Big data image-text recognition system and method based on artificial intelligence

Abstract

The invention relates to the technical field of image-text recognition, in particular to a big data image-text recognition system and method based on artificial intelligence, and the system comprises an element identification module, a text keyword extraction module, a mapping network construction module, an image-text perception adjustment module, a decision diagram construction module and a format adaptability adjustment module. In the invention, through deep pixel analysis and visual feature matching, accurate analysis of the relationship between the object in the image and the visual label is realized, the accuracy of image-text data identification is improved, the specific association between the image content and the related text data is accurately understood, and the constructed image network is subjected to semantic association calculation, so that the accuracy of image-text data identification is improved. According to the method, the semantic model can dynamically adapt to new data input, the decision path of the image-text content is optimized, the image-text recognition efficiency is effectively improved, the classification decision is more accurate, the image-text recognition quality and speed in the big data environment are optimized, and the processing performance and the application flexibility are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of image and text recognition technology, and in particular to a big data image and text recognition system and method based on artificial intelligence. Background Art

[0002] The field of image and text recognition technology involves using computer vision to identify and interpret text information in images. It relies on deep learning models, such as convolutional neural networks (CNN) and recurrent neural networks (RNN), to process and analyze image data, identify text elements in images, and understand the semantic content of text. Image and text recognition systems are used in automated content indexing, interactive experiences in augmented reality environments, and automated document analysis and content management systems, playing a key role in advertising analysis, security monitoring, and digital media management.

[0003] Among them, the AI ​​big data image and text recognition system involves advanced image and text recognition solutions, which use AI technology to process batch data sets to identify and parse image and text information. It has a wide range of uses, for example, it can automatically tag image content on social media platforms, provide search engine optimized image search results, or automate data entry and classification in legal and medical record management. By integrating AI and big data technology, it can improve recognition accuracy and processing speed, and support data-intensive application scenarios.

[0004] Existing image and text recognition technologies rely on traditional deep learning models to process image data. Although the models perform well in image and text recognition tasks, they require batch data training and are not sensitive enough when processing batch data sets that change in real time. Especially in diverse application scenarios, such as social media platforms or legal and medical record management, traditional technologies have difficulty adapting to new data types or sudden data changes quickly, resulting in reduced recognition accuracy or insufficient processing speed. For example, in the field of security monitoring, if the system cannot update the recognition algorithm in real time to adapt to new scenarios, important events will be missed or misreported, indicating that existing technologies have limitations in adaptability and scalability in dynamic data environments and need to be further optimized to meet the needs of highly variable modern applications. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a big data image and text recognition system and method based on artificial intelligence.

[0006] In order to achieve the above purpose, the present invention adopts the following technical scheme: a big data image and text recognition system based on artificial intelligence includes:

[0007] The element identification module performs pixel analysis on big data images, analyzes the relationship between objects in the image and visual labels through visual feature matching, and generates a visual element index;

[0008] The text keyword extraction module uses artificial intelligence to scan the input big data text data according to the visual element index, performs word frequency analysis and semantic parsing, associates the visual element index with the text, and establishes an initial semantic link;

[0009] The mapping network construction module constructs a graph network from the initialized semantic links, wherein the nodes represent the graph and text elements, and generates an optimized semantic network by calculating the semantic relevance and adjusting the connection strength between the nodes;

[0010] The image and text perception adjustment module uses the optimized semantic network to monitor the semantic changes of big data images and texts in real time through artificial intelligence, adjusts the semantic network, and obtains a dynamically adjusted semantic model;

[0011] The decision diagram construction module extracts decision-making data from the dynamically adjusted semantic model, adjusts the decision path according to changes in the image and text content, optimizes the image and text recognition efficiency, and generates a classification decision diagram;

[0012] The format adaptability adjustment module uses the classification decision diagram to evaluate the consistency and efficiency of big data image and text recognition through artificial intelligence, adjusts the adaptability of the image and text format, and generates an image and text recognition log.

