JPS62121586A - Automatic recognition device for number plate - Google Patents

Abstract

PURPOSE:To simplify a processor by detecting a number plate position and a serial number with one television camera, performing an image pickup with plural television cameras, and recognizing complicated characters, such as District Land Transport Bureau, the class of a vehicle, and a usage code within the plate. CONSTITUTION:When the television camera performs the image pickup of the vehicle in travelling, image data fetched in from a television camera 2 having a wide view field are processed, and the existing position of a number plate within a screen is detected, and a serial number part which as large characters and is easy to recognize is taken out and recognized. And the addresses within the screen of the class of the vehicle, and the usage code on the plate are found, and the addresses that correspond to the picture planes of plural cameras having narrow view fields provided rotating by 90 degrees are calculated. Based on the bit of above information, required parts within static picture data fetched in from plural television cameras 3-7 are processed, and the District Land Transport Bureau, the class of the vehicle, and the usage code within the plate are recognized. Thereby, the data to be processed can be reduced remarkably, and furthermore, since the algorithm of a process can be reduced, a data memory part or a data processing part can be simplified.

Description

[Detailed Description of the Invention] Industrial Application Field The present invention provides an automatic license plate recognition device that images a traveling vehicle with a television camera, processes the obtained still image, and recognizes the vehicle's license plate i~. Regarding.

When inputting images using conventional technology, for example, a television camera, it is generally ideal to realize this with one camera. However, if high resolution and wide viewing range are required, expensive television cameras are required. As a method for realizing this using an inexpensive standard television camera, there is a so-called multi-camera image reduction method in which a plurality of television cameras form a moving image device.

If the resolution of the TV camera is n (TV lines), the resolution required for recognition is a (mm/TV lines), and the required field of view is L (mm), then the number of cameras is N.
must satisfy N≧L/(na). Also, −
2m to compare the license plate of a car on the lane
For the field of view range, and for character recognition of image data,
It has been experimentally confirmed that a resolution of 2 mm/T lines is required. For this reason, the standard 30OT
When forming this ham device with V television cameras, four television cameras are required.

FIG. 4 shows the field of view according to this multi-camera image withdrawal method. However, with this method, when the license plate to be recognized is on the boundary line of the camera field of view, it becomes extremely difficult to recognize it. In order to solve this problem, a method was proposed in which the camera field of view overlaps a portion of the width of the plate. Considering the case where the above resolution and field of view range are achieved with the above camera using this method, 1
This means that 0 cameras are required. FIG. 5 shows the field of view according to this multi-camera imaging method.

In this figure, P is the width of the large plate (440 mm
). Fig. 3 is a block diagram of an automatic license plate recognition device configured by this multi-camera imaging method. In the figure, 22 is a vehicle passing position detection sensor;
3 to 32 are 30OT V TV cameras (E Nochie 0
．． 6m), 33-42 are A/D converters, 43-52
is a binary memory, 53-62 is a multi-value memory, 63-72
is a plate position detection section, and γ3 is a character recognition section.

In the conventional device shown in FIG. 3, when a vehicle approaches, the position sensor 22 issues a shutter trigger signal, and the television camera 2
3 to 32 capture the image as a still image. The captured image is A/D converted by A/D converters 33 to 42.
The data is converted and stored as digital data in multivalued image memories 53 to 62 with 8 bits per pixel and 256 gradations. At this time, the data is also binarized at a certain level,
It is also stored in the binary image memories 43-52. Next, the plate detection units 63-72 process the data in the binary memories 43-52 and detect the plate portion. Based on this information, only a certain part of the character is extracted from the multi-value memories 53 to 62, and is recognized by the character recognition section 73. In this way, even by using this multi-camera method, it is possible to configure the imaging section of the license plate self-vJ recognition device.

Problems to be Solved by the Invention However, in the conventional multi-camera imaging method described above, it is necessary to process all image data from multiple cameras in a short time, and to achieve this, high-speed image processing that detects the plate position is required. It was necessary to have multiple devices corresponding to the number of cameras. Therefore, there was a problem that the device was expensive and large. In addition, the vertical field of view of these 300 V cameras is 0.45 m (0, f3 x 0
075), in order to image the license plate within this range, a high-performance position sensor with little deviation in sensing timing depending on the vehicle type is required.

The present invention solves the above problems, and provides an automatic license plate recognition device using an excellent multi-camera method that can delete processing data and simplify processing equipment. The purpose is to provide

Means for Solving the Problems In order to achieve the above object, the present invention provides a television camera with a wide field of view for detecting the position of the license plate and recognizing the serial number, and We propose the use of a plurality of television cameras that are rotated 90 degrees around the imaging optical axis in a plane perpendicular to the imaging optical axis to recognize vehicle types, vehicle types, and usage codes.

Effects The present invention has the following effects due to the above configuration. In other words, when a television camera captures an image of a moving vehicle, the image data captured by a single television camera with a wide field of view.
to detect the location of the license plate on the screen. Then, the serial number part, which has large characters and is easy to recognize, is extracted and recognized.

