WO2023246580A1

WO2023246580A1 - Information processing method and apparatus, and electronic device

Info

Publication number: WO2023246580A1
Application number: PCT/CN2023/100157
Authority: WO
Inventors: 杨晓东; 兰荣华
Original assignee: 京东方科技集团股份有限公司; 成都京东方智慧科技有限公司
Priority date: 2022-06-20
Filing date: 2023-06-14
Publication date: 2023-12-28
Also published as: CN115131779A

Abstract

Embodiments of the present application relate to the technical field of computers, and disclose an information processing method and apparatus and an electronic device. The method comprises: obtaining the image information of a license plate to be recognized; and inputting the image information into a preset recognizer to obtain license plate information of the license plate to be recognized, wherein the license plate information at least comprises text information and color information, and the recognizer executes an asymmetric convolutional neural network operation to output the text information and the color information. The embodiments of the present application solve the problem in related art that the recognition efficiency of license plate recognition algorithms is low.

Description

Information processing methods, devices and electronic equipment

Technical field

This application relates to the field of computer technology. Specifically, this application relates to an information processing method, device and electronic equipment.

Background technique

In the field of smart transportation technology, license plate recognition is widely used in various scenarios, such as vehicle positioning, car anti-theft, automated highway speeding supervision, red light electronic police, highway toll stations and other scenarios; license plate recognition is important for maintaining traffic safety and urban security. It is of great significance to prevent traffic jams and realize automated traffic management. In related technologies, license plate recognition algorithms have been widely used in various scenarios. However, license plate recognition algorithms still have the problem of low recognition efficiency in application scenarios. For example, the recognized vehicle often needs to stay for several minutes to be effectively recognized. Or the vehicle to be recognized needs to be kept relatively straight to the image acquisition module and the speed is slow to be effectively recognized. In special cases, the vehicle license plate needs to be photographed with fill-in light, which is very restrictive; therefore, the problem of low recognition efficiency of the license plate recognition algorithm needs to be solved.

Contents of the invention

The embodiment of the present application provides an information processing method to solve the problem of low recognition efficiency of license plate recognition algorithms in related technologies.

Correspondingly, embodiments of the present application also provide an information processing device, an electronic device, and a storage medium to ensure the implementation and application of the above method.

In order to solve the above problems, an embodiment of the present application discloses an information processing method. The method includes: obtaining the image information of the license plate to be recognized; inputting the image information into a preset recognizer to obtain the image information of the license plate to be recognized. License plate information; wherein the license plate information at least includes text information and color information, and the recognizer performs asymmetric convolutional neural network operation processing to output the text information and color information respectively.

Optionally, inputting the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized includes: performing spatial transformation correction processing on the image information through the recognizer to obtain the corrected license plate information. image information; the recognizer performs the asymmetric convolutional neural network operation processing on the corrected image information, and outputs the text information and color information respectively.

Optionally, performing the asymmetric convolutional neural network operation processing on the corrected image information through the identifier and outputting the text information and color information respectively includes: The information is subjected to at least three inverted residual processing operations in sequence; wherein the number of feature extractions for each of the inverted residual processing operations is different; the features extracted by at least three of the inverted residual processing operations are spliced, and all the features are output respectively. Describe text information and color information.

Optionally, the number of feature extractions for each inverse residual processing operation is increased according to the execution order of the at least three inverse residual processing operations.

Optionally, said sequentially performing at least three inverse residual processing operations on the corrected image information includes: sequentially performing at least three inverse residual processing operations on the corrected image information through an asymmetric convolution layer. operation, and during the execution of the last inverse residual processing operation, dimensionality reduction processing is performed on the extracted feature information.

Optionally, obtaining the image information of the license plate to be recognized includes: detecting the license plate to be recognized; acquiring the license plate image including the license plate to be recognized, scaling the license plate image to a resolution of N×M, and obtaining the license plate to be recognized. Recognize the image information of the license plate; where N and M are preset values.

