JPH085316A - Method and apparatus for recognizing moving body - Google Patents

Abstract

PURPOSE:To provide a method and an apparatus for recognizing a moving body in a high accuracy by improving the follow-up property for the moving body, which is moving at a high speed. CONSTITUTION:Respective photographing regions 100 of a plurality of cameras 1-1 to 1-n are overlapped by a specified width 101 in the moving direction of a vehicle 10, and the measuring area is set. When the entrance of the moving body 10 is detected by the camera 1-1 at the head part at first, the respective cameras take the photograph based on the synchronizing signal from a camera control part 3. Image processing parts 2-1 to 2-n process the pictures from the corresponding cameras, narrow the number-plate region, and extract the partial data for every factor such as the ground-transport-bureau code of the vehicle number and the serial number by character recognition. The control part 5 synthesizes the partial data of the same vehicle number of the partial data from the respective image processing parts, forms the completed data and outputs the data into a central control device 6 and the like.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recognition system, and more particularly to a recognition method and apparatus suitable for identifying a moving body.

[0002]

2. Description of the Related Art As an image recognition system for a moving body, for example, as disclosed in Japanese Examined Patent Publication No. 62-7588, an image of the moving body is detected by a sensor, an ITV camera takes an image, and image processing is performed. What you recognize, ITV
In general, a part of the image pickup area of the camera is used to detect a moving object, and thereafter, an image of the previous image pickup area is photographed to perform recognition processing.

Further, Japanese Patent Application Laid-Open No. 62-139079 discloses an image processing apparatus utilizing a plurality of cameras. This method measures the distance between the camera and the moving object by overlapping all the imaging areas of a plurality of cameras, reduces the thinning amount of the image as the moving object approaches, and increases the resolution.
Recognize the distance, size, shape, etc. of moving objects existing in a dimension. Note that there is no description about the arrangement of a plurality of cameras.

According to this, a moving body approaching from any direction can be quickly detected at a distant position, and high-resolution image processing is performed at the approaching position, so that the ability to follow a target object moving at high speed can be improved. It is explained that it can be increased.

[0005]

In the above-mentioned conventional technique, the image for recognition processing is not limited to the case where a single camera is used, but also when a plurality of cameras are used. Limited in scope. Therefore,
At the time when the moving object is detected and the recognition image is captured next time, the high-speed moving object moves to the outside of the imaging area, which makes it difficult to capture the recognition image.

Further, if the camera angle is correctly set for the moving body by detecting the distances of the moving body by a plurality of cameras, an image of the leading moving body can be obtained in a close state, but the moving body of the succeeding moving body can be obtained. An image cannot be captured or deteriorates, resulting in extremely low recognition accuracy. In any case, it cannot be applied to a high-speed moving object or a large number of moving objects that need to be recognized at high speed, such as a vehicle recognition or article inspection line.

An object of the present invention is to provide a highly accurate recognition method and apparatus which can follow a moving body moving at a high speed or a moving body moving in a large amount and have high accuracy.

An object of the present invention is to provide a vehicle number recognition system capable of accurately recognizing the vehicle number of a vehicle moving at high speed.

An object of the present invention is to provide a parts inspection system which recognizes a large number of parts moving on an inspection line and identifies defective products.

[0010]

SUMMARY OF THE INVENTION An object of the present invention is to provide a moving body recognition method for recognizing an object in a measurement area in which image pickup areas of a plurality of cameras are connected by image processing. By setting the matching imaging areas to overlap each other by a predetermined width necessary to avoid erroneous recognition of the object or its element, and performing recognition processing of the object based on images taken simultaneously by each camera. To be achieved.

It is characterized in that the predetermined width is equivalent to an imaging width in the moving direction of the object or its element.

Another object of the present invention is to perform recognition processing for each image simultaneously captured by each camera, detect that recognition data obtained from a plurality of images have a partial relationship with the target object, and detect these parts. This is achieved by synthesizing the recognition data and recognizing the object.

[0013]

According to the present invention, the measurement area of the moving body is formed by connecting the image pickup areas of the plurality of cameras so as to overlap each other by a predetermined width in the moving direction of the moving body so that the plurality of cameras take images at the same time. , The measurement area can be expanded in the moving direction.

According to this, the followability to a high speed moving body is improved. Alternatively, high-speed processing can be performed by simultaneously recognizing a large number of moving objects. Furthermore, it is possible to recognize the relationship between moving bodies or groups of moving bodies that cannot be imaged by one camera.