[0013] As a further solution of the present invention, the visual element index includes object boundaries, color features and shape descriptions identified in big data images, the initialized semantic links include association tags, frequencies and contextual fits between keywords and visual elements, the optimized semantic network includes strength-adjusted node connections, node attribute adjustment data and key path displays, the dynamically adjusted semantic model includes real-time updated semantic strength parameters, semantic evolution tracking records and adjusted node influence assessments, the classification decision diagram includes decision node configurations, path selection logic and image-text matching metrics, and the image-text recognition log includes recorded format adjustment history, recognition efficiency optimization and format compatibility evaluation results.

[0014] As a further solution of the present invention, the element identification module includes:

[0015] The pixel analysis submodule analyzes the pixels of the big data image, uses convolutional neural network technology to identify and distinguish the objects and background in the image, extracts the basic elements of the image content, and identifies the spatial positioning information of the object to obtain analytical positioning data;

[0016] The feature matching submodule identifies the visual features of the big data image based on the analysis and positioning data, matches the shape, color, and texture of the object with the preset visual tags, associates the object features with the visual tags, and records the matching association strength to obtain feature association data;

[0017] The formatting processing submodule records the matching status of each object and the associated visual label according to the feature association data, performs formatting processing, synchronizes the consistency of big data image recognition, and obtains a visual element index.

[0018] As a further solution of the present invention, the formula of the convolutional neural network technology is as follows:

[0019] F l =σ((W l ·k+D l )*F l-1 α+b l +C l )

[0020] Among them, F l is the feature map of the lth layer, W l is the convolution kernel weight of the lth layer, D l is the dynamic adjustment coefficient, k is the learning rate, * is the convolution operation, F l-1 is the feature map of the previous layer, α is the attenuation factor, b l is the bias term, C l is the additional bias, and σ is the nonlinear activation function.

[0021] As a further solution of the present invention, the text keyword extraction module includes:

[0022] The data scanning submodule uses the visual element index and artificial intelligence to scan the input big data text data, identify key information in the text, extract text fragments associated with the visual elements, perform initial text segmentation and labeling, and obtain key text fragments;

[0023] The word frequency analysis submodule uses the text key fragments to perform word frequency analysis on the big data text data, identifies the keywords and phrases in the text by statistically analyzing the frequency of the keywords, and associates the keywords with the tags in the visual index to obtain keyword frequency data;

[0024] The semantic analysis submodule performs semantic analysis through the keyword frequency data, analyzes the semantic relationship between words, combines the visual element index, associates the text keywords with the visual elements, and obtains the initialization semantic link.

[0025] As a further solution of the present invention, the mapping network construction module includes:

[0026] The graph network initialization submodule uses the initialized semantic links to create a graph network, where each node represents a graph and text element. Basic semantic attributes and connection weights are assigned through node initialization, node attribute marking and connection settings are performed, and an initialized graph network structure is obtained;

[0027] The relevance calculation submodule performs semantic relevance calculation between nodes based on the initialization graph network structure, measures the semantic fit and relevance between multiple nodes, adjusts the connection weights, and performs semantic fit update and weight adjustment to obtain adjusted node relevance data;

[0028] The network optimization submodule performs optimization adjustment of the network structure through the adjusted node association data, optimizes the semantic recognition efficiency of the overall network by adjusting the connection strength between nodes and the network layout, and obtains an optimized semantic network.

[0029] As a further solution of the present invention, the image-text perception adjustment module includes:

[0030] The real-time monitoring submodule uses artificial intelligence based on the optimized semantic network to perform real-time monitoring of big data graphic content, capture semantic changes in graphic streams, and record dynamic features of graphic data in real time, perform data stream analysis and change capture, and obtain graphic change records;

[0031] The change assessment submodule performs semantic change analysis of big data graphics based on the graphics and text change records, identifies key semantic change points and patterns, analyzes the impact of the changes on the entire network, and evaluates the semantic criticality and analytical impact of the change points to obtain semantic change analysis results;

[0032] The weight adjustment submodule uses the semantic change analysis results to adjust the connection strength and semantic weight between nodes, match the semantic changes captured in real time, optimize the semantic connection and node response strategy, and obtain a dynamically adjusted semantic model.

