicle traveling in front of the vehicle on which the vehicle is mounted.

As described above, the correlation calculator 40 calculates and outputs a correlation value between the input image read from the search image memory 33 and the image output from the second output switch 39. On the other hand, the second output switch 39
Of the preceding vehicle read out from the tracking image memory 38 and output.

Accordingly, the correlation calculator 40 calculates the input image read from the search image memory 33 and the reference image memory 3
A process for calculating and outputting a correlation value with an input image read from the image storage device 4, 35, or a correlation value between an input image read from the search image memory 33 and an image of a preceding vehicle read from the tracking image memory 38. Is calculated and output.

Here, the calculation of the correlation value between the input image read from the search image memory 33 and the input images read from the reference image memories 34 and 35 will be described later. This is performed in order to derive the movement vector of the minute area, and is performed in units of the minute area. This small area is shown in FIG.
As shown in (1), the input image read out from the reference image memories 34 and 35 is defined by, for example, partitioning into 8 × 8 pixels.

The correlation calculator 40 calculates the correlation value between the input image read from the search image memory 33 and the minute area of the input image read from the reference image memories 34 and 35, as shown in FIG. As shown in the figure, the address generator 42 sets the origin position O1 of the minute area of the input image stored in the reference image memories 34 and 35, and sets the origin position O1 in the input image stored in the search image memory 33. Given the origin position O2 of the correlation operation area set to surround the micro area, when the pixel values are read from the search image memory 33 and the reference image memories 34 and 35,
As shown in FIG. 6, for example, D _{m, n} = ΣΣ | X _{i, j} −Y _{i + m, j + n} while changing the offset amount (m, n) from the origin position O2 of the correlation calculation area. Where で is obtained by calculating the correlation value D _mn at each offset amount (m, n) according to the formula for calculating the sum of i and j.

Here, X _{i, j} shown in this calculation expression is a pixel value of a minute area of the input image stored in the reference image memory 34, 35, and Y _{i + m, j + n} is a search image. The pixel values of the input image stored in the memory 33.

Further, the correlation calculator 40 stores the input image read from the search image memory 33 and the tracking image memory 3.
When the calculation of the correlation value with the image of the preceding vehicle read out from the address generator 8 is executed, as shown in FIG.
2, the origin position O3 of the car image stored in the tracking image memory 38 and the correlation calculation area set in the input image stored in the search image memory 33 so as to surround the car image. When the pixel value is read from the search image memory 33 and the tracking image memory 38 by giving the origin position O4, as shown in FIG. 8, the offset amount (m, n) of the correlation operation area from the origin position O4 is obtained. ) Is changed, for example, D _{m, n} = ΣΣ | X _{i, j} −Y _{i + m, j + n} | where ΣΣ is the offset amount (m, This is performed by _{obtaining the} correlation value D _mn in n).

Here, X _{i, j} shown in this calculation expression is a pixel value of the car image stored in the tracking image memory 38, and Y
_{i + m and j + n} are pixel values of the input image stored in the search image memory 33.

The correlation value calculated by this calculation formula shows the smallest value at the place where the correlation becomes highest.
For example, if the example illustrated in FIG. 9 is used, when the same image as the image of the minute area is located at an offset amount of “m = 5, n = 4” from the origin position O1 of the correlation operation area, The correlation value shows the smallest value at the position of “m = 5, n = 4”.

As described above, the correlation peak position detector 4
1 is input with the correlation value output from the correlation calculator 40,
A process of detecting and outputting the peak position of the correlation value output from the correlation calculator 40 is performed.

That is, the correlation peak position detector 41
When the correlation calculator 40 enters the correlation calculation process by giving the origin position from the address generator 42 and sequentially outputs the correlation value at each offset amount, the peak position of the correlation value is detected and output. . In the case of the example shown in FIG. 9, a peak position "m = 5, n = 4" is detected and output.

FIGS. 10 and 11 show one embodiment of the processing flow executed by the control computer 43. FIG. Next, the present invention will be described in detail according to this processing flow. When a request to detect a vehicle traveling ahead of the on-board vehicle is issued, the control computer 43 first controls the input switch 36 in step 1 as shown in the processing flow of FIGS. 10 and 11. Thus, for the first reference image memory 34,
As shown in FIG. 12, which is read by the TV camera 31, the first output switch 37 is switched to the first reference image memory 34.

