Image-based character recognition method, device and equipment combining RPA and AI

Abstract

The present disclosure proposes an image-based character recognition method, device and equipment combining RPA and AI, the method comprising: using RPA (Robotic Process Automation) to obtain a character image, using AI (Artificial Intelligence) to determine the multidimensional feature information of the character image, determining the reference position information of the character image, processing the multidimensional feature information according to the reference position information to obtain target feature information, and identifying the character to be identified from the character image according to the target feature information. Through the present disclosure, the multidimensional feature information of the character image can be processed according to the reference position information to obtain more accurate target feature information. When the character to be identified is identified from the character image based on the target feature information, the accuracy of the representation of the feature of the character to be identified in the character image can be effectively guaranteed, and the character recognition effect and recognition accuracy can be effectively improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to an image-based character recognition method, device and equipment combining RPA (Robotic Process Automation) and AI (Artificial Intelligence). Background Art

[0002] Robotic Process Automation (RPA), referred to as Robotic Process Automation, uses specific "robot software" to simulate human operations on computers and automatically execute process tasks according to rules.

[0003] Artificial Intelligence (AI) is a technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence.

[0004] In the related art, when characters in an image are recognized, contrast features are usually selected by referring to image classification.

[0005] In this way, the reference image classification selects contrast features, which is prone to image representation confusion, resulting in poor character recognition in the image. Summary of the invention

[0006] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0007] To this end, the purpose of the present disclosure is to propose an image-based character recognition method, device, electronic device and storage medium that combines RPA and AI, which can process the multi-dimensional feature information of the character image according to the reference position information to obtain more accurate target feature information. When the character to be recognized is identified from the character image based on the target feature information, the representation accuracy of the feature of the character to be recognized in the character image can be effectively guaranteed, thereby effectively improving the character recognition effect and recognition accuracy.

[0008] The first aspect of the present disclosure proposes an image-based character recognition method combining RPA and AI, including: using robotic process automation (RPA) to acquire a character image; using artificial intelligence (AI) to determine multidimensional feature information of the character image, wherein the feature information is used to describe the features of the character to be recognized in the character image based on corresponding dimensions; determining reference position information of the character image; processing the multidimensional feature information according to the reference position information to obtain target feature information, wherein the number of dimensions of the target feature information is less than the number of dimensions of the multidimensional feature information; and recognizing the character to be recognized from the character image according to the target feature information.

[0009] In one implementation, the determining the reference position information of the character image includes: determining the relative position information of the character to be recognized in the character image, and using the relative position information as the reference position information.

[0010] In one embodiment, the use of artificial intelligence AI to determine the multi-dimensional feature information of the character image includes: using the AI ​​to determine the size feature information of the character to be recognized in the character image; determining the morphological feature information of the character to be recognized in the character image; determining the depth feature information of the character to be recognized in the character image; and using the size feature information, the morphological feature information, and the depth feature information together as the multi-dimensional feature information.

[0011] In one embodiment, processing the multidimensional feature information according to the reference position information to obtain target feature information includes: fusing the reference position information and the multidimensional feature information to obtain fused feature information; and sampling the fused feature information to obtain target feature information.

[0012] In one embodiment, the fusing the reference position information and the multi-dimensional feature information to obtain the fused feature information includes: fusing the reference position information into the size feature information of the multi-dimensional feature information to obtain the fused feature information.

[0013] In one implementation, the sampling process of the fused feature information to obtain the target feature information includes: determining a required sampling scale; and upsampling and / or downsampling the fused feature information according to the required sampling scale to obtain the target feature information.

[0014] In one embodiment, the fused feature information includes: height feature information and width feature information; wherein, upsampling and / or downsampling the fused feature information according to the required sampling scale to obtain the target feature information includes: upsampling and / or downsampling the height feature information based on the required sampling scale to obtain first target feature information; and / or upsampling and / or downsampling the width feature information based on the required sampling scale to obtain second target feature information, wherein the first target feature information and / or the second target feature information is used as the target feature information.

[0015] In one embodiment, identifying the character to be identified from the character image based on the target feature information includes: inputting the target feature information into a character recognition model to obtain a character sequence feature that matches the target feature information, wherein the character recognition model is trained using the self-supervised learning method in the AI; matching the character to be identified from the character image based on the character sequence feature.

[0016] The image-based character recognition device combining RPA and AI proposed in the second aspect of the embodiment of the present disclosure includes: an acquisition module, used to acquire a character image using robotic process automation (RPA); a first determination module, used to determine multidimensional feature information of the character image using artificial intelligence (AI), wherein the feature information is used to describe the features of the character to be recognized in the character image based on corresponding dimensions; a second determination module, used to determine reference position information of the character image; a processing module, used to process the multidimensional feature information according to the reference position information to obtain target feature information, wherein the number of dimensions of the target feature information is less than the number of dimensions of the multidimensional feature information; and a recognition module, used to recognize the character to be recognized from the character image according to the target feature information.

[0017] In one implementation, the second determination module is specifically configured to determine relative position information of the character to be recognized in the character image, and use the relative position information as the reference position information.

[0018] In one embodiment, the first determination module is specifically used to: use the AI ​​to determine the size feature information of the character to be recognized in the character image; determine the morphological feature information of the character to be recognized in the character image; determine the depth feature information of the character to be recognized in the character image; and use the size feature information, the morphological feature information, and the depth feature information together as the multi-dimensional feature information.

[0019] In one embodiment, the processing module includes: a fusion submodule for fusing the reference position information and the multidimensional feature information to obtain fused feature information; and a processing submodule for sampling the fused feature information to obtain target feature information.

[0020] In one implementation, the fusion submodule is specifically used to fuse the reference position information into the size feature information of the multi-dimensional feature information to obtain the fused feature information.

[0021] In one implementation, the processing submodule is specifically used to: determine a required sampling scale; and perform upsampling and / or downsampling processing on the fused feature information according to the required sampling scale to obtain the target feature information.

