KR102077396B1 - Method for generating license plate image using noise pattern and apparatus therefor - Google Patents

Abstract

Disclosed are a vehicle license plate image generating method using a noise pattern and an apparatus thereof. Vehicle license plate image generation method according to an embodiment of the present invention comprises the steps of receiving a first image for the license plate; And generating a second image to which the noise pattern is applied to the vehicle license plate by applying at least one noise pattern to the first image, wherein the plurality of images include noises using a pre-learned network. Extracting a noise pattern from each, and generating a noise pattern filter to which each of the extracted noise patterns is applied, wherein generating the second image comprises at least one noise of the generated noise pattern filters The second image may be generated by applying a pattern filter to the first image.

Description

Method for generating vehicle license plate image using noise pattern and its device {METHOD FOR GENERATING LICENSE PLATE IMAGE USING NOISE PATTERN AND APPARATUS THEREFOR}

The present invention relates to a technology for generating a vehicle license plate image using a filter to which a noise pattern is applied. More specifically, a training data set for a vehicle license plate image required for deep learning is used as a filter to which a noise pattern is applied. It relates to a method and an apparatus that can be generated by the.

Various types of security systems are used, for example, by sensing an area within a road to detect the presence of a car, photographing the license plate of the car according to the detection result, and extracting letters and numbers from the captured image, Identify or recognize license plates.

For example, a license plate recognition system may extract a license plate of a vehicle using a high performance license plate recognition (LPR) engine. The LPR engine may perform optical character recognition (OCR) on a license plate area in a photographed camera image. An optical character recognition function may be provided to output a recognition result as text.

The LPR system is an apparatus for automatically extracting vehicle number data from an image taken by a camera, and is widely used in areas such as billing, enforcement, security, and parking. Generally, the license plate recognition system captures noisy images due to contrast between natural phenomena and obstacles, heavy noise in rainy weather, sunlight in the daytime, and reflection of vehicle lights at night, which makes it difficult to extract license plates. There is this.

In LPR artificial intelligence using deep learning, in order to improve the inference performance of these noisy (or high-noise) images, machine learning requires a large amount of images, but it is difficult to obtain near-realistic data. .

In the past, a method of synthesizing a high noise license plate image using computer graphic technology was used to secure image data necessary for deep learning.However, in order to infer the number from an image captured in such a low visibility environment, the image was taken in the same environment. Although a large amount of photorealistic images are required, it is very difficult and expensive to secure an image through shooting.

Accordingly, there is a need for a method that can easily generate vehicle license plate images in various environments required for deep learning.

Embodiments of the present invention provide a vehicle license plate image generating method and apparatus using a noise pattern that can generate a training data set for a vehicle license plate image required for deep learning using a filter to which a noise pattern is applied.

Embodiments of the present invention provide a vehicle license plate image generation method using a noise pattern capable of generating or synthesizing a vehicle license plate image to which each of the noise patterns is applied by applying noise patterns that may occur in various environments to the vehicle license plate image. Provide the device.

Vehicle license plate image generation method according to an embodiment of the present invention comprises the steps of receiving a first image for the license plate; And generating a second image to which the noise pattern is applied to the vehicle license plate by applying at least one noise pattern to the first image.

The generating may locally generate the second image to which the noise pattern is locally applied by applying the noise pattern to the vehicle license plate locally.

The generating may sequentially generate the second image to which the plurality of noise patterns are applied by sequentially applying each of the plurality of noise patterns to the first image.

The generating may include applying each of the plurality of noise patterns to the first region to a local region differently set with respect to the first image, thereby generating the second image applied to the different region where the plurality of noise patterns are differently set. Can be generated.

Furthermore, the vehicle license plate image generating method according to an embodiment of the present invention extracts a noise pattern from each of a plurality of images including noises using a pre-learned network, and applies a noise pattern to each of the extracted noise patterns. The method may further include generating a filter, wherein generating the second image may generate the second image by applying at least one noise pattern filter among the generated noise pattern filters to the first image. have.

Vehicle license plate image generating apparatus according to an embodiment of the present invention includes a receiving unit for receiving a first image for the license plate; And a synthesizer configured to apply at least one noise pattern to the first image to generate a second image to which the noise pattern is applied to the license plate.