The overlapping predetermined width is equal to or larger than the image width of the object to be recognized or the image width of the element when the object is composed of elements. here,
“Equivalent” refers to a range in which the object can be recognized, and is a width from a little small to a little large.

According to this, when the object is imaged by a plurality of cameras, the object or the element is always imaged in one imaging area, and therefore, the object or the element is not unrecognizable. In particular, if the overlapping predetermined width is set to the "equal" above,
Since the minimum overlapping width that does not cause a recognition error can be obtained, the measurement area can be widened, and unnecessary recognition processing due to overlapping can be reduced, and the processing speed can be improved.

Further, by combining the partial data obtained from each image, even when the elements of the object are imaged by a plurality of cameras, the data can be provided as complete data matching the object. Highly useful for monitoring. Further, partial data for which complete data cannot be obtained by the synthesizing process can be recognized as a defective product in a component inspection or the like, so that it is highly used.

[0018]

Embodiments of the present invention will now be described in detail with reference to the drawings. 1 to 10 show a first embodiment, which is an example applied to a vehicle number recognition system.

FIG. 1 is a schematic block diagram of a vehicle number recognition system. A plurality of ITV cameras 1-1 to 1 as imaging devices
-N is arranged so as to occupy the imaging region 100 in the traveling direction 11 of the vehicle 10 on the traveling lane. The overlapping areas 101 are provided in the imaging areas 100 of adjacent cameras,
It is a measurement area.

Each camera 1 uses the shutter system to pick up the image in the image pickup area at the same time. However, in the speed range where the image blur due to the movement of the vehicle is not a problem for the vehicle number recognition,
Imaging by the NTSC system, high-definition system, PAL system, or the like may be used.

FIG. 2 is a schematic diagram for explaining the image pickup area 100 by the plurality of ITV cameras 1. Cameras 1-1 to 1
Imaging areas 100-1 to 100 based on the visual field (dotted line) of -n
-N is the overlapping area 101-1 to 101- (n- of the width w.
1) is set. The width w of the overlapping area in the case of vehicle number recognition is set by the camera control unit 3 to be equal to or slightly larger than the vertical width of one line of the characters (numbers) on the license plate.

The image processing units 2-1 to 2-n instruct the corresponding camera 1 via the coaxial cable 8 to take in the image for recognition processing, extract the vehicle number data which is the recognition object, and extract the data from the bus. It is transmitted to the control unit 5 via (BUS). Therefore, each image processing unit 2 includes a control unit 21, an input / output processing unit 22, a difference unit 23, a binarization unit 24, a feature amount extraction unit 25, a recognition processing unit 26, a data buffer 27, and the like.

In response to an instruction from the control unit 5, the camera control unit 3 includes a synchronization signal output unit 31 for outputting a synchronization signal to each camera 1 and an overlapping region setting unit 32 for setting an overlapping width between adjacent image pickup regions. Have The overlapping area setting means may be provided as an adjusting mechanism of the ITV camera 1.

The control unit 5 takes in the vehicle number data recognized and processed by instructing the image processing unit 2 and the like, and totalizes the vehicle number data from the plurality of image processing units 2 to generate complete vehicle number data. To do. The result is used for the central control device 6 and the external device 7.
Or to display on the display version. Therefore, it is provided with a control means 51, an input / output processing means 52, a combining processing means 53, a data converting means 54, a buffer 55 and the like.

Next, the outline of the operation of the vehicle number recognition system configured as described above will be described with reference to the flowchart shown in FIG.

This processing is performed by the control unit 5 to the image processing unit 2-1.
Then, the vehicle detection process is instructed and activated. The image processing unit 2-1 instructs the ITV camera 1-1 to capture a moving body detection image (s101), and performs a moving body extraction process on the captured image (s102). When the moving body extraction process is completed, the presence or absence of a vehicle is determined (s103), and when the vehicle is not detected, the above-described moving body detection process of s101 to s103 is periodically repeated.

The moving body extraction processing (s102) is a well-known method, in which the background image in the picked-up area is removed and binarized from the picked-up image for detection, and the binarized image is transferred to the moving body. The moving body image is extracted by extracting the characteristic amount such as the corresponding area.