[0022] In one embodiment, the fused feature information includes: height feature information and width feature information; wherein the processing submodule is further used to: upsample and / or downsample the height feature information based on the required sampling scale to obtain first target feature information; and / or upsample and / or downsample the width feature information based on the required sampling scale to obtain second target feature information, wherein the first target feature information and / or the second target feature information is used as the target feature information.

[0023] In one embodiment, the recognition module is specifically used to: input the target feature information into a character recognition model to obtain a character sequence feature that matches the target feature information, wherein the character recognition model is trained using the self-supervised learning method in the AI; and match the character to be recognized from the character image based on the character sequence feature.

[0024] The electronic device proposed in the third aspect embodiment of the present disclosure includes: at least one processor and a memory; the memory stores computer execution instructions; at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the image-based character recognition method combining RPA and AI proposed in the first aspect embodiment of the present disclosure.

[0025] The computer-readable storage medium proposed in the fourth aspect embodiment of the present disclosure stores computer execution instructions. When the processor executes the computer execution instructions, the image-based character recognition method combining RPA and AI proposed in the first aspect embodiment of the present disclosure is implemented.

[0026] The advantages or beneficial effects of the above technical solution include at least: obtaining character images by adopting robotic process automation RPA, determining the multidimensional feature information of the character image by adopting artificial intelligence AI, determining the reference position information of the character image, processing the multidimensional feature information according to the reference position information to obtain target feature information, and identifying the characters to be identified from the character image according to the target feature information; being able to process the multidimensional feature information of the character image according to the reference position information to obtain more accurate target feature information; when the characters to be identified are identified from the character image based on the target feature information, the representation accuracy of the features of the characters to be identified in the character image can be effectively guaranteed, and the character recognition effect and recognition accuracy can be effectively improved.

[0027] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present disclosure will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments according to the present disclosure and should not be regarded as limiting the scope of the present disclosure.

[0029] Figure 1 It is a flowchart of an image-based character recognition method combining RPA and AI proposed in one embodiment of the present disclosure;

[0030] Figure 2 is a schematic diagram of a graphic verification code in an embodiment of the present disclosure;

[0031] Figure 3 is a flowchart of an image-based character recognition method combining RPA and AI proposed in another embodiment of the present disclosure;

[0032] Figure 4 is a flowchart of an image-based character recognition method combining RPA and AI proposed in another embodiment of the present disclosure;

[0033] Figure 5 is a schematic diagram of a character recognition process in an embodiment of the present disclosure;

[0034] Figure 6 is a structural diagram of an image-based character recognition device combining RPA and AI proposed in one embodiment of the present disclosure;

[0035] Figure 7is a structural diagram of an image-based character recognition device combining RPA and AI proposed in another embodiment of the present disclosure;

[0036] Figure 8 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0037] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present disclosure, and cannot be understood as limiting the present disclosure.

[0038] The following describes an image-based character recognition method combining RPA and AI according to an embodiment of the present disclosure with reference to the accompanying drawings.

[0039] Figure 1 It is a flowchart of an image-based character recognition method combining RPA and AI proposed in one embodiment of the present disclosure.

[0040] In this embodiment, an image-based character recognition method combining RPA and AI is configured as an image-based character recognition device combining RPA and AI. In this embodiment, the image-based character recognition method combining RPA and AI can be configured in an image-based character recognition device combining RPA and AI. The image-based character recognition device combining RPA and AI can be set in a server, or can also be set in an electronic device, and the embodiments of the present disclosure do not limit this.

[0041] This embodiment takes the image-based character recognition method combining RPA and AI as an example, which is configured in an electronic device, such as a smart phone, a tablet computer, a personal digital assistant, an e-book, and other hardware devices with various operating systems.

[0042] It should be noted that the execution subject of the embodiments of the present disclosure may be, for example, a central processing unit (CPU) in a server or electronic device in terms of hardware, or may be, for example, a related background service in a server or electronic device in terms of software, without limitation.

[0043] In addition, the "recognition" in the embodiments of the present disclosure refers to the process of image-based character recognition that combines robotic process automation (RPA) and artificial intelligence (AI). That is to say, the image-based character recognition process is a fully automated image-based character recognition process, and the image-based character recognition process is also combined with artificial intelligence (AI) to realize automated recognition of characters in images in the field of natural language processing (NLP).

[0044] In the description of this disclosure, the term "robotic process automation (RPA)" refers to the automatic execution of process tasks on a computer according to rules through robotic application software.

[0045] In the description of this disclosure, the term "artificial intelligence AI" refers to the discipline that studies how computers can simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), including both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning, deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0046] In the description of the present disclosure, the term "character image" refers to an image containing characters to be recognized and processed. The character image can be, for example, a graphic verification code image for authorizing login to a system, or can be any type of image containing characters, without limitation.

[0047] In the description of the present disclosure, the term "character to be recognized" refers to a character in a character image that is currently to be recognized.

[0048] In the description of the present disclosure, the term "feature information" refers to data information used to characterize the features of the characters to be recognized in the character image. The multi-dimensional feature information of the character image can be used to describe the features of the characters to be recognized in the character image based on the corresponding dimensions.

[0049] In the description of the present disclosure, the term "reference position information" refers to the position information obtained by vectorizing the position feature information of a character image. The reference position information may be two-dimensional position information obtained by vectorizing the height feature information and width feature information of the character image.

[0050] In the description of the present disclosure, the term "target feature information" refers to feature information obtained by performing dimensionality reduction processing on multidimensional feature information, where the dimension of the target feature information is smaller than the number of dimensions of the multidimensional feature information.

[0051] In the description of the present disclosure, the term "size feature information" is used to describe the size features of the characters to be recognized in the character image. The size feature information may be height feature information and width feature information of the characters to be recognized in the character image.