Optionally, before obtaining the image information of the license plate to be recognized, the method further includes: training the recognizer according to the sample license plate information; wherein the sample license plate information at least includes sample text information of the sample license plate and sample Color information; the sample text information includes a first preset number of characters, and the sample color information includes a second preset number of colors.

Optionally, when the number of characters in the sample text information is less than a third preset number, spaces or preset characters are added between characters in the sample text information.

An embodiment of the present application also discloses an information processing device. The device includes: an acquisition module, used to acquire the image information of the license plate to be recognized; and a recognition module, used to input the image information to a preset recognizer to obtain The license plate information of the license plate to be recognized; wherein the license plate information at least includes text information and color information, and the recognizer performs asymmetric convolutional neural network operation processing to output the text information and color information respectively.

Optionally, the recognition module includes: an STN processing sub-module, used to perform spatial transformation correction processing on the image information through the recognizer to obtain corrected image information; an output sub-module, used to use the recognition The device performs the asymmetric convolutional neural network operation processing on the corrected image information, and outputs the text information and color information respectively.

Optionally, the output sub-module includes: an inverse residual processing unit, configured to sequentially perform at least three inverse residual processing operations on the corrected image information; wherein, each of the inverse residual processing operations The number of feature extraction times is different;

A splicing unit is used to splice the features extracted by at least three of the inverse residual processing operations, and output the text information and color information respectively.

Optionally, the splicing unit is configured to sequentially perform at least three inverse residual processing operations on the corrected image information through an asymmetric convolution layer, and during the execution of the last inverse residual processing operation , perform dimensionality reduction processing on the extracted feature information.

Optionally, the acquisition module includes: a detection sub-module, used to detect the license plate to be recognized; acquire the license plate image including the license plate to be recognized, scale the license plate image to a resolution of N×M, and obtain the license plate to be recognized. Recognize the image information of the license plate; where N and M are preset values.

Optionally, the device further includes: a training module for training the recognizer based on sample license plate information; wherein the sample license plate information at least includes sample text information and sample color information of the sample license plate; the sample The text information includes a first preset number of characters, and the sample color information includes a second preset number of colors.

Optionally, in this embodiment of the present application, when the number of characters in the sample text information is less than a third preset number, spaces or preset characters are added between characters in the sample text information.

Embodiments of the present application also disclose an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, one or more of the aspects described in the embodiments of the present application are implemented. Methods.

Embodiments of the present application also disclose a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the method as described in one or more of the embodiments of the present application is implemented. .

An embodiment of the present application also discloses a computer program product, which includes a computer program. When the computer program is executed by a processor, the computer program implements one or more of the methods described in the embodiments of the present application.

The beneficial effects brought by the technical solutions provided by the embodiments of the present application are: in the embodiments of the present application, the image information of the license plate to be recognized is obtained, and the image information is input to a preset recognizer to obtain the license plate of the license plate to be recognized. information; wherein, the recognizer performs asymmetric convolutional neural network operation processing and outputs the text information and color information of the license plate to be recognized respectively; the asymmetric convolution layer is used to reduce the amount of calculation and simultaneously output two feature information of text and color. Further reduce the amount of calculation and improve the recognition accuracy.

Additional aspects and advantages of the embodiments of the application will be set forth in part in the description which follows, and will be apparent from the description, or may be learned by practice of the application.

Description of the drawings

The above and/or additional aspects and advantages of the present application will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

Figure 1 is a flow chart of an information processing method provided by an embodiment of the present application;

Figure 2 is a schematic diagram of a first example provided by the embodiment of the present application;

Figure 3 is a schematic diagram of a second example provided by the embodiment of the present application;

Figure 4 is a schematic diagram of a third example provided by the embodiment of the present application;

Figure 5 is one of the schematic diagrams of the fourth example provided by the embodiment of the present application;

Figure 6 is the second schematic diagram of the fourth example provided by the embodiment of the present application;

Figure 7 is a schematic structural diagram of an information processing device provided by an embodiment of the present application;

FIG. 8 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.