FIG. 4 shows an image pickup area at the time of the moving body detection processing by the ITV camera 1-1. In the traveling direction 11 of the vehicle in the imaging area 100-1, about 1/2 to 4/5 from the beginning is set as the moving object detection image area 102. That is, the image processing unit 2-1 acquires the image of the area 102 at the same time as the image capturing by the
At the speed of, the moving object is extracted and the presence / absence of the moving object is determined within the time corresponding to reading the remaining area 103.

The wider the detection image area 102 is, the faster the moving object can be detected. With the arrangement of only one camera as in the conventional art, the detection image area 102 is ⅓.
There is a limit to the extent, which may cause a detection error of a high-speed moving body. In the case of this example, since the recognition images are captured by a plurality of cameras as described later, the moving object detection screen area 102 of the leading camera can be enlarged, and the NTSC image (60 frames / 60 frames / Seconds) processing conditions,
When the area 102 has a width of ½, 80 km / h and a width of 4/5 can detect a vehicle of 150 km / h.

When a vehicle is detected by this moving object detection processing, the control unit 5 instructs the camera control unit 3 to output a synchronization signal and also instructs each image processing unit 2 to perform vehicle number recognition processing (s104). ).

Each image processing unit 2 instructs the corresponding ITV camera 1 to capture an image for recognition processing, and each camera 1 is synchronized with the image capturing area 100 shown in FIG. 2 at the timing of the synchronization signal from the camera control unit 3. Imaging of -1 to 100-n is performed simultaneously (s105).

Each image processing unit 2 performs the detection processing of the recognition target area, that is, the license plate candidate area on the captured recognition processing image (s106), and the vehicle number of the object is the object number of the detected candidate area. Data extraction processing is performed (s107).

FIG. 5 is an image transition diagram for explaining the license plate area detection processing. The same figure (a) shows one I
In the example of the recognition processing image by the TV camera 1, the imaging area 1
The image 200 has a vehicle at 00. For this image,
The background image is removed and the binarization processing is performed, and the binary image 2 in FIG.
Get 01. License plate candidate extraction processing is performed on the black part of this binary image by extracting the feature amount such as area,
The candidate images 202 (a) to (e) having the possibility of being a license plate are obtained as shown in FIG. In this example, an image area corresponding to the area of one line of the characters on the license plate is extracted as a candidate image so that the element data of the vehicle number can be detected.

Character recognition processing is sequentially performed on the area of each candidate image, and when a character (number) is recognized, the data is stored in a buffer (not shown) of the image processing unit 2 as data indicating a vehicle number or its element. store.

FIG. 6 shows the format of the vehicle number data 120 extracted from the candidate image area of the license plate 12. The vehicle number data 120 is stored in the buffer 27 as data (k, i) (the same applies to the buffer 55 after transfer). k is the number of the image processing unit 2 (or ITV camera 1), and i is the number of partial data. Partial data is Land Transport Bureau code 121 (Mito), vehicle type number 122 (55), application code 123 (sa), serial number 124 (12-34)
And position data 125. In order to manage a plurality of vehicle number data recognized by the same image processing unit, the plate number number m is used and the vehicle number data 120 is stored as data (k, m, i). The set of all partial data indicating the vehicle number will be called complete data, and the case of only a part will be called incomplete data.

FIG. 7 is an image showing the positional relationship of vehicle numbers by the image pickup area 100 of one camera. The figure (a) is
This is the case where some elements of the vehicle number are obtained as incomplete data at the top of the image 200. From this image, the Land Transport Office code 121 (Mito) and the vehicle model number 122 (55) cannot be recognized, and the application code 123 (sa) and the serial number 124 (12).
-34) is recognized as incomplete data. Positional data 125 is added to these incomplete data and stored in the buffer 21.

The position data 125 includes "HIGH",
There are three types of "MID" and "LOW", and the distance 204 from the upper end of the captured image to the lower end of the recognized character area of the vehicle number.
Is the distance 205 from the upper end to the lower end of the entire car number area
“HIGH” when the distance is shorter, and the distance 204 is the imaging area 20.
When the distance from the upper end to the lower end of 0 is longer than 206, “LO
W "and the distance 204 is between" HIGH "and" LOW "," MID "is given.

In the case of FIG. 9A, the position data 125 is "HIGH". However, the (101-1) image having no overlapping area above the imaging area 200 is “MID”. FIG. 11B shows an example in which the position data 125 is “MID”. FIG. 6C shows an example in which the position data 125 is “LOW”, and the Land Transport Office code 121 (Mito) and the vehicle model number 122.
(55) is obtained as incomplete data. However,
An image in which the imaging region 200 in (c) does not have an overlapping region below (101-n) is “MID”.