[0052] In the description of the present disclosure, the term "morphological feature information" is used to describe the morphological features of the character to be recognized in the character image. The morphological position information may be, for example, the type of the character to be recognized, and morphological information such as whether there is flipping.

[0053] In the description of the present disclosure, the term "depth feature information" is used to describe the number of storage bits required to store a character to be recognized in a character image.

[0054] In the description of the present disclosure, the term "relative position information" refers to two-dimensional vector position information used to describe the relative position of a character to be recognized in a character image.

[0055] In the description of the present disclosure, the term "fused feature information" refers to feature information obtained by fusing reference position information and multi-dimensional feature information, and the fused feature information can be used to perform subsequent character recognition processing tasks.

[0056] In the description of the present disclosure, the term "required sampling scale" refers to the scale required when sampling and processing the fused feature information. The required sampling scale can be adaptively configured according to the requirements of the neural network for processing the fused feature information.

[0057] In the description of the present disclosure, the term "first target feature information" refers to feature information obtained by upsampling and / or downsampling the height feature information in the fusion feature information according to the required sampling scale.

[0058] In the description of the present disclosure, the term "second target feature information" refers to feature information obtained by upsampling and / or downsampling width feature information in fusion feature information according to a required sampling scale.

[0059] In the description of the present disclosure, the term "character sequence feature" refers to a sequence feature obtained after performing feature transformation processing on a one-dimensional feature.

[0060] In the description of the present disclosure, the term "character recognition model" refers to a model trained using a self-supervised learning method in artificial intelligence (AI). The character recognition processing model can perform character recognition processing tasks based on character sequence features.

[0061] The present disclosure can be specifically applied to the field of natural language processing (NLP) of artificial intelligence AI, which is a field of computer science, artificial intelligence, and linguistics that focuses on the interaction between computers and human (natural) language.

[0062] For example, in the embodiment of the present disclosure, the process of image-based character recognition based on the full-process automation can realize the full-process automation execution by using robotic process automation RPA to obtain character images, using RPA combined with artificial intelligence AI to determine the multi-dimensional feature information of the character image, determining the reference position information of the character image, processing the multi-dimensional feature information according to the reference position information to obtain target feature information, and identifying the characters to be identified from the character image according to the target feature information.

[0063] like Figure 1 As shown, the image-based character recognition method combining RPA and AI includes:

[0064] S101: Acquire character images using Robotic Process Automation (RPA).

[0065] Among them, Robotic Process Automation (RPA) refers to the automatic execution of process tasks on computers according to rules through robot application software.

[0066] The character image refers to an image containing characters to be recognized and processed, and the character image may be, for example, a graphic verification code image for authorizing a login to a system, or may be any type of image containing characters, without limitation.

[0067] That is to say, an application scenario in an embodiment of the present disclosure may be, for example, using robotic process automation (RPA) to obtain a graphic verification code for an authorized login system, and then using the obtained graphic verification code as a character image, performing image-based character recognition processing on the obtained characters to obtain corresponding character recognition processing results. Alternatively, the image-based character recognition method combining RPA and AI described in an embodiment of the present disclosure may also be applied to any other possible character recognition scenario, without limitation.

[0068] In the embodiments of the present disclosure, when using Robotic Process Automation (RPA) to obtain character images, an application data interface can be pre-configured, and the RPA robot receives a graphic verification code for authorizing login to the system through the application data interface according to a pre-set software operation process, and uses the obtained graphic verification code image as a character image. Alternatively, a data transmission interface can be provided on an image-based character recognition device that combines RPA and AI, and an image containing characters transmitted by other electronic devices can be received through the data transmission interface as the obtained character image. Alternatively, any other possible method can be used to obtain the character image, and there is no limitation to this.

[0069] For example, Figure 2 As shown, Figure 2 It is a schematic diagram of the graphic verification code in the embodiment of the present disclosure. Robotic process automation RPA can be used to obtain the graphic verification code of the authorized login system, and the obtained graphic verification code image can be used as a character image. Then, character recognition processing can be performed on the character image to realize the recognition of the characters to be recognized in the character image.

[0070] S102: Using artificial intelligence AI to determine multi-dimensional feature information of the character image, wherein the feature information is used to describe the features of the character to be recognized in the character image based on corresponding dimensions.

[0071] Among them, the character to be recognized in the character image can be called the character to be recognized, the feature information is data information used to characterize the features of the character to be recognized in the character image, and the multi-dimensional feature information of the character image can be used to describe the features of the character to be recognized in the character image based on the corresponding dimensions.

[0072] After the character image is acquired by using the Robotic Process Automation (RPA) as described above, the embodiment of the present disclosure may use artificial intelligence (AI) to determine the multi-dimensional feature information of the character image.

[0073] In the disclosed embodiment, when artificial intelligence AI is used to determine the multi-dimensional feature information of a character image, the acquired character image can be input into a convolutional neural network (CNN), and the convolutional neural network can be used to perform feature analysis on the character image to obtain the output result of the convolutional neural network, and the output result of the convolutional neural network is used as the multi-dimensional feature information of the character image.

[0074] In other embodiments, the acquired character image can also be input into a self-attention-based visual neural network (Vision Transformer, VIT), and the self-attention-based visual neural network can be used to perform feature analysis on the character image to obtain the output result of the self-attention-based visual neural network as the multi-dimensional feature information of the character image, or any other artificial intelligence AI method can be used to determine the multi-dimensional feature information of the character image, without limitation.

[0075] S103: Determine reference position information of the character image.

[0076] The reference position information refers to the position information obtained by vectorizing the position feature information of the character image. The reference position information may be two-dimensional position information obtained by vectorizing the height feature information and width feature information of the character image.

[0077] In the disclosed embodiment, when determining the reference position information of a character image, a convolutional neural network or a self-attention-based visual neural network can be used to analyze and process the character image to obtain height feature information, width feature information, and depth feature information of the character image. The height feature information and the width feature information can then be vectorized to obtain a two-dimensional position vector, and the obtained two-dimensional position vector can be used as the reference position information.