Detailed ways

The embodiments of the present application are described below with reference to the drawings in the present application. It should be understood that the embodiments described below in conjunction with the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application, and do not limit the technical solutions of the embodiments of the present application.

Those skilled in the art will understand that, unless expressly stated otherwise, the singular forms "a", "an", "the" and "the" used herein may also include the plural form. It should be further understood that the terms "comprising" and "including" used in the embodiments of this application mean that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and/or components, but do not exclude Implementation is other features, information, data, steps, operations, elements, components and/or their combinations supported by the technical field. It should be understood that when we refer to an element being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or one element and the other element may be connected to the other element through intervening elements. Establish connections. Additionally, "connected" or "coupled" as used herein may include wireless connections or wireless couplings. The term "and/or" is used herein to indicate at least one of the items defined by the term. For example, "A and/or B" can be implemented as "A", or as "B", or as "A and B"".

In order to make the purpose, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

Referring to Figure 1, an embodiment of the present application provides an information processing method. Optionally, the method is applied to an electronic device. The electronic device may be a license plate recognition device for license plate recognition, or a device including a license plate recognition device. Systems; such systems include vehicle positioning systems, car anti-theft systems, highway speeding automated supervision systems, red light electronic police systems, highway toll station systems, video surveillance systems, vehicle analysis systems, intelligent car finding systems, road analysis systems, etc.

As shown in Figure 1, the method may include the following steps 101 to 102.

Step 101: Obtain the image information of the license plate to be recognized.

Among them, the image information of the license plate to be recognized can come from the shooting of the image acquisition device or the transmission of the client. The image acquisition device can be the image acquisition device of the aforementioned system; the client can be the client of the electronic device; the license plate is a kind of vehicle Important features are an important part of vehicle structured information and have many application requirements in the field of smart transportation. After the electronic device obtains the image information of the license plate to be recognized, it performs information processing on the license plate image to identify the license plate feature information.

Step 102: Input the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein the license plate information at least includes text information and color information, and the recognizer executes an asymmetric convolutional neural network. Network operation processing outputs the text information and color information respectively.

The image information is input to a preset recognizer, and the recognizer performs license plate recognition on the image information to obtain the license plate information of the license plate to be recognized. The license plate information at least includes text information and color information; text information such as text, numbers and letters. For example, usually, the license plate text mainly includes: the abbreviation of the province, 6 digits and letters; the last digit of the special license plate may be a Chinese character , such as "Xue" (coach car), etc.; the color information of the license plate usually includes a limited number of colors, such as blue, yellow, green, white, and black.

Specifically, the recognizer is pre-trained, and the recognizer includes an asymmetric convolutional neural network structure, such as an asymmetric convolution layer, and outputs the text information and color information respectively; wherein, usually, The asymmetric convolution layer, that is, the convolution layer is conv(p×q) and conv(q×p), p and q represent numbers, and p≠q; on the one hand, asymmetric convolution can reduce the amount of calculation; on the other hand, p and q can be set with reference to the length-to-width ratio of the license plate, so that the operation performed by the recognizer is more consistent with the length-to-width characteristics of the license plate, thereby improving the recognition accuracy of the recognizer; for example, if the length-to-width ratio of the license plate is 3: 1, then the convolutional layers are conv(1×3) and conv(3×1).

In addition, the recognizer outputs the text information and color information respectively; that is, the recognizer performs a non- Symmetrical convolutional neural network operation processing outputs text information and color information respectively. In this way, after the recognizer recognizes the two feature information of text and color, there is no need to fuse the two features, further reducing the amount of calculation and improving the recognition speed. ;At the same time, the same recognizer is no longer used to fuse the two features to avoid reducing the accuracy of the recognizer.