Next, the control unit 5 receives the partial data shown in FIG. 6 from each image processing unit 2, stores it in the vehicle number data area 551 of the buffer 55, and synthesizes the complete vehicle number data. Processing is performed (s108), data conversion according to specifications is further performed, and the data is output to the central control device 6 and the display board (s109).

FIG. 8 is a flowchart showing details of the combining process. To simplify the description, the number of vehicles recognized by one image processing unit 2 is 1 or less. First, the partial data number i, the incomplete data number j, and the image processing unit number k
Is initialized (s201), and the partial data of FIG.
0) to (1, 4) are sequentially read while incrementing i (s202), and it is determined whether the data group read in k is complete data consisting of all partial data of i = 1 to 5 (s203). ).

If it is complete data, the partial data of (1,0) to (1,3) excluding the position data is stored in the buffer 55.
Stored in the combined data buffer area 552 of the
4). The image processing unit number k is incremented (s20
5), it is determined whether k exceeds the final number n (s20
6), if not exceeded, the process returns to step s201.

On the other hand, when it is determined in step s203 that the data is incomplete, the incomplete data number j is incremented and the incomplete data buffer area 553 of the buffer 55 is incremented.
Temporarily store partial data including position data (s20
7) and proceeds to step s205. The incomplete data number j is updated for each partial data set within the partial data i = 0 to 4 in units of k.

Next, at step s206, when it is confirmed that the processing up to k> n is completed, the incomplete data number j is judged (s208), and if j = 0, the processing is completed.

On the other hand, if j ≧ 1, partial data including position data is sequentially read from the incomplete data buffer 553 (s209), and it is determined whether or not there is a partial data set that can be combined. That is, it is judged whether the position data of the partial data is "MID" (s210), and if it is not "MID", the rationality of the camera number k and the data number i is judged between the partial data sets of j unit (s211). , If the rationality between the sets is established, then those data are combined into one complete data (s212), and the combined data buffer area 552
(S213).

On the other hand, the position data of the partial data is "MI
In the case of “D”, since FIG. 7A corresponds to the first screen or FIG. 7C corresponds to the last screen, complete data cannot be obtained at this imaging timing. Further, it is conceivable that the reason cannot be obtained in step s211, for example, when all the partial data could not be correctly recognized due to dirt on the license plate.

These incomplete data are stored in the combined data buffer area 552 in units of j (s213). By setting the incomplete data in units of j, it is possible to distinguish from the vehicle numbers of other vehicles, but it is useful for identifying the vehicle although it is incomplete. Then, all the processes for the incomplete data number j are confirmed (s214).
When this happens, this process ends.

As a result, the composite data buffer area 552
Is the complete vehicle number obtained from the image of one camera, the complete vehicle number obtained from the images of the multiple cameras by the combining process, and the images of the multiple cameras for the vehicle passing at that timing. One or more of the incomplete vehicle numbers that have been assigned are accumulated. The control unit 5 converts the complete vehicle number and the incomplete vehicle number into transmission data and display data of a predetermined frame by the data conversion unit 54, and outputs them.

The central control unit 6 receives these pieces of information and uses them for monitoring and controlling the traffic volume or detecting a stolen vehicle or a violating vehicle. In this case, it is desirable that the data is complete data, but according to the present embodiment, even incomplete data can be identified for each vehicle, so that it can be used as effective data. The display process of the display plate may be performed from the control device 6.

FIG. 9 schematically illustrates the above-mentioned combining process. Overlapping area 10 of imaging areas 100-1 and 100-2
The width w of 1 is the serial number 12 with the largest vertical width of the car number character.
4 It is set so as to be 1 or more for the vertical width of "12-34".

Now, it is assumed that the license plate 12 is imaged as shown in (a) across the two imaging regions. When this is seen in each imaging area, it becomes like (b) and (c). That is, “Mito” and “56” are the imaging area 101-1.
“Sa” and “12-34” are recognized as partial data (1,0) and (1,1) of the image pickup area 101-2, respectively, as partial data (2,2) and (2,3). .

Therefore, in step s203, these partial data sets are determined to be incomplete data, and (d),
As in (e), the partial data (1, 4): "LOW" and (2, 4): "HIGH" position data are added and stored in the incomplete data buffer. This partial dataset is
Since the data position is not “MID”, step s
In the determination of 211, it is determined as "reasonable" from the relationship between the position data 125, the camera number k, and the partial data number i, and in step s212, it is combined as shown in (f) to form complete data.