[0078] S104: Process the multidimensional feature information according to the reference position information to obtain target feature information, wherein the number of dimensions of the target feature information is smaller than the number of dimensions of the multidimensional feature information.

[0079] The target feature information refers to feature information obtained by performing dimensionality reduction processing on multi-dimensional feature information, and the dimension of the target feature information is smaller than the number of dimensions of the multi-dimensional feature information.

[0080] In the disclosed embodiment, after the multi-dimensional feature information of the character image is determined by artificial intelligence AI and the reference position information of the character image is determined, the multi-dimensional feature information can be processed according to the reference position information to obtain the target feature information.

[0081] In the embodiments of the present disclosure, when processing multidimensional feature information according to reference position information to obtain target feature information, the multidimensional feature information can be reduced in dimension according to the reference position information, and the two-dimensional vector position information in the reference position information can be reduced in dimension to expand the two-dimensional vector position information into one-dimensional vector information, so as to process the multidimensional feature information according to the reference position information and obtain the target feature information.

[0082] S105: Identify the characters to be recognized from the character image according to the target feature information.

[0083] Among them, the characters currently to be recognized in the character image can be called characters to be recognized.

[0084] After the multi-dimensional feature information is processed according to the reference position information to obtain the target feature information, the embodiment of the present disclosure can recognize the character to be recognized from the character image according to the target feature information.

[0085] In the disclosed embodiment, when identifying the characters to be identified from the character image based on the target feature information, a self-supervised contrastive learning framework can be pre-trained, and then the parameters of the pre-trained contrastive learning framework can be optimized and adjusted to obtain the optimized and adjusted contrastive learning framework. Then, the character image can be recognized and processed using the optimized and adjusted contrastive learning framework based on the target feature information to obtain the output result of the contrastive learning framework, and the output result of the contrastive learning framework can be used as the characters to be identified from the character image.

[0086] In other embodiments, a multi-layer perceptron (MLP) neural network can be used to perform character recognition processing on the character image according to the target feature information to obtain the analysis and processing results of the multi-layer perceptron neural network, and then the analysis and processing results of the multi-layer perceptron neural network are input into a connectionist temporal classification model (CTC) for encoding processing to obtain the characters output by the connectionist temporal classification model, and the characters output by the connectionist temporal classification model are used as the characters to be recognized from the character image, or any other artificial intelligence AI method can be used to recognize the characters to be recognized from the character image according to the target feature information, without limitation.

[0087] In this embodiment, robot process automation (RPA) is used to acquire character images, artificial intelligence (AI) is used to determine multidimensional feature information of the character image, reference position information of the character image is determined, and the multidimensional feature information is processed according to the reference position information to obtain target feature information. According to the target feature information, the characters to be recognized are identified from the character image. The multidimensional feature information of the character image can be processed according to the reference position information to obtain more accurate target feature information. When the characters to be recognized are recognized from the character image based on the target feature information, the representation accuracy of the features of the characters to be recognized in the character image can be effectively guaranteed, thereby effectively improving the character recognition effect and recognition accuracy.

[0088] Figure 3 It is a flowchart of an image-based character recognition method combining RPA and AI proposed in another embodiment of the present disclosure.

[0089] like Figure 3As shown, the image-based character recognition method combining RPA and AI includes:

[0090] S301: Acquire character images using Robotic Process Automation (RPA).

[0091] The description of S301 can be seen from the above-mentioned embodiment, and will not be repeated here.

[0092] S302: Using AI to determine the size feature information of the character to be recognized in the character image.

[0093] The size feature information is used to describe the size feature of the character to be recognized in the character image. The size feature information may be height feature information and width feature information of the character to be recognized in the character image.

[0094] In the embodiments of the present disclosure, after using Robotic Process Automation (RPA) to acquire a character image, AI may be used to determine the size feature information of the character to be recognized in the character image, and a convolutional neural network may be used to perform feature analysis on the character to be recognized in the character image to obtain height feature information and width feature information output by the convolutional neural network as the size feature information of the character to be recognized in the character image.

[0095] S303: Determine the morphological feature information of the character to be recognized in the character image.

[0096] The morphological feature information is used to describe the morphological features of the character to be recognized in the character image. The morphological position information may be, for example, the type of the character to be recognized and morphological information such as whether there is flipping.

[0097] In the embodiments of the present disclosure, when determining the morphological feature information of the character to be recognized in the character image, the character image can be input into a visual neural network for feature analysis processing to obtain the output result of the visual neural network, and the morphological feature information output by the visual neural network is used as the morphological feature information of the character to be recognized in the character image.

[0098] S304: Determine depth feature information of the character to be recognized in the character image.

[0099] The depth feature information is used to describe the number of storage bits required to store the character to be recognized in the character image.

[0100] In the embodiment of the present disclosure, when determining the depth feature information of the character to be recognized in the character image, the character image can be processed using a depth feature extraction model to obtain the feature information extracted by the depth feature extraction model as the depth feature information of the character to be recognized in the character image.

[0101] S305: Take the size feature information, the shape feature information, and the depth feature information together as multi-dimensional feature information.

[0102] After the embodiment of the present disclosure uses AI to determine the size feature information, morphological feature information and depth feature information of the character to be recognized in the character image, the size feature information, morphological feature information and depth feature information of the character to be recognized in the character image can be used together as multi-dimensional feature information.

[0103] In this embodiment, AI is used to determine the size feature information of the character to be recognized in the character image, determine the morphological feature information of the character to be recognized in the character image, and determine the depth feature information of the character to be recognized in the character image. The size feature information, morphological feature information, and depth feature information are used together as multi-dimensional feature information, thereby obtaining relatively rich multi-dimensional feature information to ensure the comprehensiveness of feature extraction of the character image. Since the multi-dimensional feature information can be used to fuse with the reference position information to become fused feature information for character recognition, the recognition effect of character recognition on the character image can be guaranteed, and the accuracy of character recognition can be effectively improved.