In the embodiment of the present application, the image information of the license plate to be recognized is obtained, and the image information is input to a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein, the recognizer executes an asymmetric convolutional neural network In the operation processing, the text information and color information of the license plate to be recognized are output respectively; the amount of calculation is reduced through the asymmetric convolution layer, and the two feature information of text and color are output respectively to further reduce the amount of calculation and improve the recognition accuracy. The embodiments of the present application solve the problem of low recognition efficiency of license plate recognition algorithms in related technologies.

In an optional embodiment, inputting the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized includes: performing spatial transformation correction processing on the image information through the recognizer , obtain corrected image information; perform the asymmetric convolutional neural network operation processing on the corrected image information through the identifier, and output the text information and color information respectively.

As a first example, refer to Figure 2, which shows a specific example of the recognizer in the embodiment of the present application; the recognizer includes: an input module, a spatial transformation correction (STN) module, a backbone ) module as well as text output (plate num) module and color output (color feature) module.

Among them, the input module inputs the image information of the license plate to be recognized to the STN module. The STN module performs spatial transformation and correction processing on the image information, and inputs it to the backbone module for asymmetric convolutional neural network operation processing, and finally outputs it through the text output module. The text information of the license plate to be recognized, and the color information of the license plate to be recognized are output through the color output module.

Spatial Transformer Networks (STN) mainly performs affine transformation on the image information to complete the adaptive correction of the original license plate image. As a second example, see Figure 3, STN processing mainly includes the following steps: The input image passes through the convolution layer for feature extraction: first, preliminary feature extraction is performed through the convolution layer of conv(3×3), and then through Maxpool2×2, Use stride = 2 to perform maximum pooling downsampling to obtain the sampled features; then pass the sampled features through the conv (5×5) convolution layer for feature extraction, and then use Maxpool maximum pooling downsampling to retain the more significant features. features, and finally the affine transformation matrix (1*6 matrix) is regressed through the fully connected layer (Linear), and the adaptive correction of the original image is finally completed through the affine transformation (affine_grid).

In an optional embodiment, the discriminator performs the non-alignment on the corrected image information. Called convolutional neural network operation processing, outputting the text information and color information respectively includes: sequentially performing at least three inverse residual processing operations on the corrected image information; wherein each of the inverse residual processing operations The number of times of feature extraction is different; the features extracted by at least three inverse residual processing operations are spliced, and the text information and color information are output respectively.

Among them, the inverted residual processing operation (inverted Residual) is mainly used for finer-grained feature extraction of image information; as a third example, see Figure 4, the image information transformed by the STN layer first uses convolution for preliminary feature extraction. , followed by three different inverted Residual processing operations to perform fine-grained feature extraction. Among them, the number of feature extractions for each inverted residual processing operation is different. As shown in Figure 4, inverted Residual×3 means that 3 feature extractions are performed; the output data of the mth inverted residual processing operation is used as the m+th The input data of one inverted residual processing operation is used for feature splicing after downsampling. For example, as shown in Figure 4, the output data of the first inverted residual processing operation is used as the third inverse residual processing operation in the direction indicated by arrows S11 and S12. The input data of the quadratic inverse residual processing operation and the downsampling process (downsampling 1) are used for feature splicing. The output data of the second inverse residual processing operation is used as the input data of the third inverse residual processing operation in the direction indicated by arrows S21 and S22, and is used for feature splicing after downsampling processing (downsampling 2).

Among them, the features extracted by each of the inverted residual processing operations are spliced. As shown in Figure 4, the output data of the first inverted residual processing operation follows the direction indicated by arrow S12, and the output data of the second inverted residual processing operation follows the direction indicated by arrow S12. The output data of the operation follows the direction indicated by arrow S22 to retain the features extracted by the previous inverted residual processing operation, and is finally input into the splicing operation for feature splicing; as the number of inverted residual processing operations gradually increases, the image will become smaller , the extracted features pay more and more attention to details, while the previous features retain more general information of the image, and the information output by the first two inverted residual processing operations and the information output by the last inverted residual processing operation are feature spliced and fused. , not only retains the more detailed information of each part of the image, but also retains the most original generalization information of the image, and the accuracy and generalization of the convolutional neural network will be further improved.