[0033] As a further solution of the present invention, the decision diagram construction module includes:

[0034] The data extraction submodule identifies key image-text association data and change trends from the dynamically adjusted semantic model, classifies and organizes semantic features and structural data that affect decision-making, and obtains a decision-making data set;

[0035] The path adjustment submodule adjusts the decision path according to the decision-making data set, and performs real-time update and logic reconstruction of the decision path according to the real-time changes and semantic dynamic adjustment of the big data graphic content to obtain the adjusted decision path;

[0036] The efficiency optimization submodule optimizes the image and text recognition efficiency based on the adjusted decision path, optimizes the processing time and recognition efficiency of big data images and texts, performs performance adjustment and efficiency optimization, and obtains a classification decision diagram.

[0037] As a further solution of the present invention, the format adaptability adjustment module includes:

[0038] The image and text evaluation submodule performs consistency evaluation of the image and text content of the big data according to the classification decision diagram, analyzes the degree of image and text matching and the consistency level, measures the matching degree between the image and text, records the areas and factors of identification deviation, and obtains the consistency evaluation result;

[0039] The efficiency calculation submodule uses the consistency evaluation results to evaluate the efficiency of big data image and text recognition, calculate the speed and response time of image and text processing, record efficiency bottlenecks and potential risk points, and obtain efficiency evaluation results;

[0040] The adaptability adjustment submodule uses the efficiency evaluation result to adjust the adaptability of the big data image and text format, optimizes the big data image and text processing flow, and obtains the image and text recognition log by optimizing the format and structure of the image and text display.

[0041] A method for big data image and text recognition based on artificial intelligence, wherein the method for big data image and text recognition based on artificial intelligence is performed based on the above-mentioned big data image and text recognition system based on artificial intelligence, and comprises the following steps:

[0042] S1: Scan the big data image at the pixel level, identify and record the color value and brightness level of each pixel, calculate the color distribution and texture characteristics of the differentiated area, and generate a visual element index;

[0043] S2: Scanning the big data text according to the visual element index, recording the frequency of each word, and analyzing the correlation between the word and the visual element in the image data to generate a semantic correlation index;

[0044] S3: Create a graph network through the semantic association index, where each node represents a graphic element, calculate the connection strength between nodes, and adjust the structure of the graph network with reference to the context association of the vocabulary and the fit of the visual elements to generate an optimized semantic network;

[0045] S4: using the optimized semantic network, monitoring the changes of graphics and texts in the real-time data stream, dynamically adjusting the nodes and connections in the network, optimizing the relationship between nodes according to the real-time update of the graphics and text content, and generating a dynamic semantic model;

[0046] S5: Based on the dynamic semantic model, the decision path of the image and text content is evaluated, the image and text recognition path is optimized, the decision path is adjusted according to the recognition efficiency and consistency, the adaptability of the image and text format is adjusted, and the image and text recognition log is generated.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, through in-depth pixel analysis and visual feature matching, accurate analysis of the relationship between objects and visual labels in the image is achieved, the accuracy of image and text data recognition is improved, and the system is allowed to more accurately understand the specific relationship between image content and related text data. By performing word frequency analysis and semantic parsing, image data and text content are associated, and the contextual understanding ability of data is further enhanced, so that the system can extract richer and more accurate information from multidimensional data sources. The constructed graph network adjusts the connection strength between nodes in real time through semantic relevance calculation, increases the network's response sensitivity to image and text changes, and ensures that the semantic model can dynamically adapt to new data inputs. The decision path of image and text content is optimized, effectively improving the efficiency of image and text recognition, making classification decisions more accurate, optimizing the quality and speed of image and text recognition in a big data environment, and significantly improving processing performance and application flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a system flow chart of the present invention;

[0050] Figure 2 It is a schematic diagram of the system framework of the present invention;

[0051] Figure 3 It is a flow chart of the element identification module of the present invention;

[0052] Figure 4 It is a flow chart of the text keyword extraction module of the present invention;

[0053] Figure 5 A flowchart of a mapping network construction module of the present invention;