[0104] S306: Determine the relative position information of the character to be recognized in the character image, and use the relative position information as reference position information.

[0105] The relative position information refers to the two-dimensional vector position information used to describe the relative position of the character to be recognized in the character image.

[0106] In the embodiments of the present disclosure, when determining the relative position information of the character to be recognized in the character image, the character image can be processed using a convolutional neural network or a self-attention-based visual neural network to obtain the height feature information, width feature information, and depth feature information of the character image. The height feature information and the width feature information can then be vectorized to obtain a two-dimensional position vector, and the obtained two-dimensional position vector can be used as the relative position information.

[0107] In the disclosed embodiment, after determining the relative position information of the character to be recognized in the character image, the relative position information can be used as reference position information, and then the multi-dimensional feature information can be processed according to the reference position information. The specific description can be found in the subsequent embodiments.

[0108] S307: Fusing the reference position information and the multi-dimensional feature information to obtain fused feature information.

[0109] The fused feature information refers to the feature information obtained by fusing the reference position information and the multi-dimensional feature information. The fused feature information can be used to perform subsequent character recognition processing tasks.

[0110] After the embodiment of the present disclosure uses AI to determine the size feature information, morphological feature information and depth feature information of the character to be recognized in the character image, and determines the relative position information of the character to be recognized in the character image as the reference position information, the reference position information and the multi-dimensional feature information can be fused to obtain fused feature information.

[0111] In the embodiments of the present disclosure, when fusing reference position information and multi-dimensional feature information to obtain fused feature information, an artificial intelligence (AI) processing model can be pre-trained. The pre-trained artificial intelligence (AI) processing model can perform feature fusion tasks, and then the reference position information and multi-dimensional feature information are input into the artificial intelligence (AI) processing model for fusion processing to obtain the fused feature information as the fused feature information.

[0112] S308: Sampling the fused feature information to obtain target feature information.

[0113] In the embodiment of the present disclosure, after the reference position information and the multi-dimensional feature information are fused to obtain the fused feature information, sampling processing may be performed on the fused feature information to obtain the target feature information.

[0114] In the embodiment of the present disclosure, when sampling the fused feature information to obtain the target feature information, a sampling scale can be set in advance, and then the fused feature information can be upsampled or downsampled according to the set sampling scale to obtain the feature information after sampling as the target feature information.

[0115] In other embodiments, when the fused feature information is upsampled or downsampled according to the set sampling scale, the height feature information and width feature information in the fused feature information can be sampled with different synchronizations, that is, the height feature information can be sampled and the width feature information can not be sampled, or the width feature information can be sampled and the height feature information can not be sampled, so as to obtain the processed feature information as the target feature information.

[0116] In this embodiment, the reference position information and the multi-dimensional feature information are fused to obtain the fused feature information, and the fused feature information is sampled to obtain the target feature information. Since the fused feature information is sampled, the amount of data in the subsequent recognition processing tasks can be reduced, and the character recognition processing efficiency can be effectively improved. The reference position information and the multi-dimensional feature information are fused to obtain the target feature information, so that more effective target feature information can be obtained to perform recognition processing of the characters to be recognized in the character image, thereby helping to improve the character recognition processing effect.

[0117] S309: Identify the characters to be recognized from the character image according to the target feature information.

[0118] The description of S309 can be seen from the above-mentioned embodiment, and will not be repeated here.

[0119] In this embodiment, a character image is acquired by using a robot process automation (RPA), and multi-dimensional feature information of the character image is determined by using artificial intelligence (AI), and reference position information of the character image is determined. The multi-dimensional feature information is processed according to the reference position information to obtain target feature information. According to the target feature information, the character to be recognized is identified from the character image. The multi-dimensional feature information of the character image can be processed according to the reference position information to obtain relatively accurate target feature information. When the character to be recognized is recognized from the character image based on the target feature information, the representation accuracy of the feature of the character to be recognized in the character image can be effectively guaranteed, and the character recognition effect and recognition accuracy can be effectively improved. By using AI to determine the size feature information of the character to be recognized in the character image, the morphological feature information of the character to be recognized in the character image is determined, and the depth feature information of the character to be recognized in the character image is determined, and the size feature information, morphological feature information, and depth feature information are combined into a single pixel. The feature information is used together as multi-dimensional feature information, so that richer multi-dimensional feature information can be obtained, ensuring the comprehensiveness of feature extraction of character images. Since the multi-dimensional feature information can be used to fuse with the reference position information to become fused feature information for character recognition, the recognition effect of character recognition on character images can be guaranteed, and the accuracy of character recognition can be effectively improved. The reference position information and the multi-dimensional feature information are fused to obtain the fused feature information, and the fused feature information is sampled to obtain the target feature information. Since the fused feature information is sampled, the amount of data in subsequent recognition processing tasks can be reduced, effectively improving the character recognition processing efficiency. The reference position information and the multi-dimensional feature information are fused to obtain the target feature information, so that more effective target feature information can be obtained for the recognition of the characters to be recognized in the character image, thereby assisting in improving the character recognition processing effect.

[0120] Figure 4 It is a flowchart of an image-based character recognition method combining RPA and AI proposed in another embodiment of the present disclosure.

[0121] like Figure 4 As shown, the image-based character recognition method combining RPA and AI includes:

[0122] S401: Acquire character images using robotic process automation (RPA).

[0123] S402: Using AI to determine the size feature information of the character to be recognized in the character image.

[0124] S403: Determine the morphological feature information of the character to be recognized in the character image.

[0125] S404: Determine depth feature information of the character to be recognized in the character image.

[0126] S405: Take the size feature information, the shape feature information, and the depth feature information together as multi-dimensional feature information.