In an optional embodiment, the number of feature extractions for each inverted residual processing operation increases according to the execution order of the at least three inverted residual processing operations; for example, as shown in Figure 4, the three inverted residual processing operations The number of feature extraction times is 3, 4, and 6 respectively. As the inverse residual processing operations gradually increase, by increasing the number of feature extraction times, the feature extraction granularity is refined to obtain more feature details.

In an optional embodiment, performing at least three inverse residual processing operations on the corrected image information in sequence includes: performing at least three inverse residual processing operations on the corrected image information through an asymmetric convolution layer. An inverse residual processing operation is performed, and during the execution of the last inverted residual processing operation, dimensionality reduction processing is performed on the extracted feature information.

As a fourth example, see Figures 5 and 6. Figure 5 is a schematic process diagram of the inverse residual processing operations (for example, the first n-1) before the last inverse residual processing operation (for example, the nth). , Figure 6 is a schematic process diagram of the last inverse residual processing operation; among them, in Figure 5, the initial input data is dimensionally reduced in the convolution layer conv(1×1), and then passed through the asymmetric convolution layer conv(3 ×1), the asymmetric convolution layer conv(1×3) for feature extraction, and the Dwise (3×3) convolution layer to separate the network channel correlation and spatial correlation, and finally use the convolution layer conv(1× 1) Overlay (Add) the initial input data to achieve dimensionality increase; among them, the relu layer represents the activation function layer.

In Figure 6, in the last inverted residual processing operation, there is no need to superimpose the initial input data to achieve dimensionality reduction of the feature data. Among them, as shown in Figure 5 and Figure 6, in the Dwise (3×3) convolution layer, the step size of the first n-1 Dwise strides is 1, and the step size of the nth stride is 2; in the Dwise step When the length is 1, the features that can be learned at this scale are repeatedly extracted, the previously learned features are added to the subsequent features for enhancement, and the step size (stride) of Dwise is 2 to control the further downward movement. feature extraction.

In an optional embodiment, obtaining the image information of the license plate to be recognized includes:

Detect the license plate to be recognized; for example, the license plate to be recognized is captured by an image collection device or transmitted by the client;

Obtain the license plate image including the license plate to be recognized, scale the license plate image to a resolution of N×M, and obtain the image information of the license plate to be recognized; where N and M are preset values; for example, the length of the license plate The width ratio is usually 3:1. Set N:M to 3:1 or an approximate value to improve the accuracy of recognition. For example, if the resolution is 32*96, license plate recognition can also be performed when the resolution is low. Reduce the dependence of the license plate recognition algorithm on the environment when the license plate image is taken, such as the dependence on light and scene.

In an optional embodiment, before obtaining the image information of the license plate to be recognized, the method further includes:

The recognizer is trained according to the sample license plate information; wherein the sample license plate information at least includes sample text information and sample color information of the sample license plate; the sample text information includes a first preset number of characters, and the sample color The information includes a second preset number of colors.

Among them, when training the recognizer, in the sample data set, the text recognition of the license plate and the color recognition of the license plate use the same data set, and are trained at the same time to improve the correlation of the two feature recognitions, thereby improving the accuracy.

The sample text information includes a first preset number of characters, such as text, numbers, and letters. For example, usually, the license plate text usually includes 7 characters or 8 characters. For example, the license plate length of an ordinary car is 7 characters. digits, the length of the new energy license plate is 8 digits; the license plate text mainly includes: the abbreviation of the province, 6 digits and letters; the last digit of the special license plate may be a Chinese character, such as "学" (coach vehicle), etc.; the sample color information Including a second preset number of colors, the sample color information, such as the color information of the license plate, usually includes a limited number of colors, such as There are five types: blue, yellow, green, white, and black.