In addition, the address on the screen of His Majesty, car model, and usage code in the plate is obtained, and its position is located on multiple 90 cars with a narrow field of view.
Calculate the address that corresponds to the camera screen installed by rotating the camera. Based on this information, the system processes the necessary parts of still screen data captured from multiple television cameras to recognize the name, vehicle type, and usage code on the plate. As a result, it is possible to increase the speed of image data processing and simplify the apparatus.

Embodiment FIG. 1 shows the configuration of an embodiment of the present invention. In FIG. 1, 1 is a position sensor, 2 is a television camera whose viewing range is set to 2 m, and 3 to 7 are data cameras whose viewing range is set to 0.6 m and are rotated 90 degrees.

Because it was rotated 90 degrees, the horizontal field of view of one camera was 0.4
It will be 5m. The resolution of the TV camera here is 30OT.
It will be a V book. Also, 8 to 13 are A/D converters, 14
1 is a binary memory, 15 to 1 are multi-value memories, 20 is a plate position detection section, and 21 is a character y2H section. Further, FIG. 2 shows the visual field range according to this embodiment. In the figure, C is the size of one character of the license plate's Majesty, vehicle type, and purpose code (40 mm).

Next, the operation of the above embodiment will be explained. After vehicle entry 2
The process up to the point where still image data is stored in the value memory and the multi-value memory is the same as that described in FIG. 3. Next, the plate position detection section 20 processes the data in the 21-direction meter 14 and detects the position of the plate section. Since the binarization memory 14 contains all the information for one license plate, by processing this f-number, it is possible to determine which part (one or two) of the number plates from ■ to ■ in Figure 2. You can get information about what is included. The resolution of one bale of this image is about 8 mm/TV line, which is poor resolution, but it is sufficient to cut out the entire plate, which has a large amount of information, and to recognize serial numbers. These two fins at the same time
Only the serial numbers in the format memory 14 are detected, and the data is passed to the character recognition section 21 to recognize the serial numbers. Therefore, since it is only necessary to recognize the data in the multi-value memories 15 to 19 such as His Majesty, car type, and usage code, the camera overlap can be 40 mm for one character as shown in FIG. Therefore, six cameras are required.

The license plate detection information is sent to the character recognition section 21,
Based on this information, only the necessary multilevel memory is processed and characters are detected. This data is transmitted to the character recognition section 21, and the characters are recognized. Here, cameras 3 to 7 for capturing images of His Majesty, vehicle type, and usage code are installed rotated by 90 degrees in a plane perpendicular to the imaging optical axis, so they can be rotated 90 degrees in a plane perpendicular to the imaging optical axis. In order to create a shape, it is necessary to convert the data in memory. However, in this device, the pattern matching method is used for character recognition, and as shown in Figure 6, the dictionary pattern used for character recognition is also rotated 90 degrees. No such conversion is necessary, and the decoy image data can be processed as is. FIG. 6 shows II 511.

Effects of the Invention According to the present invention, as is clear from the above embodiments, plate position detection and character recognition of only the serial number are performed using images captured by a single camera with a wide field of view, and from the results, only necessary parts are detected. By processing high-resolution images taken with multiple cameras, complex characters such as the 1 code can be recognized, making it possible to significantly reduce the amount of data to be processed.

Furthermore, since the processing algorithm is shortened, the data memory section and data processing section can be simplified. In addition, multiple cameras for complex character recognition are rotated 90 degrees around the positive image optical axis in a plane perpendicular to this, providing a wider vertical field of view than if installed upright. This allows the sensing area of the position sensor to have a certain amount of leeway.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram of an automatic license plate recognition device according to an embodiment of the present invention, and FIG. 2 is a schematic block diagram of an automatic license plate recognition device according to an embodiment of the present invention. Figure 3 is a schematic block diagram of an automatic license plate recognition device using the conventional multi-camera imaging method, Figures 4 and 5 are diagrams showing the field of view using the conventional multi-camera 1m method, and Figure 6 The figure shows data conversion of a detected character part. 1.22...Vehicle passing position detection sensor 2-7, 23-
32...TV camera (30 OTV books) 8~13.3
3-42... A/D converter 14, 43-52...
・21iII data storage method 15-19.53-6
2...Multi-value data storage memory 20, 63-72...
Plate position detection unit 21, 73...Character recognition unit Patent applicant Matsushita Electric Industrial Co., Ltd. 42 Liao Figure 1 Figure 2 Figure 3

Claims (1)

[Claims] A single television camera that captures an image over one lane of view detects the location of the license plate, recognizes the serial number, and rotates the same field of view 90 degrees around the imaging optical axis in a plane perpendicular to this. An automatic license plate recognition device that captures images using multiple television cameras installed at the parking lot and recognizes complex characters such as land, vehicle type, and usage codes on the plate.