[0127] S406: Determine the relative position information of the character to be recognized in the character image, and use the relative position information as reference position information.

[0128] The description of S401 - S406 can be referred to the above embodiment for examples, and will not be repeated here.

[0129] S407: Fusing the reference position information into the size feature information of the multi-dimensional feature information to obtain fused feature information.

[0130] In the disclosed embodiment, when the reference position information and the multi-dimensional feature information are fused to obtain the fused feature information, the reference position information may be fused into the size feature information of the multi-dimensional feature information to obtain the fused feature information.

[0131] In the embodiment of the present disclosure, when the reference position information is fused into the size feature information of the multidimensional feature information to obtain the fused feature information, the reference position information and the size feature information of the multidimensional feature information can be input into a pre-trained artificial intelligence AI processing model for fusion processing to obtain the fused feature information as the fused feature information.

[0132] S408: Determine the required sampling scale.

[0133] The required sampling scale refers to the scale required for sampling and processing the fused feature information. The required sampling scale can be adaptively configured according to the requirements of the neural network for processing the fused feature information.

[0134] In the embodiment of the present disclosure, when determining the required sampling scale, the required sampling scale can be configured according to the requirements of the neural network for processing the fused feature information, or a suitable sampling scale can be selected from the pre-set sampling scales as the required sampling scale, and there is no limitation on this.

[0135] After determining the required sampling scale, the embodiment of the present disclosure may perform sampling processing on the fusion feature information according to the required sampling scale.

[0136] S409: upsampling and / or downsampling the fused feature information according to the required sampling scale to obtain target feature information.

[0137] In the embodiment of the present disclosure, after the required sampling scale is determined as described above, the fused feature information can be upsampled and / or downsampled according to the required sampling scale to obtain target feature information, and the size feature information in the fused feature information can be upsampled and / or downsampled according to the required sampling scale to obtain the feature information after upsampled and / or downsampled as the target feature information.

[0138] Optionally, in some embodiments, the fused feature information includes: height feature information and width feature information, wherein, when the fused feature information is upsampled and / or downsampled according to the required sampling scale to obtain the target feature information, the height feature information can be upsampled and / or downsampled based on the required sampling scale to obtain the first target feature information, and / or the width feature information can be upsampled and / or downsampled based on the required sampling scale to obtain the second target feature information, wherein the first target feature information and / or the second target feature information is used as the target feature information, so that the height feature information and / or the width feature information can be upsampled and / or downsampled to obtain the target feature information. Since the required sampling scale is set according to the processing requirements of the neural network for the feature information, the target feature information obtained by the sampling process can adapt to the feature processing requirements of the neural network, thereby ensuring the recognition processing effect of character recognition based on the target feature information.

[0139] The first target feature information refers to feature information obtained by upsampling and / or downsampling the height feature information in the fusion feature information according to the required sampling scale.

[0140] The second target feature information refers to feature information obtained by upsampling and / or downsampling the width feature information in the fusion feature information according to the required sampling scale.

[0141] In the embodiment of the present disclosure, when the fused feature information is upsampled and / or downsampled according to the required sampling scale to obtain the target feature information, the height feature information can be upsampled and / or downsampled based on the required sampling scale to obtain the first target feature information, and / or the width feature information can be upsampled and / or downsampled based on the required sampling scale to obtain the second target feature information.

[0142] In the embodiment of the present disclosure, when the fused feature information is upsampled and / or downsampled according to the required sampling scale to obtain the target feature information, the first target feature information and / or the second target feature information can be used as the target feature information, that is, the first target feature information can be used as the target feature information, or the second target feature information can be used as the target feature information, or the first target feature information and the second target feature information can be used together as the target feature information.

[0143] S410: Inputting the target feature information into a character recognition model to obtain a character sequence feature matching the target feature information, wherein the character recognition model is trained using a self-supervised learning method in AI.

[0144] The character sequence feature refers to the sequence feature obtained after performing feature transformation processing on the one-dimensional feature.

[0145] Among them, the character recognition model refers to the one trained by the self-supervised learning method in artificial intelligence AI. The character recognition processing model can perform character recognition processing tasks according to the characteristics of the character sequence.

[0146] In the disclosed embodiments, the artificial intelligence processing model can be pre-trained by self-supervised learning, that is, the artificial intelligence processing model learns directly from unlabeled data. The artificial intelligence processing model trained by self-supervised learning can perform character recognition processing tasks according to the characteristics of character sequences.

[0147] In the embodiment of the present disclosure, after upsampling and / or downsampling the fused feature information according to the required sampling scale to obtain the target feature information, the target feature information can be input into the character recognition model, and the character recognition model can be used to perform feature transformation processing on the target feature information to obtain the sequence features output by the character recognition model, and the sequence features output by the character recognition model are used as the character sequence features for matching the target feature information.

[0148] S311: Match the characters to be recognized from the character image according to the character sequence features.

[0149] In the disclosed embodiment, after the target feature information is input into the character recognition model to obtain the character sequence features matching the target feature information, the character to be recognized can be matched from the character image according to the character sequence features.

[0150] In the disclosed embodiment, when matching the characters to be recognized from the character image based on the character sequence features, the character sequence features and the character image can be input into the matching processing model, and the character image can be matched and recognized using the matching processing model to obtain the characters output by the matching processing model as the characters to be recognized matched from the character image.

[0151] For example, Figure 5 As shown, Figure 5 It is a schematic diagram of the character recognition process in the embodiment of the present disclosure, which uses robotic process automation RPA to obtain a verification code image as a character image, and then uses a convolutional neural network or a self-attention-based visual neural network to perform feature analysis on the character image to obtain multidimensional feature information of the character image, determine the reference position information of the character image, and then process the multidimensional feature information according to the reference position information to obtain target feature information, and input the target feature information into the character recognition model for feature conversion processing to obtain character sequence features matching the target feature information, and then input the character sequence features into the multi-layer perception neural network and the connection temporal classification model in turn for processing to obtain the characters to be recognized from the character image.