Subsequently, in step 2, the second output switch 39 is switched to the side of the reference image memories 34, 35, and in step 3, the input switch 36 is switched. At this point, the input switch 36 switches so as to select the second reference image memory 35. Then, step 4
Then, the image read by the TV camera 31 is input to the search image memory 33, and the TV camera 31 is input to the reference image memories 34 and 35 selected by the input switch 36.
Enter the image to be read.

As described above, at this time, the first reference image memory 34 has the input image stored in the second reference image memory 34 (the same as the input image stored in the search image memory 33). The input image of the previous frame is stored.

Subsequently, at step 5, the address generator 4
2 to the reference image memories 34 and 35 output from the second output switch 39 (to store the input image of the frame immediately before the input image stored in the search image memory 33). The origin position O1 of the minute area of the input image stored in the reference image memories 34 and 35 is sequentially given, and the search image memory 33 is provided with the correlation calculation area shown in FIG. By giving the origin position O2, the movement vector of each minute area of the input image stored in the reference image memories 34 and 35 is obtained.

That is, the reference image memories 34 and 35 output from the second output switch 39 are sequentially supplied with the origin position O1 of the minute area of the input image stored in the reference image memories 34 and 35. When the search operation memory 33 is provided with the origin position O2 of the correlation calculation area surrounding the micro area shown in FIG. 5, the correlation calculator 40 calculates the origin position O2 for each micro area according to the above-described processing. the correlation value D _mn are successively calculated in the offset amount from the (m, n), in response to this, the correlation peak position detector 41, for each minute region, for detecting the peak position of the correlation value D _mn.

The input image stored in the reference image memories 34 and 35 output from the second output switch 39 is the input image of the frame immediately before the input image stored in the search image memory 33, From this, the correlation peak position detected by the correlation peak position detector 41 at this time is determined by the movement destination of each minute area of the input image stored in the reference image memories 34 and 35 output from the second output switch 39. Is shown. Therefore, as shown in FIG. 13, according to the processing in step 5, the movement vector of each minute area of the input image is obtained.

In step 5, the movement vector of each minute area of the input image is obtained. Subsequently, in step 6, the movement vector of each minute area is compared with the movement vector of the background so that the movement vector is obtained. It is determined whether there is something other than the background. When the mounted vehicle is running, the background movement vector is, as shown in FIG.
1 shows a movement extending radially from a specific point (vanishing point) defined from the mounting position of No. 1. On the other hand, a vehicle traveling ahead shows a motion different from the background motion vector as shown in FIG. From this, by comparing the movement vector of each minute region with the movement vector of the background, it is determined whether or not there is something other than the background ahead.

Subsequently, in step 7, it is determined whether or not there is an area having a movement vector different from the background. When it is determined that there is an area having a movement vector different from the background, the process proceeds to step 8. By performing labeling processing on a minute region having a movement vector different from the background, a lump of minute regions having a movement vector different from the background is extracted. That is, as shown in FIG. 15, a block of minute regions (connected vertically and horizontally) filled with black in the figure having a movement vector different from the background is obtained by a known labeling process, so that a movement vector different from the background is obtained. It extracts the lump of the minute area having.

Subsequently, in step 9 (the processing flow in FIG. 11), it is determined whether or not any of the labeled lump has a size larger than a specified value, so that there is a car traveling ahead. Is detected. That is, among the labeled masses, those having a size not exceeding the specified size are determined to be noise, and by removing them, it is detected whether or not there is a car traveling ahead.

In step 9, when it is determined that there is a vehicle traveling ahead by judging that some of the labeled masses (images) have a size larger than a specified value, the process proceeds to step 10. The process proceeds to determine whether or not an image of the detected car is registered in the tracking image memory 38, and when it is determined that the image is registered, the process proceeds to step 11 and the image detected in step 9 is determined. Is determined to overlap with the image of the registered car.