If the width w of the overlapping area 101 is smaller than the maximum vertical width l of the car number character, the areas 100-1 and 100-1.
There is a case where it cannot be recognized by any of -2. On the other hand, if the overlapping width w is too large than the maximum character width l,
The measurement area consisting of the imaging areas of a plurality of cameras becomes narrower by that amount, which makes it impossible to support high-speed movement. In addition, useless recognition processing by overlapping images increases.

FIG. 10 illustrates the relationship between the imaging area and the recognizable vehicle speed. In this example, three ITV cameras are used, and the imaging areas 100-1 to 100-3 include an overlapping area and a measurement area is set. FIG. 7A shows the case where the vehicle is low speed. The moving body detection image 203 is captured at the position 14 of the image capturing area 100-1, the detection of the vehicle is confirmed after moving to the distance 15, and the recognition processing image 204 of the image capturing area 100-1 is captured. . In this case 1
It can be recognized by one camera. On the other hand, FIG. 2B shows a case where the vehicle is at high speed, and from the imaging of the moving object detection image 203 at the position 14 of the imaging area 100-1 to the imaging area 10 of the third unit.
The recognition image 204 is captured at the position 15 of 0-3.

For example, according to the processing conditions of the NTSC image, the vehicle speed at which the vehicle number can be recognized by one camera is about 8
0km / h or less About 150k when using 2 cameras
m / h or less About 220km / with 3 cameras
h.

In the above combining process, one image pickup area 10
It is assumed that the number of license plate plates recognized as 0 is m ≦ 1. However, when the speed of the moving body is slow, a plurality of plates are recognized. In this case, the recognition plate number m is managed for each image processing unit number (k), and the complete data is determined for each number m, so that a plurality of vehicle numbers can be recognized in one imaging area. If so, the synthesizing process becomes possible.

According to the vehicle number recognition system of the present embodiment, the monitoring area is expanded by combining the image pickup areas of a plurality of cameras. Therefore, all the camera images are picked up in accordance with the detection of the vehicle picked up by the first camera. The vehicle number can be recognized by using the image, and the vehicle number of a vehicle moving at high speed can be recognized.

Further, the image pickup areas of the adjacent cameras have an overlapping area in the traveling direction of the vehicle so that it can be recognized even when the license plate is imaged across two cameras. Moreover, since the width of the overlapping area is set to be equal to or slightly larger than the width of one line of the characters (numbers) on the license plate, one line of characters is always caught in one of the imaging areas and becomes unrecognizable. In addition, by setting the overlapping width to a minimum, the monitoring area (imaging area by a plurality of cameras) as a whole is maximized to expand the range of recognizable moving speed.

Furthermore, the partial data of the vehicle number taken over a plurality of cameras can be combined and provided as accurate vehicle number data, so that subsequent use is easy and highly reliable. In addition, since incomplete data that does not become a complete vehicle number is identified as a partial data set for each vehicle through synthesis processing,
Although it is incomplete, it can be used later.

Next, a second embodiment in which the present invention is applied to a parts inspection system will be described.

FIG. 11 shows a schematic structure of a nameplate character inspection system for semiconductor parts. Inspection line 82
1 are IC or LSI parts 816, 817 ,. ． ．
820 ,. ． ． Are placed at substantially equal intervals and are configured to move at a constant speed. Above the line, three cameras 805, 806 and 807 are mounted on rails 811 and 8
12 are set on each of fixed shafts 808, 809, 810.

The positions of the fixed shafts 808 to 810 are controlled by the camera control unit 804, and the imaging area 81 of each camera is controlled.
3, 814 and 815 are adjusted to have a predetermined overlapping area w, and the measurement area is set. Of course, it may be manually adjusted. The overlapping area w is represented by the parts 816 ,. ． ． It is set to be equal to or slightly larger than the character width of the nameplate characters such as the manufacturer's name and model printed on.

The cameras 805, 806, and 807 simultaneously shoot images by a synchronization signal from the camera control unit 804.
The image processing units 801, 802, and 803 binarize the photographed image in the same manner as described above, and then recognize the nameplate character on the component by matching with the pattern character, and recognize the recognized character data via the bus 822. To the control unit 800. The control unit 800 synthesizes the name plate character data of the image processing units 801 to 803, divides the data into complete data and incomplete data, totalizes the data, and outputs the inspection result.