[0152] In this embodiment, a character image is acquired by using robotic process automation (RPA), artificial intelligence (AI) is used to determine the multidimensional feature information of the character image, the reference position information of the character image is determined, the multidimensional feature information is processed according to the reference position information to obtain target feature information, and the character to be identified is identified from the character image according to the target feature information. The multidimensional feature information of the character image can be processed according to the reference position information to obtain relatively accurate target feature information. When the character to be identified is identified from the character image based on the target feature information, the representation accuracy of the feature of the character to be identified in the character image can be effectively guaranteed, and the character recognition effect and recognition accuracy can be effectively improved. The height feature information and / or width feature information are upsampled and / or downsampled to obtain the target feature information. Since the required sampling scale is set according to the processing requirements of the neural network for the feature information, the target feature information obtained by the sampling process can adapt to the feature processing requirements of the neural network, thereby ensuring the recognition processing effect of character recognition according to the target feature information.

[0153] Figure 6 It is a structural diagram of an image-based character recognition device combining RPA and AI proposed in one embodiment of the present disclosure.

[0154] like Figure 6 As shown, the image-based character recognition device 60 combining RPA and AI is applied to the field of natural language processing (NLP), including:

[0155] An acquisition module 601 is used to acquire a character image using a robotic process automation (RPA);

[0156] A first determination module 602 is used to determine multi-dimensional feature information of a character image using artificial intelligence AI, wherein the feature information is used to describe features of a character to be recognized in the character image based on corresponding dimensions;

[0157] The second determination module 603 is used to determine the reference position information of the character image;

[0158] A processing module 604 is used to process the multi-dimensional feature information according to the reference position information to obtain target feature information, wherein the number of dimensions of the target feature information is smaller than the number of dimensions of the multi-dimensional feature information; and

[0159] The recognition module 605 is used to recognize the characters to be recognized from the character image according to the target feature information.

[0160] In some embodiments of the present disclosure, the second determining module 603 is specifically configured to:

[0161] The relative position information of the character to be recognized in the character image is determined, and the relative position information is used as reference position information.

[0162] In some embodiments of the present disclosure, the first determining module 602 is specifically configured to:

[0163] AI is used to determine the size feature information of the character to be recognized in the character image;

[0164] Determine the morphological feature information of the character to be recognized in the character image;

[0165] Determine the depth feature information of the character to be recognized in the character image;

[0166] The size feature information, shape feature information and depth feature information are taken together as multi-dimensional feature information.

[0167] In some embodiments of the present disclosure, Figure 7 As shown, Figure 7 Another embodiment of the present disclosure proposes an image-based character recognition device 60 combining RPA and AI, wherein the processing module 604 includes:

[0168] A fusion submodule 6041 is used to fuse the reference position information and the multi-dimensional feature information to obtain fused feature information;

[0169] The processing submodule 6042 is used to perform sampling processing on the fused feature information to obtain target feature information.

[0170] In some embodiments of the present disclosure, the fusion submodule 6041 is specifically used to:

[0171] The reference position information is fused into the size feature information of the multi-dimensional feature information to obtain fused feature information.

[0172] In some embodiments of the present disclosure, the processing submodule 6042 is specifically used to:

[0173] Determine the sampling scale required;

[0174] The fused feature information is upsampled and / or downsampled according to the required sampling scale to obtain the target feature information.

[0175] In some embodiments of the present disclosure, the fused feature information includes: height feature information and width feature information;

[0176] The processing submodule 6042 is further used for:

[0177] The fused feature information is upsampled and / or downsampled according to the required sampling scale to obtain the target feature information, including:

[0178] Performing upsampling and / or downsampling processing on the height characteristic information based on the required sampling scale to obtain the first target characteristic information; and / or

[0179] The width feature information is upsampled and / or downsampled based on a required sampling scale to obtain second target feature information, wherein the first target feature information and / or the second target feature information is used as the target feature information.

[0180] In some embodiments of the present disclosure, the identification module 605 is specifically used to:

[0181] Input the target feature information into the character recognition model to obtain the character sequence features matching the target feature information, wherein the character recognition model is trained using the self-supervised learning method in AI;

[0182] According to the characteristics of the character sequence, the characters to be recognized are matched from the character image.

[0183] With the above Figures 1 to 5 Corresponding to the image-based character recognition method combining RPA and AI provided in the embodiment, the present disclosure also provides an image-based character recognition device combining RPA and AI. Figures 1 to 5The embodiment provides a corresponding image-based character recognition method combining RPA and AI, so the implementation of the image-based character recognition method combining RPA and AI is also applicable to the image-based character recognition device combining RPA and AI provided in the embodiment of the present disclosure, and will not be described in detail in the embodiment of the present disclosure.

[0184] The functions of each module in each device in the embodiments of the present disclosure can be found in the corresponding description in the above method, and will not be repeated here.

[0185] In this embodiment, robot process automation (RPA) is used to acquire character images, artificial intelligence (AI) is used to determine multidimensional feature information of the character image, reference position information of the character image is determined, and the multidimensional feature information is processed according to the reference position information to obtain target feature information. According to the target feature information, the characters to be recognized are identified from the character image. The multidimensional feature information of the character image can be processed according to the reference position information to obtain more accurate target feature information. When the characters to be recognized are recognized from the character image based on the target feature information, the representation accuracy of the features of the characters to be recognized in the character image can be effectively guaranteed, thereby effectively improving the character recognition effect and recognition accuracy.

[0186] In order to implement the above embodiments, the present disclosure also proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the image-based character recognition method combining RPA and AI proposed in the above embodiments of the present disclosure is implemented.