[0054] Figure 6 This is a flow chart of the image and text perception adjustment module of the present invention;

[0055] Figure 7 Building a modular flow chart for the decision diagram of the present invention;

[0056] Figure 8 It is a flow chart of the format adaptability adjustment module of the present invention;

[0057] Fig. 9 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0060] See also Figure 1 to Figure 2 The present invention provides a technical solution, a big data image and text recognition system based on artificial intelligence, comprising:

[0061] The element identification module performs pixel analysis on big data images, identifies visual features of color and shape, analyzes the relationship between objects in the image and visual labels through visual feature matching, and generates a visual element index;

[0062] The text keyword extraction module uses artificial intelligence to scan the input big data text data based on the visual element index, perform word frequency analysis and semantic parsing, mark the vocabulary and semantic dense points in the text, associate the visual element index with the text, and establish an initial semantic link;

[0063] The mapping network construction module constructs a graph network from the initial semantic links. The nodes represent the elements of the graph and text. By calculating the semantic relevance, the connection strength between nodes is adjusted, the overall network structure is optimized, and the dynamic changes at the semantic level are matched to generate an optimized semantic network.

[0064] The image and text perception adjustment module uses the optimized semantic network and artificial intelligence to monitor the semantic changes of big data images and texts in real time. By adjusting the semantic network, it matches the dynamic changes of big data image and text content and obtains a dynamically adjusted semantic model.

[0065] The decision diagram construction module extracts decision-making data from the dynamically adjusted semantic model, adjusts the decision path according to the changes in the image and text content, optimizes the image and text recognition efficiency and response speed, and generates a classification decision diagram;

[0066] The format adaptability adjustment module uses classification decision diagrams and artificial intelligence to evaluate the consistency and efficiency of big data image and text recognition. It adjusts the adaptability of image and text formats, handles diverse structures and format changes, and generates image and text recognition logs.

[0067] The visual element index includes the object boundaries, color features and shape descriptions identified in big data images. The initialized semantic links include the association tags, frequencies and contextual fits between keywords and visual elements. The optimized semantic network includes strength-adjusted node connections, node attribute adjustment data and key path display. The dynamically adjusted semantic model includes real-time updated semantic strength parameters, semantic evolution tracking records and adjusted node influence assessments. The classification decision diagram includes decision node configuration, path selection logic and image-text matching metrics. The image-text recognition log includes recorded format adjustment history, recognition efficiency optimization and format compatibility evaluation results.

[0068] See also Figure 2 , 3 , the element identification module includes:

[0069] The pixel analysis submodule analyzes the pixels of the big data image, uses convolutional neural network technology to identify and distinguish the objects and background in the image, extracts the basic elements of the image content, and identifies the spatial positioning information of the object. The execution process of obtaining the analysis positioning data is as follows;

[0070] The pixel analysis submodule analyzes the pixels of the big data image and extracts effective pixel information. The system constructs an entropy value array by calculating the entropy value of each pixel interval. The average value of the array is calculated as the initial entropy threshold. The threshold is used to screen representative pixel intervals so as to focus on processing the parts with large amounts of information in the image. The system gradually lowers the entropy threshold, decreasing a preset step size each iteration, so as to gradually eliminate pixel intervals with low amounts of information. The process continues until the change in entropy value is lower than a predetermined small threshold or reaches a stable state to determine the optimal entropy threshold, which not only improves the efficiency of data processing, but also enhances the accuracy of image analysis. It can effectively distinguish the foreground and background in the image, accurately extract the basic elements of the image content, including the spatial positioning information of the object, and obtain analytical positioning data.

[0071] The formula for the convolutional neural network technique is as follows:

[0072] F l =σ((W l ·k+D l )*F l-1 α+b l +C l )

[0073] Among them, F lis the feature map of the lth layer, W l is the convolution kernel weight of the lth layer, D l is the dynamic adjustment coefficient, k is the learning rate, * is the convolution operation, F l-1 is the feature map of the previous layer, α is the attenuation factor, b l is the bias term, C l is the additional bias, and σ is the nonlinear activation function.