The synthesizer may generate the second image to which the noise pattern is locally applied by locally applying the noise pattern to the vehicle license plate.

The synthesizer may generate the second image to which the plurality of noise patterns are applied by sequentially applying each of the plurality of noise patterns to the first image.

The synthesis unit may generate the second image applied to the local area in which the plurality of noise patterns are differently set by applying each of the plurality of noise patterns to the first image to the local area. Can be.

Furthermore, the apparatus for generating a vehicle license plate image according to an embodiment of the present invention extracts a noise pattern from each of a plurality of images including noises using a pre-learned network, and applies a noise pattern to which each of the extracted noise patterns is applied. The apparatus may further include a generation unit configured to generate a filter, and the synthesis unit may generate the second image by applying at least one noise pattern filter among the generated noise pattern filters to the first image.

According to embodiments of the present invention, a training data set for a vehicle license plate image required for deep learning may be generated using a filter to which a noise pattern is applied. For example, the present invention may generate or synthesize a vehicle license plate image to which each of the noise patterns is applied by applying noise patterns that may occur in various environments to the vehicle license plate image.

According to embodiments of the present invention, by extracting the noise patterns from the vehicle license plate image including the actual noise photographed in various environments, by applying the extracted noise patterns to generate a vehicle license plate image including the noise pattern, Deep learning data sets can be synthesized and obtained as desired, thereby enabling learning data to be acquired at low cost and high efficiency.

According to embodiments of the present invention, a new image is generated by locally applying a plurality of noise patterns to different regions of the license plate image, and the learning time may be reduced when deep learning is performed using the generated images. The recognition accuracy of the high noise image can be improved.

1 is a flowchart illustrating an operation of a vehicle license plate image generating method according to an exemplary embodiment of the present invention.
2 illustrates an example of a process of generating a vehicle license plate image to which a noise pattern is applied according to the method of the present invention.
Figure 3 shows another example of the process of generating a vehicle license plate image to which the noise pattern by the method of the present invention is applied.
Figure 4 shows another example of the process of generating a vehicle license plate image to which the noise pattern is applied by the method of the present invention.
5 illustrates a configuration of an apparatus for generating a license plate image according to an embodiment of the present invention.

Advantages and features of the present invention, and methods for achieving them will be apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various different forms, only these embodiments are intended to make the disclosure of the present invention complete, and common knowledge in the art to which the present invention pertains. It is provided to fully inform the person having the scope of the invention, which is defined only by the scope of the claims.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, “comprises” and / or “comprising” refers to a component, step, operation and / or element that is one or more of the other components, steps, operations and / or elements. It does not exclude existence or addition.

Unless otherwise defined, all terms used in the present specification (including technical and scientific terms) may be used in a sense that can be commonly understood by those skilled in the art. In addition, terms that are defined beforehand that are generally used should not be interpreted ideally or excessively unless they are clearly specifically defined.

Hereinafter, with reference to the accompanying drawings, it will be described in detail preferred embodiments of the present invention. The same reference numerals are used for the same elements in the drawings, and duplicate descriptions of the same elements are omitted.

Embodiments of the present invention, the main idea is to generate a training data set for the vehicle license plate image required for deep learning using a filter to which the noise pattern is applied.

Here, the present invention may generate or synthesize a vehicle license plate image to which each of the noise patterns is applied by applying noise patterns that may occur in various environments to the vehicle license plate image.

In this case, the noise patterns may be extracted from a vehicle license plate image including actual noise photographed in various environments using a network previously trained on the noise patterns, for example, a convolutional neural network (CNN).

Furthermore, the present invention synthesizes or applies a noise pattern to a vehicle license plate image by using the extracted noise pattern or a filter to which the noise pattern is applied and a network as a license plate image, and thereby applies a high noise noise pattern to the vehicle. License plate images can be created or synthesized.

In this case, the present invention may apply a plurality of noise patterns to the vehicle license plate image, and the method of applying the noise patterns may be sequentially applied to all areas, or different noise patterns may be applied to different preset local areas. have.

In addition, the license plate image used as an input in the present invention may be a vehicle license plate image actually photographed, a vehicle license plate image produced using graphic technology, a noise pattern from the vehicle license plate image including the actual photographed noise It may also be a vehicle license plate image removed. Of course, the license plate image used in the present invention is not limited to the aforementioned license plate image, and may include all kinds of license plate images available in the present invention.