Optionally, during the training process, the label of the license plate text can be a 7-digit or 8-digit number, with the number representing the number of each character; the label of the license plate color can be a label of 0-4, representing 5 colors respectively.

In an optional embodiment, when the number of characters in the sample text information is less than a third preset number, spaces or preset characters are added between characters in the sample text information.

Since the text length of the sample license plate is not uniform during the training of the recognizer, the dimensions of all data need to be aligned during the training of the convolutional neural network. Therefore, the length of all license plates will be a third preset number (for example, 8 bits). Character representation, use special characters to replace the deficiencies, such as adding spaces or preset characters to meet the third preset number; in addition, after adding spaces or preset characters, another list can be used to record the number of each sample license plate The actual length, CTC loss is finally used for training convergence.

In the embodiment of the present application, the image information of the license plate to be recognized is obtained, and the image information is input to a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein, the recognizer executes an asymmetric convolutional neural network Computational processing outputs the text information and color information of the license plate to be recognized respectively, with lower requirements for image information; based on the asymmetric convolutional neural network, it can recognize the license plate information in a variety of complex situations, and is suitable for capturing photos of fast vehicles. scene; in addition, the asymmetric convolution layer is used to reduce the amount of calculations, and at the same time, two feature information of text and color are output to further reduce the amount of calculations and improve the recognition accuracy. The embodiments of this application solve the problem of low recognition efficiency of license plate recognition algorithms in related technologies.

Based on the same principles as the methods provided in the embodiments of the present application, the embodiments of the present application also provide an information processing device. The device is applied to an electronic device. The electronic device may be a license plate recognition device for license plate recognition. , or a system including license plate recognition equipment; such systems include vehicle positioning systems, car anti-theft systems, automated highway speeding supervision systems, red light electronic police systems, highway toll station systems, video surveillance systems, vehicle analysis systems, and intelligent car finding systems. , road analysis system, etc.

As shown in Figure 7, the device includes:

The acquisition module 701 is used to acquire the image information of the license plate to be recognized.

The recognition module 702 is used to input the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein the license plate information at least includes text information and color information, and the recognizer performs asymmetric Convolutional neural network operation processing outputs the text information and color information respectively.

In addition, the recognizer outputs the text information and color information respectively; that is to say, the recognizer performs an asymmetric convolutional neural network operation and outputs the text information and color information respectively. In this way, the recognizer recognizes both text and color. After obtaining the feature information, there is no need to fuse the two features, further reducing the amount of calculation and improving the recognition speed; at the same time, the same recognizer is no longer used to fuse the two features to avoid reducing the accuracy of the recognizer.

Optionally, in this embodiment of the present application, the identification module 702 includes:

The STN processing submodule is used to perform spatial transformation correction processing on the image information through the identifier to obtain corrected image information;

An output submodule is configured to perform the asymmetric convolutional neural network operation processing on the corrected image information through the identifier, and output the text information and color information respectively.

Optionally, in this embodiment of the application, the output sub-module includes:

An inverse residual processing unit, configured to perform at least three inverse residual processing operations on the corrected image information in sequence; wherein the number of feature extractions for each inverted residual processing operation is different;

Optionally, in this embodiment of the present application, the number of feature extraction times for each inverse residual processing operation is increased according to the execution order of the at least three inverse residual processing operations.

Optionally, in this embodiment of the application, the splicing unit is used for:

Perform at least three inverse residual processing operations sequentially on the corrected image information through an asymmetric convolution layer, and perform dimensionality reduction processing on the extracted feature information during the execution of the last inverse residual processing operation. .

Optionally, in this embodiment of the present application, the acquisition module 701 includes:

The detection sub-module is used to detect the license plate to be recognized;

Obtain a license plate image including the license plate to be recognized, scale the license plate image to a resolution of N×M, and obtain image information of the license plate to be recognized; where N and M are preset values.