[0074] The execution flow is as follows:

[0075] Calculate the dynamic adjustment coefficient D of each layer based on the input data l The coefficients are based on the activation map of the previous layer and the weights to be trained in the current layer, and are dynamically adjusted through the mini-batch gradient descent method. The weights of each layer are W l After multiplying by the learning rate k, l Adding can make the model adaptively adjust the weights according to different data characteristics during the training process, and compare the adjusted weights with the feature map F of the previous layer. l-1 Perform convolution operation and multiply by attenuation factor α, which helps balance the influence between different layers, plus bias term b l and additional bias C l After that, the feature map F of the current layer is calculated by the nonlinear activation function σ l , which not only enhances the model’s ability to learn features, but also improves the generalization ability of the overall network.

[0076] The feature matching submodule analyzes the positioning data, identifies the visual features of the big data image, matches the shape, color, and texture of the object with the preset visual label, associates the object features with the visual label, and records the matching association strength to obtain the execution process of the feature association data as follows;

[0077] The feature matching submodule further identifies the visual features in the image based on the analysis of the positioning data and defines a set of visual feature parameters, including shape, color, and texture. By comparing with the preset visual labels, the visual features of each object in the image are identified and the matching degree with each visual label is calculated. In order to accurately quantify the matching results, a weighted scoring method is adopted to obtain the comprehensive visual feature score of each object by weighting the matching degree of shape, color, and texture. It not only ensures a high degree of matching accuracy, but also enhances the efficiency and consistency of image recognition through systematic data processing, and obtains feature association data. The formula used is:

[0078] R=θ·S+λ·C+ψ·T

[0079] In the formula, R is the comprehensive visual feature score, which indicates the total correlation strength of the match, θ, λ, and ψ are the weights of shape, color, and texture, respectively, and S, C, and T are the matching scores of the corresponding visual features in the image.

[0080] The formatting processing submodule records the matching of each object and the associated visual label according to the feature association data, and performs formatting processing to synchronize the consistency of big data image recognition, and obtains the execution process of the visual element index as follows;

[0081] The formatting submodule organizes and formats the data according to the feature association data, and designs a special data structure to store the matching relationship between each identified object and the visual label. By formatting the data, it ensures that all image data can be standardized after recognition, improves the readability and usability of the data, and synchronously processes multiple image files to optimize the processing speed and reduce system resource consumption, which not only ensures the efficiency of data processing, but also improves the overall performance of the image processing system and obtains the visual element index.

[0082] See also Figure 2 , 4 ,The text keyword extraction module includes:

[0083] The data scanning submodule uses the visual element index and artificial intelligence to scan the input big data text data, identify the key information in the text, and extract the text fragments associated with the visual elements, perform initial text segmentation and labeling, and obtain the execution process of the key text fragments as follows;

[0084] The data scanning submodule uses the visual element index and combines artificial intelligence technology to identify and extract key information in the text, identify entities in the text, such as names, places, and proper nouns, and associate the entities with relevant image data in the visual element index. The text is segmented and marked, and the text is divided into independent fragments, and each fragment is semantically labeled based on the relevance of the label to the visual element. Not only does it extract text fragments directly associated with the visual element, but it also further processes and analyzes the text and outputs key text fragments. The formula used is:

[0085]

[0086] Among them, S represents the scan score, T i is the i-th term in the text, V is the visual element index, AI represents the artificial intelligence recognition function, and k is the number of terms in the text segment.

[0087] The word frequency analysis submodule uses text key fragments to perform word frequency analysis on big data text data. It statistically analyzes the frequency of keywords, identifies keywords and phrases in the text, and associates keywords with tags in the visual index. The execution process of obtaining keyword frequency data is as follows;

[0088] The word frequency analysis submodule uses key text segments to perform frequency analysis on the words in the segments. It uses statistical methods to calculate the frequency of occurrence of each keyword and phrase in the text and identify high-frequency keywords in the text. Keywords are matched with tags in the visual element index to form an association between words and images. This not only provides detailed language features of text data, but also enhances the association between text and visual content. It is suitable for subsequent semantic analysis and provides a rich data foundation to obtain keyword frequency data. The formula used is:

[0089]

[0090] Among them, F k represents the average frequency of keywords, w j is the jth keyword, freq(w j ) is the frequency of occurrence of the word in the text, and n is the number of keywords.