In step 11, when it is determined that the image detected in step 9 does not overlap with the car image registered in the tracking image memory 38,
When it is determined that no detected car image is registered in the tracking image memory 38, the image detected in step 9 is determined as a newly detected car image, and the process proceeds to step 12. After the detected image is registered by copying it to the tracking image memory 38, step 16
Then, the first output switch 39 is switched, and the process returns to step 2 (the processing flow in FIG. 10).

Here, for the tracking image memory 38,
Instead of registering the image detected in step 9 as it is, a method of registering an image cut out from a circumscribed rectangle may be adopted. On the condition that the image does not completely overlap with the image of the car registered in the image memory 38 for tracking, a method is adopted in which a new image is not registered but is registered when the overlap is small. You may.

On the other hand, when it is determined in step 7 that there is no area having a movement vector different from the background, and when it is determined in step 9 that there is no vehicle traveling ahead, the movement of the mounted vehicle due to vibration or the like occurs. Proceeding to step 13 in consideration of the vector detection error, the tracking image memory 3
It is determined whether or not an image of the detected car has been registered in step 8. When it is determined that the image has not been registered, the process proceeds to step 16 without performing steps 14 and 15 described below. And the first output switch 3
9 and then returns to step 2 (the processing flow of FIG. 10).

When it is determined in step 13 that the detected car image is registered in the tracking image memory 38, the image detected in step 9 is stored in the tracking image memory 38 in step 11. When it is determined that the registered image overlaps with the registered image of the preceding vehicle, the process proceeds to step 14, where the second output switch 39 is set to the tracking image memory 38.
Switch to the side.

Subsequently, in step 15, the origin image O3 of the car image stored in the tracking image memory 38 is stored in the tracking image memory 38 via the address generator 42.
Is given to the search image memory 33, and the origin position O4 of the correlation calculation area surrounding the vehicle image shown in FIG. 7 is given to the new position of the vehicle image stored in the tracking image memory 38. Ask for.

That is, the tracking image memory 38 is provided with the origin position O3 of the vehicle image stored in the tracking image memory 38, and the search image memory 33 is provided with the surrounding image shown in FIG. Given the origin position O4 of the correlation calculation area shown, the correlation calculator 40 performs the offset amount (m,
The correlation value D _mn in n) is sequentially calculated, and in response _thereto ,
The correlation peak position detector 41 detects the peak position of the correlation value _Dmn .

The input image stored in the search image memory 33 is the latest image. Therefore, the correlation peak position detected by the correlation peak position detector 41 at this time is stored in the tracking image memory 38. The latest position of the car image is shown. From this, the latest position of the car image stored in the tracking image memory 38, that is, the latest position of the car, is obtained in accordance with the processing of step 15.

When the latest position of the vehicle traveling ahead is determined in step 15, the process proceeds to step 16, where the first output switch 39 is switched, and then the process proceeds to step 2 (the processing flow in FIG. 10). Go back.

As described above, when the control computer 43 issues a request to detect a vehicle traveling ahead of the mounted vehicle, FIG.
As shown in the time chart of FIG.
While the latest input image is stored in one of the first reference image memory 34 and the second reference image memory 35, and the input image one frame before is stored in the other. I will do it. Then, the latest input image stored in the search image memory 33 and the reference image memory 3
By calculating the correlation with the minute area of the input image one frame before the latest image stored in 4,35, the image of the car showing a motion different from the background is detected and stored in the tracking image memory 38. As well as calculating the correlation between the latest input image stored in the search image memory 33 and the image of the vehicle stored in the tracking image memory 38, the detected vehicle position is tracked. Process it.

Although the description of this processing flow is omitted, when the image of the vehicle once detected cannot be detected because the vehicle traveling ahead is out of sight, it is stored in the image memory 38 for tracking. The image of the car is deleted.

Of the vehicles traveling ahead, those that are particularly important for realizing the safe traveling of the on-board vehicles are the preceding vehicle traveling in the same lane ahead and the interrupt that interrupts the same lane. It is a car.

In order to detect the preceding vehicle traveling ahead in the same lane, the processing flow of FIG. 17 is executed instead of the processing flow of FIG. 11 of the processing flow executed in FIGS. Will be. Also, in order to detect an interrupted vehicle coming into the same lane, the processing flow of FIG. 18 is executed instead of the processing flow of FIG. 11 among the processing flows executed in FIG. 10 and FIG. Become.