[0187] Figure 8 FIG. 1 is a block diagram showing a structure of an electronic device according to an embodiment of the present disclosure. Figure 8 As shown, the electronic device 80 includes: a memory 810 and a processor 820, wherein the memory 810 stores a computer program that can be run on the processor 820. When the processor 820 executes the computer program, the image-based character recognition method combining RPA and AI in the above embodiment is implemented. The number of the memory 810 and the processor 820 can be one or more.

[0188] The electronic device 80 further includes:

[0189] The communication interface 830 is used to communicate with external devices and perform data exchange transmission.

[0190] If the memory 810, the processor 820 and the communication interface 830 are implemented independently, the memory 810, the processor 820 and the communication interface 830 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0191] Optionally, in a specific implementation, if the memory 810, the processor 820 and the communication interface 830 are integrated on a chip, the memory 810, the processor 820 and the communication interface 830 can communicate with each other through an internal interface.

[0192] An embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present disclosure when the program is executed by a processor.

[0193] The embodiment of the present disclosure also provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present disclosure.

[0194] The embodiment of the present disclosure also provides a chip, including: an input interface, an output interface, a processor and a memory. The input interface, the output interface, the processor and the memory are connected through an internal connection path. The processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided in the embodiment of the application.

[0195] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the Advanced RISC Machines (ARM) architecture.

[0196] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DR RAM).

[0197] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present disclosure is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0198] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0199] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0200] Any process or method description described in the flowchart or otherwise herein can be understood to represent a module, segment or portion of code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiments of the present disclosure includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

[0201] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0202] It should be understood that the various parts of the present disclosure may be implemented with hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented with software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method may be completed by instructing the relevant hardware through a program, which may be stored in a computer-readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0203] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a disk or an optical disk, etc.

[0204] The above is only a specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. An image-based character recognition method combining RPA and AI, characterized in that: The method comprises: Use Robotic Process Automation (RPA) to obtain character images; Using artificial intelligence AI to determine multi-dimensional feature information of the character image, wherein the feature information is used to describe the features of the character to be recognized in the character image based on the corresponding dimensions; Determining reference position information of the character image; fusing the reference position information and the multi-dimensional feature information to obtain fused feature information; Performing sampling processing on the fused feature information to obtain target feature information, wherein the number of dimensions of the target feature information is smaller than the number of dimensions of the multi-dimensional feature information; and According to the target feature information, the character to be recognized is recognized from the character image.

2. The method according to claim 1, characterized in that The determining of the reference position information of the character image comprises: The relative position information of the character to be recognized in the character image is determined, and the relative position information is used as the reference position information.

3. The method according to claim 2, characterized in that The method of using artificial intelligence (AI) to determine the multi-dimensional feature information of the character image includes: Using the AI ​​to determine the size feature information of the character to be recognized in the character image; Determining morphological feature information of the character to be recognized in the character image; Determining depth feature information of the character to be recognized in the character image; The size feature information, the morphology feature information, and the depth feature information are collectively used as the multi-dimensional feature information.

4. The method according to claim 1, characterized in that The fusing the reference position information and the multi-dimensional feature information to obtain fused feature information includes: The reference position information is fused into the size feature information of the multi-dimensional feature information to obtain the fused feature information.

5. The method according to claim 4, characterized in that The sampling process of the fused feature information to obtain target feature information includes: Determine the sampling scale required; The fused feature information is upsampled and / or downsampled according to the required sampling scale to obtain the target feature information.

6. The method according to claim 5, characterized in that The fused feature information includes: height feature information and width feature information; The upsampling and / or downsampling of the fused feature information according to the required sampling scale to obtain the target feature information includes: Performing upsampling and / or downsampling processing on the height characteristic information based on the required sampling scale to obtain first target characteristic information; and / or The width feature information is upsampled and / or downsampled based on the required sampling scale to obtain second target feature information, wherein the first target feature information and / or the second target feature information is used as the target feature information.

7. The method according to any one of claims 1 to 6, characterized in that: The step of identifying the to-be-identified character from the character image according to the target feature information comprises: Inputting the target feature information into a character recognition model to obtain a character sequence feature matching the target feature information, wherein the character recognition model is trained using a self-supervised learning method in the AI; According to the character sequence features, the characters to be recognized are matched from the character image.

8. An image-based character recognition device combining RPA and AI, characterized in that: The device comprises: An acquisition module, used to acquire character images using robotic process automation (RPA); A first determination module is used to determine multi-dimensional feature information of the character image using artificial intelligence AI, wherein the feature information is used to describe features of the character to be recognized in the character image based on corresponding dimensions; A second determining module, used to determine the reference position information of the character image; A fusion module, used for fusing the reference position information and the multi-dimensional feature information to obtain fused feature information; a processing module, configured to perform sampling processing on the fused feature information to obtain target feature information, wherein the number of dimensions of the target feature information is smaller than the number of dimensions of the multi-dimensional feature information; and A recognition module is used to recognize the character to be recognized from the character image according to the target feature information.

9. The device according to claim 8, characterized in that The second determining module is specifically used to: The relative position information of the character to be recognized in the character image is determined, and the relative position information is used as the reference position information.

10. The device according to claim 9, characterized in that The first determining module is specifically configured to: Using the AI ​​to determine the size feature information of the character to be recognized in the character image; Determining morphological feature information of the character to be recognized in the character image; Determining depth feature information of the character to be recognized in the character image; The size feature information, the morphology feature information, and the depth feature information are collectively used as the multi-dimensional feature information.

11. The device according to claim 8, characterized in that The fusion module is specifically used for: The reference position information is fused into the size feature information of the multi-dimensional feature information to obtain the fused feature information.

12. An electronic device, characterized in that: include: at least one processor and memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the image-based character recognition method combining RPA and AI as described in any one of claims 1-7.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When the processor executes the computer-executable instructions, the image-based character recognition method combining RPA and AI as described in any one of claims 1 to 7 is implemented.

Images