[0091] The semantic analysis submodule performs semantic analysis through keyword frequency data, analyzes the semantic relationship between words, combines visual element indexes, associates text keywords with visual elements, and obtains the execution process of initializing semantic links as follows;

[0092] The semantic analysis submodule uses keyword frequency data to conduct in-depth semantic relationship analysis, not only analyzing the direct semantic associations between terms, but also exploring connections in a wider context. By combining terms with visual element indexes, a comprehensive semantic link between text keywords and corresponding visual elements can be established. This improves the depth of understanding of text content, making the conversion from text to visual content smoother and more intuitive, and obtaining an initial semantic link using the formula:

[0093]

[0094] Where L is the strength of the semantic link, c i is the ith keyword, sem(c i ,V) represents word c i The semantic relevance between the visual element index V, freq(c i ) is the word c i The frequency of n is the number of keywords.

[0095] See also Figure 2 , 5 , the mapping network building blocks include:

[0096] The graph network initialization submodule uses the initialization semantic links to create a graph network. Each node represents a graph element. Basic semantic attributes and connection weights are assigned through node initialization, and node attribute marking and connection settings are performed. The execution flow of the initialization graph network structure is as follows;

[0097] The graph network initialization submodule uses the initial semantic links to create a new graph network, where each node represents a graphic element. Basic semantic attributes are assigned to each node, including relevant vocabulary, concepts, and key image features. The connection weights between each node are initially set based on semantic similarity and content relevance. Through an automated tagging process, the system identifies and annotates node attributes, and configures preliminary connections between nodes. To ensure that the graph network can accurately map the complex relationship between text and visual data, after the settings are completed, the initial graph network structure is generated, using the formula:

[0098]

[0099] Among them, G represents the structural score of the graph network, W ij is the connection weight between nodes i and j, A i and A j are the semantic attribute strengths of nodes i and j respectively, and p is the total number of nodes.

[0100] The relevance calculation submodule performs semantic relevance calculation between nodes based on the initialization graph network structure, measures the semantic fit and relevance between multiple nodes, adjusts the connection weights, and performs semantic fit update and weight adjustment. The execution process of obtaining the adjusted node relevance data is as follows;

[0101] The relevance calculation submodule performs semantic relevance calculations between nodes based on the initial graph network structure. It measures and analyzes the semantic fit between multiple nodes, including evaluating the shared attributes, related concepts, and consistency between graphs and texts. According to the analysis results, the connection weights are adjusted to reflect the actual semantic connection strength between nodes. The semantic fit is updated and the weights are adjusted periodically to ensure that the network can adapt to changes in content and new data inputs, and the adjusted node relevance data is output using the formula:

[0102]

[0103] Where C represents the average correlation score, sin(N k ,N l ) is node N k and N l The semantic similarity between kl is the connection weight between nodes, and q is the number of node pairs evaluated.

[0104] The network optimization submodule performs optimization adjustment of the network structure through the adjusted node correlation data, and optimizes the semantic recognition efficiency of the overall network by adjusting the connection strength and network layout between nodes. The execution process of the optimized semantic network is as follows;

[0105] The network optimization submodule performs network structure optimization through the adjusted node association data, focusing on optimizing the structure and efficiency of the entire semantic network by dynamically adjusting the strength of the connections between nodes and the network layout. Taking into account new semantic data and user feedback, the network is adjusted to improve the accuracy and speed of semantic recognition, which not only improves the responsiveness of the network, but also ensures that the overall network can process and interpret batch data sets more effectively, and obtains the optimized semantic network, using the formula:

[0106]

[0107] Among them, O represents the optimized network efficiency, U st is the utility function of node s under configuration t, D st is the corresponding demand satisfaction, S is the set of nodes in the network, and r is the number of network configurations.