In this example, when the sensor 823 detects that the heads of a plurality of parts rows have reached fixed positions on the line, the camera control unit 804 outputs a synchronization signal to each camera to perform imaging. . In FIG. 11, five parts enter the image pickup area by three cameras, and these five parts are inspected at once.

FIG. 12 is an explanatory view showing the name plate image taken by each camera and the result of the combining process. Images 90 by the cameras 805, 806 and 807
1, 902 and 90 are photographed. The name plate character that is the recognition object in this example is "HIT" in which the first line indicates the manufacturer name.
ACHI ", the second line shows the model" HM621100 "
Is. In addition, "HITACH
It has an overlapping width w corresponding to the line width of "I".

The image processing units 801 to 803 character-recognize the corresponding images 901 to 903 to form the format of the nameplate data 1000 shown in FIG.
Transmit to 0. The nameplate data 1000 includes its own image processing unit number (k) 1001 and the number of nameplates it has recognized (m).
1002 and 1 to m name plate character data 1003. Name plate character data 1003 is maker name code 10
04, format code 1005 and position data 1006, both data 1003 and 1004 are complete data, and only one is incomplete data.

The image processing unit 801 recognizes the complete data 816 'of the component 816 and the incomplete data 817' of only the manufacturer name of the component 817 from the image 901 (904).
It is transmitted to the control unit 800 in the format of 3. Similarly,
The image processing unit 802 recognizes incomplete data 817 ′ of only the model of the component 817 and complete data 818 ′ of the component 818 from the image 902 (905), and the image processing unit 803
Complete data 81 for parts 819, 820 from image 903
9 ', 820' are recognized (906), and the control unit 8
To 00.

The control unit 800 performs the same synthesizing process as in the first embodiment, synthesizes incomplete data that can be synthesized, and outputs the processing result 908. In the example of FIG. 12, incomplete data of the images 901 and 902 are combined, and finally, five sets of complete data are totaled. That is, in this inspection, all five parts are determined to be acceptable. On the other hand, if there is a rejected part, incomplete data remains in the totalized result, so that it is reported by an alarm or the corresponding incomplete data is displayed on the CRT.

According to this embodiment, a plurality of parts can be inspected at once by using the enlarged surveillance area by combining a plurality of cameras, so that the inspection speed can be greatly improved.

Further, since the measurement areas overlap the image pickup areas of adjacent cameras by the minimum necessary width so as to prevent recognition errors, accurate and high-speed inspection is possible. Further, since the overlapping width is configured to be adjustable according to the inspection object, it can be applied to the inspection of many kinds.

The present invention is not limited to the nameplate character inspection of parts. It can be applied in a wide range of fields such as inspection of a plurality of parts mounted on a printed board or inspection of patterns and shapes other than characters.

[0071]

According to the present invention, since the image pickup areas of a plurality of cameras are set so as to overlap each other in the moving direction of the moving body by a predetermined width, the measurement area can be expanded, and the moving body can be recognized at high speed and a large number of movements can be made. It has the effect of improving body recognition.

Further, since the predetermined width is set to be approximately equal to the image width of the recognition target object or its element, there is no recognition error even when the target object is imaged by a plurality of cameras. This has the effect of improving the recognition accuracy of the moving body. In addition, since the above-described predetermined width is the lower limit that does not cause a recognition error, waste of duplication processing can be reduced and the measurement area for the number of cameras can be maximized.

Furthermore, since partial data of the object extracted from the images of the cameras are combined, complete data matching the object is output even when the object extends over a plurality of cameras. It has the effect of improving reliability. In addition, since complete data can be used for identifying a specific object and incomplete data can be used for inspecting a specific object, etc., it can be applied to a wide range of applications and a flexible and easy-to-use system can be implemented.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a vehicle number recognition recognition system that is an embodiment of the present invention.

FIG. 2 is a schematic diagram illustrating an imaging area by a plurality of cameras.

FIG. 3 is a flowchart illustrating a schematic operation of the vehicle number recognition system according to the embodiment.

FIG. 4 is a schematic diagram illustrating an imaging region of a moving object detection processing image.

FIG. 5 is an image transition diagram illustrating a process of detecting a license plate candidate image from a recognition processing image.

FIG. 6 is a format diagram of a license plate and car number data.