Optionally, in this embodiment of the present application, the device further includes:

A training module configured to train the recognizer based on sample license plate information; wherein the sample license plate information at least includes sample text information and sample color information of the sample license plate; the sample text information includes a first preset number of characters , the sample color information includes a second preset number of colors.

The information processing device provided by the embodiments of the present application can implement each process implemented in the method embodiments of Figures 1 to 6. To avoid repetition, details will not be described here.

In the information processing device provided by this application, the acquisition module 701 obtains the image information of the license plate to be recognized, and the recognition module 702 inputs the image information into a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein, the recognition The processor performs asymmetric convolutional neural network operation processing and outputs the text information and color information of the license plate to be recognized respectively; the asymmetric convolution layer reduces the amount of calculation and simultaneously outputs the two feature information of text and color to further reduce the amount of calculation and improve recognition. Accuracy.

The information processing device in the embodiment of the present application can execute the information processing method provided in the embodiment of the present application. The implementation principle is similar. The actions performed by each module and unit in the information processing device in the embodiments of the present application are the same as those in the present application. Corresponding to the steps in the information processing method in each embodiment of the application, for a detailed functional description of each module of the information processing device, please refer to the description in the corresponding information processing method shown above, and will not be described again here.

Based on the same principle as the method shown in the embodiment of the present application, the embodiment of the present application also provides an electronic device, which may include but is not limited to: a processor and a memory; a memory for storing a computer program; A processor, configured to execute the information processing method shown in any optional embodiment of this application by calling a computer program. Compared with related technologies, the information processing method provided by this application obtains the image information of the license plate to be recognized, inputs the image information into a preset recognizer, and obtains the license plate information of the license plate to be recognized; wherein, the recognition The processor performs asymmetric convolutional neural network operation processing and outputs the text information and color information of the license plate to be recognized respectively; the asymmetric convolution layer reduces the amount of calculation and simultaneously outputs the two feature information of text and color to further reduce the amount of calculation and improve recognition. Accuracy.

In an optional embodiment, an electronic device is also provided, as shown in Figure 8. The electronic device 8000 shown in Figure 8 includes: a processor 8001 and a memory 8003. Among them, the processor 8001 and the memory 8003 are connected, such as through a bus 8002. Optionally, the electronic device 8000 may also include a transceiver 8004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and/or data reception. It should be noted that in practical applications, the number of transceivers 8004 is not limited to one, and the structure of the electronic device 8000 does not limit the embodiments of the present application.

The processor 8001 can be a CPU (Central Processing Unit, central processing unit), a general-purpose processor, a DSP (Digital Signal Processor, a data signal processor), an ASIC (Application Specific Integrated Circuit, an application-specific integrated circuit), or an FPGA (Field Programmable Gate Array). , field programmable gate array) or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various illustrative logical blocks, modules, and circuits described in connection with this disclosure. The processor 8001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

Bus 8002 may include a path that carries information between the above-mentioned components. The bus 8002 can be a PCI (Peripheral Component Interconnect, Peripheral Component Interconnect Standard) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 8002 can be divided into address bus, data bus, control bus, etc. For ease of presentation, only one thick line is used in Figure 8, but it does not mean that there is only one bus or one type of bus.

The memory 8003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, RAM (Random Access Memory) or other types that can store information and instructions. Dynamic storage devices can also be EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed Optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other media that can be used to carry or store computer programs and can be read by a computer, without limitation here Certainly.

The memory 8003 is used to store computer programs for executing embodiments of the present application, and is controlled by the processor 8001 for execution. The processor 8001 is used to execute the computer program stored in the memory 8003 to implement the steps shown in the foregoing method embodiments.

Among them, electronic devices include but are not limited to: mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PAD (tablet computers), PMP (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc. mobile terminals such as digital TVs, desktop computers, etc. The electronic device shown in FIG. 8 is only an example and should not impose any restrictions on the functions and usage scope of the embodiments of the present application.

Embodiments of the present application provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps and corresponding contents of the foregoing method embodiments can be implemented.