[0108] See also Figure 2 , 6 , the image and text perception adjustment module includes:

[0109] The real-time monitoring submodule uses artificial intelligence based on the optimized semantic network to conduct real-time monitoring of big data graphic content, capture semantic changes in graphic flow, and record the dynamic characteristics of graphic data in real time, perform data flow analysis and change capture, and obtain the execution process of graphic change records as follows;

[0110] The real-time monitoring submodule uses advanced artificial intelligence technology to continuously monitor the graphic content in big data based on the optimized semantic network, analyze the semantic changes in the graphic flow in real time, and identify and record the dynamic characteristics of the graphic data. Through specially designed algorithms, it can capture and analyze key information in the data flow in real time, such as emergencies, trend changes, or the evolution of important concepts. It can not only respond to immediate data changes, but also update the network status according to real-time data to ensure that the semantic network continues to reflect the latest data situation and generate graphic change records. The formula used is:

[0111]

[0112] Among them, M represents the effect measurement of monitoring, δV i is the change in the i-th data point in the time series, and l is the total number of data points analyzed during the monitoring period.

[0113] The change assessment submodule performs semantic change analysis of big data graphics based on the graphics and text change records, identifies key semantic change points and patterns, analyzes the impact of the changes on the entire network, and evaluates the semantic criticality and analytical impact of the change points. The execution process of obtaining the semantic change analysis results is as follows;

[0114] The change assessment submodule performs in-depth semantic change analysis of big data graphic content based on graphic change records, identifies and analyzes key semantic change points and patterns, and explores the impact of changes on the entire network structure and function. By using statistical and machine learning techniques, it is possible to evaluate the semantic criticality of each change point and analyze the specific impact of the change on network operations. It helps determine temporary changes and changes in long-term trends, provides scientific decision-making basis, and obtains semantic change analysis results. The formula used is:

[0115]

[0116] Among them, E represents the change assessment score, β j is the influence weight of change point j, ΔS j is the semantic change of the change point, and f is the number of change points evaluated.

[0117] The weight adjustment submodule uses the semantic change analysis results to adjust the connection strength and semantic weight between nodes, match the semantic changes captured in real time, optimize the semantic connection and node response strategy, and obtain the execution process of the dynamically adjusted semantic model as follows;

[0118] The weight adjustment submodule uses the results of semantic change analysis to adjust the structure of the network, and adjusts the connection strength and semantic weight between nodes in real time to match the captured semantic changes, optimize semantic connections and node response strategies, and not only improve the adaptability and response speed of the network, but also ensure that the network structure is continuously optimized as the data flow changes. The adjustment process takes into account the overall performance of the network and the performance of a single node to ensure that the network can efficiently process the latest semantic information and obtain a dynamically adjusted semantic model. The formula used is:

[0119]

[0120] Among them, R represents the network adjustment responsiveness, γ k is the adjustment sensitivity of node k, Ψ k is the expected state of node k after adjustment, Ω k is the current state of node k, and s is the total number of nodes in the network.

[0121] See also Figure 2 , 7 , the decision graph building blocks include:

[0122] The data extraction submodule identifies key image-text association data and change trends from the dynamically adjusted semantic model, classifies and organizes the semantic features and structural data that affect decision-making, and obtains the decision-making data set. The execution process is as follows;

[0123] The data extraction submodule identifies key image-text related data and change trends from the dynamically adjusted semantic model, and classifies and organizes semantic features and structural data that have a significant impact on decision-making. Using advanced data analysis technology, it identifies and separates key information points, such as important events, trend changes, or the increase in the frequency of keyword appearances. It can not only extract valuable information from the data stream, but also systematically classify and organize the information to form a decision-making data set. The formula used is:

[0124]

[0125] Where D represents the quality score of the decision-making dataset, χ i is the importance weight of the identified i-th semantic feature, ξ i is the data integration degree of the feature, and v is the number of key semantic features to be identified.