FIG. 7 is an explanatory diagram illustrating a relationship between a position of a vehicle number and position data of an image captured by one camera.

FIG. 8 is a flowchart illustrating a combining process according to the first embodiment.

FIG. 9 is an explanatory diagram schematically illustrating an example of a combining process.

FIG. 10 is an explanatory diagram illustrating a relative relationship between an imaging area included in a measurement area and a recognizable vehicle speed.

FIG. 11 is a configuration diagram of a parts inspection system that is a second embodiment of the present invention.

FIG. 12 is an explanatory diagram showing a recognition processing result and a compositing processing result of a name plate character image captured by each camera.

FIG. 13 is a format diagram of recognized name plate character data.

[Explanation of symbols]

1 ... ITV camera, 2 ... Image processing unit, 3 ... Camera control unit, 4 ... Bus (BUS), 5 ... Control unit, 6 ... Central control device, 10 ... Moving body (vehicle), 12 ... License plate,
100 ... Imaging area, 101 ... Overlap area, 120 ... Car number data format, 200 ... Recognition processing image, 201 ...
Binary image, 202 ... License plate candidate area image,
800 ... Control unit, 801 to 803 ... Image processing unit, 804
... camera control unit, 805-807 ... camera, 816-8
20 ... Parts (semiconductor), 821 ... Inspection line, 823 ...
Position sensor.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G08G 1/04 D H04N 7/18 D // G05D 1/02 P (72) Inventor Kazuyoshi Asada 5-2-1 Omika-cho, Hitachi City, Ibaraki, Ltd. Inside the Omika Plant, Hitachi, Ltd. (72) Inventor, Futa Ikeda 5-2-1 Omika-cho, Hitachi City, Ibaraki, Hitachi Process Computer Engineering Co., Ltd.

Claims (19)