This invention will be described with reference to FIGS. 1 to 5 as follows.

1 is a flowchart illustrating an operation of a vehicle license plate image generating method according to an exemplary embodiment of the present invention.

Referring to FIG. 1, a vehicle license plate image generating method according to an embodiment of the present invention includes a vehicle license plate image included in each of a plurality of license plate images including noises using a pre-learned network, for example, CNN. Noise patterns are extracted (S110).

Here, step S110 may extract a noise pattern included in the vehicle license plate image, for example, a high noise pattern, by using a network of a trained training model to extract a noise pattern from an input image, and may generate a network such as a CNN. Since a process of extracting a noise pattern using the same will be apparent to those skilled in the art, a detailed description thereof will be omitted.

For example, step S110 may be included in the vehicle license plate image from a certain number of various environments, such as vehicle headlights, darkness, scratches, dents, etc., for example, from each of 100 actual photographed vehicle license plate images. High noise patterns can be extracted.

When the noise patterns are extracted in step S110, a filter to which each of the extracted noise patterns is applied, that is, a noise pattern filter is generated (S120).

The steps S110 to S120 are processes of extracting a noise pattern from a vehicle license plate image including noise by using a network of a trained learning model in advance, and generating a noise pattern filter using the noise pattern filter. It can be used to create a new license plate image that includes a noise filter. Of course, the steps S110 to S120 may be periodically updated to generate a new noise pattern filter, and the updated noise pattern filter may be used to generate a vehicle license plate image. Of course, the noise pattern filter in the present invention may be generated using a plurality of noise patterns. For example, a noise pattern filter may be generated to apply a plurality of noise patterns to the same area, and a noise pattern filter may be generated to apply the plurality of noise patterns to each preset local area. That is, each noise pattern filter may be synthesized to generate a new noise pattern filter.

Steps S130 to S140 use the noise pattern to generate or synthesize a new license plate image including the noise pattern.

To generate a license plate image that includes high noise, a first image for the license plate is received and applied to the license plate of the first image by applying at least one or more noise pattern filters preset or selected to the received first image. In operation S130 and S140, a second image to which at least one noise pattern is applied is generated.

Here, the step S130 may receive an actual photographed vehicle license plate image, may receive a vehicle license plate image produced by graphic technology, or may be a vehicle license plate image from which noise is removed from the actually photographed vehicle license plate image.

Further, step S140 is a high-quality vehicle license plate image to which the noise pattern filter is applied using a pre-learned network that uses the noise pattern filter and the first image as an input, for example, a deep neural network (DNN) or a convolutional neural network (CNN). Can be generated.

In operation S140, a high noise vehicle license plate image may be generated using one noise pattern filter, and a vehicle license plate image including a plurality of noise patterns may be generated using a plurality of noise pattern filters. This will be described below with reference to FIGS. 2 to 4.

FIG. 2 illustrates an example of a process of generating a vehicle license plate image to which a noise pattern is applied according to the method of the present invention. As shown in FIG. 2A, a vehicle license plate image 210 of "Chungbuk 99 bar 5190" When the noise pattern filter 220 received and extracted and generated from the high noise vehicle license plate image as shown in FIG. 2B is selected or set, the noise pattern filter is applied to the license plate image 210 of "Chungbuk 99 bar 5190". By applying 220, a new license plate image 230 of "Chungbuk 99 bar 5190" is obtained by synthesizing the noise pattern filter 220 to the license plate image 210 of "Chungbuk 99 bar 5190." That is, FIG. 2 synthesizes a vehicle license plate image including the noise pattern by applying one noise pattern filter to the entire license plate image of “Chungbuk 99 bar 5190” as an input image.

3 is a view illustrating another example of a process of generating a vehicle license plate image to which a noise pattern is applied according to the method of the present invention. As illustrated in FIG. 3, a vehicle license plate image 310 of “Chungbuk 99 bar 5190” is shown. Is received and the first noise pattern filter 320 and the second noise pattern filter 330 generated by being extracted from the high noise vehicle license plate images are selected or set, the vehicle license plate image 310 of "Chungbuk 99 bar 5190" By sequentially applying the noise pattern filters 320 and 330 to the vehicle license plate image 310 of " Chungbuk 99 bar 5190 " Generates a license plate image 340 of Chungbuk 99 bar 5190 ". That is, FIG. 3 sequentially applies two noise pattern filters to the entire license plate image of the input image “Chungbuk 99 bar 5190” to synthesize the vehicle license plate image including the two noise patterns in the entire area. Although FIG. 3 illustrates that the two noise pattern filters are sequentially applied to the entire area of the image, the present invention is not limited thereto and more noise pattern filters may be sequentially applied.