Embodiments of the present application also provide a computer program product, including a computer program. When the computer program is executed by a processor, the steps and corresponding contents of the foregoing method embodiments can be implemented.

The terms "first", "second", "third", "fourth", "1", "2", etc. (if present) in the description and claims of this application and the above-mentioned drawings are used for Distinguishes similar objects without necessarily describing a specific order or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances so that the embodiments of the application described herein can be practiced in other than the order illustrated or described.

It should be understood that although each operation step is indicated by arrows in the flow chart of the embodiment of the present application, the order of implementation of these steps is not limited to the order indicated by the arrows. Unless otherwise specified herein, in some implementation scenarios of the embodiments of the present application, the implementation steps in each flowchart may be executed in other orders according to requirements. In addition, some or all of the steps in each flowchart are based on actual implementation scenarios and may include multiple sub-steps or multiple stages. Some or all of these sub-steps or stages may be executed at the same time, and each of these sub-steps or stages may also be executed at different times. In scenarios with different execution times, the execution order of these sub-steps or stages can be flexibly configured according to needs, and the embodiments of the present application do not limit this.

The above are only optional implementation modes of some implementation scenarios of the present application. It should be pointed out that for those of ordinary skill in the technical field, without departing from the technical concept of the solution of the present application, adopting solutions based on the technical ideas of the present application Other similar implementation means also fall within the protection scope of the embodiments of this application.

Claims

An information processing method that includes:

Obtain the image information of the license plate to be recognized;

Input the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein the license plate information at least includes text information and color information, and the recognizer performs asymmetric convolutional neural network operation processing , output the text information and color information respectively.
The information processing method according to claim 1, said inputting the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized includes:

Perform spatial transformation correction processing on the image information through the identifier to obtain corrected image information;

The recognizer performs the asymmetric convolutional neural network operation processing on the corrected image information, and outputs the text information and color information respectively.
The information processing method according to claim 2, wherein the asymmetric convolutional neural network operation processing is performed on the corrected image information through the identifier, and the text information and color information are respectively output, including:

Perform at least three inverse residual processing operations sequentially on the corrected image information; wherein the number of feature extractions for each inverse residual processing operation is different;

The features extracted by at least three inverse residual processing operations are spliced, and the text information and color information are output respectively.
According to the information processing method of claim 3, the number of feature extraction times for each inverse residual processing operation is increased according to the execution order of the at least three inverse residual processing operations.
The information processing method according to claim 3, wherein at least three inverse residual processing operations are performed on the corrected image information in sequence, including:

Perform at least three inverse residual processing operations sequentially on the corrected image information through an asymmetric convolution layer, and perform dimensionality reduction processing on the extracted feature information during the execution of the last inverse residual processing operation. .
The information processing method according to claim 1, said obtaining the image information of the license plate to be recognized includes:

Detect license plates to be recognized;

Obtain a license plate image including the license plate to be recognized, scale the license plate image to a resolution of N×M, and obtain image information of the license plate to be recognized; where N and M are preset values.
The information processing method according to claim 1, before obtaining the image information of the license plate to be recognized, the method further includes:

The recognizer is trained according to the sample license plate information; wherein the sample license plate information at least includes sample Sample text information and sample color information of the license plate; the sample text information includes a first preset number of characters, and the sample color information includes a second preset number of colors.
According to the information processing method of claim 7, when the number of characters in the sample text information is less than a third preset number, spaces or preset characters are added between characters in the sample text information.
An information processing device, including:

The acquisition module is used to obtain the image information of the license plate to be recognized;

A recognition module, configured to input the image information to a preset recognizer to obtain the license plate information of the license plate to be recognized; wherein the license plate information at least includes text information and color information, and the recognizer performs an asymmetric roll The product neural network operation process is used to output the text information and color information respectively.
An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method of any one of claims 1 to 8 is implemented.
A computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the method of any one of claims 1 to 8 is implemented.