In the processing flow of FIG. 17, between step 9 corresponding to step 9 of the processing flow of FIG. 11 and step 10 corresponding to step 10 of the processing flow of FIG. By providing a process called "step A" for determining whether or not the moving vector has a motion, the preceding vehicle traveling in the same lane is detected.

That is, as shown in FIG. 19, since the preceding vehicle traveling ahead in the same lane has a motion vector that does not move, in step A, the motion vector of the vehicle detected in step 9 does not move. Is determined, a preceding vehicle traveling ahead in the same lane is detected. The traveling attributes of the vehicle detected in this manner are stored in a memory (not shown) in association with the image of the vehicle.

In the processing flow of FIG. 18, step 9 corresponding to step 9 of the processing flow of FIG.
Step 10 corresponding to step 10 of the first processing flow
In between, a process called “Step B” is provided to determine whether the movement vector of the car detected in Step 9 indicates the horizontal direction and the inward direction.
It adopts a configuration that detects an interrupting vehicle that interrupts the same lane.

That is, as shown in FIG. 20, since the interrupted vehicle that interrupts in the same lane has a movement vector that is directed in the horizontal direction and inward, in step B, the vehicle detected in step 9 has By judging whether or not the movement vector indicates a horizontal direction and an inward direction, an interrupting vehicle that interrupts the same lane is detected. The traveling attributes of the vehicle detected in this manner are stored in a memory (not shown) in association with the image of the vehicle.

In the processing flow of FIGS. 10 and 11, a configuration is adopted in which if a vehicle is detected even once, it is determined that the vehicle is traveling ahead. It is preferable to adopt a configuration in which it is determined that the vehicle is traveling ahead on the condition that the number of times is detected.

FIGS. 21 and 22 show an embodiment of a processing flow executed when this configuration is adopted. That is, when this configuration is adopted, in step 12 corresponding to step 12 of the processing flow of FIG. 11, a newly detected image of the vehicle (in this stage, a candidate for a preceding vehicle) is copied to the tracking image memory 38. At this time, "0" is set in a counter for recording the number of times of detection, and OFF (displaying "unconfirmed") is set in a confirmation flag indicating the detection of the preceding vehicle.

On the other hand, step 11 in the processing flow of FIG.
When it is determined in step 11 that the detected image overlaps with the image of the preceding vehicle (candidate) registered in the tracking image memory 38, a newly provided step 1 is provided.
At 00, it is determined whether or not the confirmation flag is turned on, and when it is determined that the confirmation flag is turned on, step 14 (the processing flow of FIG. 22) corresponding to step 14 of the processing flow of FIG. Proceed to).

On the other hand, when it is determined that the confirmation flag is off, the process proceeds to step 101 where a new flag is provided.
When the value of the counter is incremented by one, and subsequently, in the newly provided step 102, it is determined whether or not the value of the counter has exceeded the prescribed value. Step 1 of the processing flow
The process proceeds to Step 14 corresponding to Step 4 (the processing flow in FIG. 22), and when it is determined that the value is equal to or greater than the specified value, the process proceeds to Step 103, which is newly provided. The process proceeds to step 14 (the processing flow in FIG. 22) corresponding to step 14.

As described above, in the processing flow of FIGS. 21 and 22, when the vehicle is detected a plurality of times in consideration of the vibration of the mounted vehicle, the vehicle traveling ahead is finally confirmed. Processing.

The positional relationship between the mounted vehicle and the preceding vehicle changes every moment. Even when the vehicle is traveling in a straight lane, the inter-vehicle distance changes due to the difference in speed between the two vehicles, and greatly varies while traveling in a curved lane as shown in FIG.