[Claims] 1. A method of recognizing a moving object for recognizing an object in a measurement area in which imaging areas of a plurality of cameras are connected to each other by image processing, wherein in the measurement area, adjacent imaging areas are adjacent to each other or the object. A method for recognizing a moving object, characterized in that the object recognition processing is performed based on images simultaneously set by respective cameras, which are overlapped by a predetermined width necessary for avoiding element recognition error. 2. A method of recognizing a moving object for recognizing an object in a measurement area in which image pickup areas of a plurality of cameras are connected by image processing, wherein a plurality of images are recognized by performing recognition processing for each image simultaneously captured by each camera. A method for recognizing a moving object, characterized in that the recognition data obtained from the above is detected to have a partial relationship with the target object, and the partial recognition data is combined to recognize the target object. 3. The measurement area according to claim 2, wherein the measurement areas are set so that adjacent imaging areas overlap each other by a predetermined width corresponding to an imaging width when viewed in a moving direction of the object or its element. How to recognize moving objects. 4. The partial relation according to claim 2, wherein the recognition process is performed for each element that constitutes the object, and is detected from an element number given to the recognition data. How to recognize moving objects. 5. The method for recognizing a moving body according to claim 4, wherein the predetermined width is set to correspond to the largest imaging width among the plurality of elements. 6. A moving object having an identification object drawn with characters, numbers, symbols, etc. (hereinafter referred to as a character body) is imaged in a measurement area in which the imaging areas of a plurality of cameras are connected, and image processing is performed. In the method of recognizing a moving object for recognizing the identification object, the measurement area is set so as to overlap by a predetermined width equivalent to or more than the imaging width of the adjacent imaging regions viewed in the moving direction of the character body, For each image taken simultaneously by each camera, the target image related to the identification object is narrowed down by predetermined image processing, and character recognition is performed from the target image for each line of the character style or each line element to extract partial data. Then, a method of recognizing a moving object, characterized in that all the partial data based on each camera are combined. 7. The positional data according to claim 6, wherein the partial data indicates a positional relationship between an identification number of the row unit or the element unit of the identification object and an imaging area of a target image from which the partial data is extracted. Is added, the synthesizing process combines the partial data having the same identification object relationship based on the identification number and / or the position data, and outputs the complete data of the identification object. A method of recognizing a moving object as a feature. 8. The method for recognizing a moving object according to claim 7, wherein the synthesizing process outputs non-synthesizable ones of the partial data as incomplete data of an identification object. 9. The method according to claim 6 or 7 or 8, wherein when a plurality of partial data are extracted from a plurality of target images for each imaged image, a segment number of the target image is given to each partial data, A method for recognizing a moving object, which is characterized in that the combining process is performed with reference to a division number. 10. The simultaneous photographing timing according to claim 8, wherein the simultaneous photographing is performed immediately after the moving body is detected by the image processing from the periodic image pickup of the camera arranged at the head in the approach direction of the moving body. The method of recognizing a moving body, wherein the image processing for detecting the moving body is performed only within a predetermined length from the tip of the imaging region. 11. The vehicle according to claim 10, wherein the moving body is a vehicle, the identification object is a vehicle number of a license plate, and the element for each row includes a Land Transport Office code or a serial number. How to recognize moving objects. 12. The method for recognizing a moving object according to claim 10, wherein the simultaneous image capturing timing is immediately after the leading moving object reaches a predetermined position near the exit of the measurement area. 13. The moving body according to claim 12, wherein the moving body is a component placed on a moving line, the identification object is a nameplate character group, and the element for each line includes a manufacturer code or a model code. Alternatively, a method of recognizing a moving body is characterized by simultaneously recognizing the nameplate characters of a plurality of parts. 14. A moving object having an identification object drawn with characters, numbers, symbols, etc. (hereinafter referred to as a character body) is imaged in a measurement area arranged so as to connect the imaging areas of a plurality of cameras. In the moving body recognition device for recognizing the identification object by image processing, the measurement areas are set so as to overlap each other by a predetermined width corresponding to an imaging width in which the adjacent imaging areas are viewed in the moving direction of the character body. And a synchronization signal output means for simultaneously photographing with each camera, and an image provided for each camera, which character-recognizes from the photographed image on a line-by-line or element-by-line basis of the character style to extract partial data. A combination processing unit that receives the partial data from the processing unit and the image processing unit, combines all the partial data based on each camera, and outputs character data corresponding to the identification object. ,
A moving body recognition device, comprising: 15. The apparatus for recognizing a moving body according to claim 14, further comprising adjusting means for adjusting the position of the camera so that the adjacent image pickup areas overlap each other by the predetermined width. 16. A plurality of cameras are arranged in an upper space of a traveling lane of the vehicle so that the imaging areas are connected to each other to set a measurement area, and an image of a vehicle number on a license plate of an approaching vehicle is captured and image processing is performed. In the vehicle number recognition system, the measurement areas are set so as to overlap each other by a predetermined width corresponding to the image pickup width of adjacent image pickup areas viewed in the moving direction of the characters of the vehicle number. A synchronization signal output unit that outputs a synchronization signal for simultaneously capturing images by each camera when an approach of a vehicle near the entrance is detected, and a line unit of the vehicle number from the captured image, which is provided for each camera. Alternatively, an image processing unit that character-recognizes each line element and extracts partial data, and receives the partial data from the image processing unit, Serial partial data by combining processing, vehicle number recognition system for the synthesis processing unit, characterized in that it comprises outputting the character data that exactly match the character data and / or incompletely coincides with the vehicle number. 17. The detection of a vehicle entering the measurement area according to claim 16, wherein the image of the image processing unit is detected based on periodical imaging of the camera arranged at the head in the approach direction of the measurement area. The vehicle number recognition system is characterized in that the image processing is performed by using a range of 1/2 to 4/5 of the imaging area of the leading camera. 18. A measurement image is set by arranging a plurality of cameras in the space above the movement line so that the image pickup areas are connected to each other, and images of nameplate characters of parts placed on the movement line and entering the image are picked up and imaged. In the component inspection system that is recognized by processing, the measurement areas are set so that adjacent imaging areas overlap by a predetermined width corresponding to the imaging width as viewed in the moving direction of the name plate characters, and the vicinity of the exit of the measurement area When the entry of the part is detected at a predetermined position, a synchronization signal output means for outputting a synchronization signal for simultaneously performing imaging by each camera, and provided for each camera, the name plate character Image processing means for extracting partial data obtained by character recognition in line units or line element units, and receiving the partial data from the image processing means, And combining processing means for combining all of the partial data to output character data that completely matches the nameplate characters and / or character data that completely matches the nameplate characters, and the incompletely matching character data and / or Alternatively, the component inspection system is provided with an output unit that outputs a component defect signal. 19. The image processing means according to claim 18, wherein, when the photographed image has a plurality of target images corresponding to the nameplate characters, the image data is assigned a partition number for partitioning the target image to the partial data. The component inspection system, wherein the synthesizing processing unit performs synthesizing processing with reference to the section number.