4 is a view illustrating another example of a process of generating a vehicle license plate image to which a noise pattern is applied according to the method of the present invention. As illustrated in FIG. 4, a vehicle license plate image 410 of “Chungbuk 99 bar 5190” is shown. ) Is received, and the first to fourth noise pattern filters 420 to 450, which are generated by being extracted from the high noise license plate images, are selected or set, and each filter is different from the license plate image 410. By applying to the local area, the vehicle license plate image 460 of the new "Chungbuk 99 bar 5190" combining the respective noise pattern filters 420 to 450 in each local area of the vehicle license plate image 410 of "Chungbuk 99 bar 5190" ) For example, in the license plate image 410 of "Chungbuk 99 bar 5190," the first noise pattern filter 420 is applied to a local area including "Chungbuk 9", and a second area is included in the local area including "9 bar." Applying the noise pattern filter 430, applying the third noise pattern filter 440 to the local area including "51", and applying the fourth noise pattern filter 450 to the local area including "90". As a result, a new license plate image 460 of “Chungbuk 99 bar 5190” that combines the noise pattern filters 420 to 450 in each local area can be generated. That is, FIG. 4 illustrates that a plurality of noise pattern filters are applied to each local area set for a vehicle license plate image of “Chungbuk 99 bar 5190” as an input image, whereby each local area includes a vehicle license plate image. To synthesize. Of course, the present invention is not limited to applying different noise pattern filters to different local regions as shown in FIG. 4, and noise pattern filters may be applied to overlap some regions, and a plurality of noise pattern filters may be applied to any one local region. The same noise pattern filter may be applied to two local regions. As such, the present invention can use any method that can apply a noise pattern filter to an image.

As described above, the method according to the embodiments of the present invention may generate a training data set for a vehicle license plate image required for deep learning using a filter to which a noise pattern is applied. For example, the present invention may generate or synthesize a vehicle license plate image to which each of the noise patterns is applied by applying noise patterns that may occur in various environments to the vehicle license plate image.

In addition, the method according to the embodiments of the present invention extracts the noise patterns from the vehicle license plate image including the actual noise photographed in various environments, and applies the extracted noise patterns to obtain a vehicle license plate image including the noise pattern By generating, the data set for deep learning can be synthesized and obtained as desired, and thus the learning data can be obtained at low cost and high efficiency.

In addition, the method according to the embodiments of the present invention generates a new image by applying a plurality of noise patterns locally to different areas of the license plate image, and learning time when deep learning is performed using the generated image In addition, the recognition accuracy of high-noise images can be improved.

In addition, the method according to the present invention may be applied in a complex manner to apply in the above-described illustration. That is, the present invention may synthesize or generate a new high-quality license plate image including the noise pattern by combining and applying various noise patterns that may be included in the license plate in various environments.

In addition, the method according to the present invention can generate a high quality vehicle license plate image including noise using an input image produced using graphic technology, thereby facilitating a high quality data set for deep learning to recognize a vehicle license plate. In addition, even if a new type of license plate is applied, the recognition accuracy of the modified license plate may be increased. That is, the present invention can generate a data set for recognizing a new license plate in advance and train the recognition of a new license plate in advance by using the same. It can also improve the recognition accuracy.

In addition, the present invention is not limited to a license plate image, and may be used to recognize an object, and may generate a vehicle image including a noise pattern and a license plate. By training the generated image, it is possible to improve the license plate or object recognition accuracy.

FIG. 5 illustrates a configuration of a vehicle license plate image generating apparatus according to an embodiment of the present invention, and illustrates a configuration of an apparatus for performing the method of FIGS. 1 to 4.

Referring to FIG. 5, a vehicle license plate image generating apparatus 500 according to an embodiment of the present invention includes a receiver 510, a synthesizer 520, a generator 530, and a database (DB) 540. .