To cope with this, the processing flow of FIG. 23 is executed instead of the processing flow of FIG. 22 of the processing flows executed in FIG. 21 and FIG. In the processing flow of FIG. 23, the image of the car detected by the movement vector is between step 13 corresponding to step 13 of the processing flow of FIG. 22 and step 14 corresponding to step 14 of the processing flow of FIG. In accordance with the above, the image of the preceding vehicle (candidate) registered in the tracking image memory 38 overlapping with that is updated, and the process of “Step A” of updating the position of the preceding vehicle (candidate) is provided, thereby mounting the vehicle. The image and the position of the preceding vehicle (candidate) generated by the change in the positional relationship between the vehicle and the preceding vehicle are updated.

This updating process is performed by rewriting the image of the car registered in the tracking image memory 38 to the image of the car detected by the movement vector. This can be performed by rewriting the position of the car that is present to the position of the car detected by the movement vector. However, in consideration of the detection error of the movement vector, as shown in FIG. An interpolation position (interpolation weight is determined in advance) between the car image / position registered and the car image / position registered in the tracking image memory 38 is obtained and rewritten according to the interpolation position. It is preferable to employ several configurations.

In the embodiment shown in FIG. 3, a configuration having one correlation calculator 40 is adopted, and by using this correlation calculator 40, the movement of the minute area of the input image stored in the reference image memories 34 and 35 is performed. Although a configuration for obtaining a vector and obtaining a new position of a car image registered in the tracking image memory 38 is adopted, a correlator for obtaining a movement vector of a minute area of the input image stored in the reference image memories 34 and 35 is adopted. It is also possible to adopt a configuration in which a correlator for obtaining a new position of the car image registered in the tracking image memory 38 is separately prepared.

FIG. 26 shows an embodiment of the forward vehicle detecting device 30 according to the present invention having the above configuration. In this embodiment, a first image storing an input image read by the TV camera 31 is stored.
And second search image memories 33a and 33b, and an input image read from the first search image memory 33a and an image output from the first output switch 37 (inputs read from the reference image memories 34 and 35). Image), a first correlation calculator 40a for calculating a correlation with the input image read from the second search image memory 33b, and a correlation between the input image read from the tracking image memory 38 and the image of the preceding vehicle. The second correlation calculator 40b and the first correlation calculator 4b
0a, a first correlation peak position detector 41a for detecting a peak position of a correlation value output, and a second correlation calculator 40b.
And a second correlation peak position detector 41b for detecting the peak position of the correlation value outputted by In the case of following this embodiment, it is not necessary to provide the second output switch 39 provided in the embodiment of FIG.

According to this configuration, in the forward vehicle detecting device 30 of the present invention shown in FIG. 26, the first correlation calculator 40a is used to move the minute area of the input image stored in the reference image memories 34 and 35. Vector and second
Is processed to obtain a new position of the car image registered in the tracking image memory 38 using the correlation calculator 40b. According to the embodiment of FIG. 3, there is an advantage that the amount of hardware can be reduced. According to the embodiment of FIG. 26, the movement vector of the minute area of the input image stored in the reference image memories 34 and 35 is obtained. And the content of the correlation calculation for obtaining a new position of the car image registered in the tracking image memory 38 can be made different from each other.

The present invention has been described with reference to the illustrated embodiments.
The present invention is not limited to this. For example, in the embodiment, the present invention has been described by taking the application to the forward vehicle detecting device 30 as a specific example, but the present invention is not limited to the application to the forward vehicle detecting device 30.

In the embodiment, the correlation calculator 40 and the correlation peak position detector 41 are described as being constituted by hardware. However, these functions may be constituted by software.

[0084]

As described above, the moving object detecting apparatus according to the present invention detects a moving vector in an image read by an image reading apparatus, and detects a moving vector having a motion different from the background moving vector from the detected moving vectors. By specifying the image of the moving object and identifying the position of the moving object from the correlation between the image of the moving object and the image read by the image reading device, without using the template image It is now possible to detect the position of a moving object with small hardware and at high speed, and to detect the position of the moving object without using an edge pattern. Position can be detected.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a principle configuration diagram of the present invention.

FIG. 3 is an embodiment of the present invention.

FIG. 4 is an explanatory diagram of a minute area.

FIG. 5 is an explanatory diagram of a correlation calculator.

FIG. 6 is an explanatory diagram of a correlation calculator.

FIG. 7 is an explanatory diagram of a correlation calculator.