The database 540 is a means for storing all information related to the present invention, and stores a plurality of noise pattern filters, a synthesized license plate image, an algorithm related to the present invention, and the like.

Of course, the database 540 can store not only this but all kinds of data used in the present invention, and this data can be updated.

The generation unit 530 extracts noise patterns included in the license plate image from each of a plurality of license plate images including noises using a pre-learned network, for example, CNN, and applies each of the extracted noise patterns. Create a filter, that is, a noise pattern filter.

Here, the generator 530 may generate a new noise pattern filter by synthesizing the generated noise pattern filters.

The receiver 510 receives a first image of the license plate in order to generate a license plate image including high noise.

The synthesizing unit 520 generates a second image to which at least one noise pattern is applied to the vehicle license plate of the first image by applying at least one noise pattern filter preset or selected to the first image.

Here, the synthesizer 520 may generate a high-quality vehicle license plate image to which the noise pattern filter is applied using a pre-learned network that receives the noise pattern filter and the first image, for example, a DNN or a CNN.

In addition, the synthesis unit 520 may generate a second image to which the noise pattern filter is locally applied by locally applying at least one noise pattern filter to each of the first images, and may apply each of the plurality of noise pattern filters to each other. By sequentially applying the first image, a second image to which a plurality of noise pattern filters are applied may be generated.

Furthermore, the combining unit 520 may generate a second image to which the plurality of noise patterns are applied by applying each of the plurality of noise patterns to a local area set differently with respect to the first image.

Although the description is omitted in the apparatus of FIG. 5, each constituent means constituting FIG. 5 may include all the contents described in FIGS. 1 to 4, which are obvious to those skilled in the art.

The system or apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the systems, devices, and components described in the embodiments may include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable arrays (FPAs). ), A programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions, may be implemented using one or more general purpose or special purpose computers. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For the convenience of understanding, the processing apparatus may be described as one used, but those skilled in the art will appreciate that the processing apparatus includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as parallel processors.

The software may include a computer program, code, instructions, or a combination of one or more of the above, and may configure the processing device to operate as desired, or process independently or collectively. You can command the device. Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. Or may be permanently or temporarily embodied in a signal wave to be transmitted. The software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner. Software and data may be stored on one or more computer readable recording media.

The method according to the embodiments may be embodied in the form of program instructions that may be executed by various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different manner than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the claims which follow.

Claims (10)

Receiving a first image for the license plate; And
Generating a second image to which the noise pattern is applied to the license plate by applying at least one noise pattern to the first image
Including;
Extracting a noise pattern from each of a plurality of images including noises using a pre-learned network, and generating a noise pattern filter to which each of the extracted noise patterns is applied;
More,
Generating the second image
And generating the second image by applying at least one noise pattern filter of the generated noise pattern filters to the first image.
The method of claim 1,
The generating step
And locally applying the noise pattern to the vehicle license plate to generate the second image to which the noise pattern is applied locally.
The method of claim 1,
The generating step
And sequentially applying each of the plurality of noise patterns to the first image to generate the second image to which the plurality of noise patterns are applied.
The method of claim 1,
The generating step
The second image is generated by applying each of the plurality of noise patterns to the first image to the local region differently set with respect to the first image. Vehicle license plate image generation method.
delete A receiver for receiving a first image of the license plate; And
A synthesizer configured to generate a second image to which the noise pattern is applied to the license plate by applying at least one noise pattern to the first image
Including;
A generator that extracts a noise pattern from each of the plurality of images including noises using a pre-learned network, and generates a noise pattern filter to which each of the extracted noise patterns is applied.
More,
The synthesis unit
And generating the second image by applying at least one noise pattern filter among the generated noise pattern filters to the first image.
The method of claim 6,
The synthesis unit
And locally applying the noise pattern to the license plate to generate the second image to which the noise pattern is applied locally.
The method of claim 6,
The synthesis unit
And sequentially applying each of the plurality of noise patterns to the first image, thereby generating the second image to which the plurality of noise patterns have been applied.
The method of claim 6,
The synthesis unit
The second image is generated by applying each of the plurality of noise patterns to the first image to the local region differently set with respect to the first image. Vehicle license plate image generation device.
delete

Images