FIG. 8 is an explanatory diagram of a correlation calculator.

FIG. 9 is an explanatory diagram of a correlation operation.

FIG. 10 is an example of a processing flow executed by a control computer.

FIG. 11 is an example of a processing flow executed by a control computer.

FIG. 12 is an explanatory diagram of an input image.

FIG. 13 is an explanatory diagram of a movement vector calculation process.

FIG. 14 is an explanatory diagram of a movement vector of an input image.

FIG. 15 is an explanatory diagram of a labeling process.

FIG. 16 is a time chart of a correlation operation.

FIG. 17 is another embodiment of the processing flow executed by the control computer.

FIG. 18 is another example of the processing flow executed by the control computer.

FIG. 19 is an explanatory diagram of a movement vector of a vehicle traveling forward.

FIG. 20 is an explanatory diagram of a movement vector of a vehicle that is running while being interrupted.

FIG. 21 is another embodiment of the processing flow executed by the control computer.

FIG. 22 is another example of the processing flow executed by the control computer.

FIG. 23 is another embodiment of the processing flow executed by the control computer.

FIG. 24 is an explanatory diagram of a car traveling on a curve.

FIG. 25 is an explanatory diagram of a process of updating an image and a position of a car.

FIG. 26 is another embodiment of the present invention.

FIG. 27 is an explanatory diagram of a conventional technique.

FIG. 28 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 moving object detecting device 10 image reading device 11 image memory 12 switching means 13 correlation calculating means 14 deriving means 15 specifying means 16 tracking image memory 17 detecting means 18 setting means 19 determining means 20 updating means

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification symbol FI // B60R 1/00 H04N 7/18 J H04N 7/18 G06F 15/70 460B

Claims (4)

[Claims] 1. A moving object detection device that is mounted on a moving object and detects an external moving object by processing an external image read by an image reading device, wherein an image of a partial image area that divides an input image is cut out. And deriving means for deriving a motion vector of the partial image area by calculating a correlation with an input image sampled before or after the input image. Specifying means for specifying the partial image area formed by the detection means, calculating the correlation between the input image and the image of the moving object cut out from the partial image area specified by the specifying means, and detecting the position of the moving object. And a means for detecting a moving object. 2. A moving object detection device that is mounted on a moving object and detects an external moving object by processing an external image read by an image reading device, the correlation between two images designated as processing targets. By calculating the correlation calculation unit, and activating the correlation calculation unit with the image to be cut out of the partial image area that partitions the input image and the input image sampled before or after the input image as a processing target, Deriving means for deriving a motion vector of the partial image area; specifying means for specifying a partial image area formed by a moving object from the motion vector derived by the deriving means; and extracting a partial image area specified by the specifying means Detecting means for detecting the position of the moving object by activating the correlation calculating means with the image of the moving object and the input image as processing targets; Moving object detection apparatus that the characterized. 3. A program recording medium that is mounted on a moving body and records a program used for realizing a moving object detection device that detects an external moving object by processing an external image read by an image reading device. Deriving a motion vector of the partial image region by calculating a correlation between an image obtained by cutting out a partial image region that partitions the input image and an input image sampled before or after the input image. A process for specifying a partial image region formed by a moving object from a motion vector derived by the derivation process; and a correlation between an input image and a moving object image cut out from the partial image region specified by the specific process. , A program for causing a computer to execute a detection process for detecting the position of the moving object is stored. Lamb recording medium. 4. A program recording medium that is mounted on a moving body and records a program used for realizing a moving object detection device that detects an external moving object by processing an external image read by an image reading device. A correlation calculation process for calculating a correlation between two images designated as processing targets, an image cut out of a partial image region that partitions the input image, and an input image sampled before or after the input image. By executing the above-described correlation calculation processing with the processing target as a processing target, the derivation processing for deriving the movement vector of the partial image area, and the partial image area formed by the moving object are specified from the movement vector derived by the derivation processing. The above-described correlation calculation process is performed by specifying a specific process, an image of a moving object that cuts out a partial image region specified by the specific process, and an input image. By moving, that the program for executing the detection process for detecting the position of the moving object in a computer is stored, a program recording medium characterized.