[0126] The path adjustment submodule adjusts the decision path according to the decision-making data set, and updates and reconstructs the decision path in real time according to the real-time changes and semantic dynamic adjustments of the big data graphic content. The execution process of the adjusted decision path is as follows;

[0127] The path adjustment submodule adjusts the decision path according to the decision-making data set. Taking into account the real-time changes and semantic dynamics of the big data graphic content, the decision path is updated and logically reconstructed in real time. By evaluating the key information and trend changes in the data set, the decision logic and strategy are adjusted in a timely manner to respond to changes in the external environment. Dynamic adjustment ensures that the decision path always remains in the optimal state, reflects the latest data analysis and business needs, and generates the adjusted decision path. The formula used is:

[0128]

[0129] Among them, P represents the efficiency index of the adjusted decision path, λ j is the adjustment sensitivity of decision point j, δ j is the amount of logic change, and g is the number of nodes in the decision path.

[0130] The efficiency optimization submodule optimizes the image and text recognition efficiency based on the adjusted decision path, optimizes the processing time and recognition efficiency of big data images and texts, performs performance adjustment and efficiency optimization, and obtains the following execution flow of the classification decision diagram;

[0131] The efficiency optimization submodule optimizes the image and text recognition efficiency based on the adjusted decision path, and uses advanced performance adjustment technology to optimize the processing time and recognition efficiency of big data images and texts. By evaluating the operating efficiency and processing time of each node, the efficiency bottleneck is identified, and the system is adjusted accordingly. This includes measures such as improving data processing algorithms, increasing resource allocation, and adjusting task priorities, and obtaining a classification decision diagram. The formula used is:

[0132]

[0133] Among them, O represents the overall efficiency optimization result, t k is the processing time of task k, σ l is the efficiency improvement factor of task l, ω l is the corresponding weight factor, s is the number of tasks to be optimized, and K is the set of all tasks.

[0134] See also Figure 2 , 8 , the format adaptation module includes:

[0135] The image and text evaluation submodule performs consistency evaluation of the image and text content of big data according to the classification decision diagram, analyzes the degree of image and text matching and the consistency level, measures the matching degree between the image and text, records the areas and factors of identification deviation, and obtains the execution process of the consistency evaluation result as follows;

[0136] The image and text evaluation submodule performs consistency evaluation of big data image and text content based on the classification decision diagram, and uses advanced analysis technology to deeply analyze the degree of image and text matching and the consistency level to ensure accurate expression and matching of image and text content. By calculating the matching degree between the image and text, it is possible to identify specific areas and influencing factors of recognition deviation, such as inconsistent image and text descriptions, image quality problems, or text interpretation errors. It helps to identify key points for improvement and improve the overall consistency of the content. It includes not only quantitative analysis of matching degree, but also specific improvement suggestions and strategies to provide a basis for subsequent optimization and adjustment, and obtain consistency evaluation results. The formula used is:

[0137]

[0138] Among them, C A represents the consistency evaluation score, α i is the matching importance weight of the i-th image-text pair, μ i is the matching metric of the image-text pair, and u is the number of image-text pairs evaluated.

[0139] The efficiency calculation submodule uses the consistency evaluation results to evaluate the efficiency of big data image and text recognition, calculate the speed and response time of image and text processing, record efficiency bottlenecks and potential risk points, and obtain the execution process of the efficiency evaluation results as follows;

[0140] The efficiency calculation submodule uses the consistency evaluation results to evaluate the overall efficiency of big data image recognition, comprehensively calculates the speed and response time of image processing, and identifies efficiency bottlenecks and potential risk points. By systematically recording and analyzing data, it is possible to identify potential efficiency obstacles and propose specific optimization measures. Measures include improving data processing algorithms, increasing resource allocation, or optimizing task scheduling strategies, providing important data support and decision-making basis for further improving system performance, and obtaining efficiency evaluation results. The formula used is:

[0141]

[0142] Among them, E C represents the efficiency evaluation score, r j is the processing speed of task j, t k is the response time of task k, and v is the total number of processed tasks.

[0143] The adaptability adjustment submodule uses the efficiency evaluation results to adjust the adaptability of the big data image and text format and optimize the big data image and text processing flow. By optimizing the format and structure of the image and text display, the execution flow of the image and text recognition log is obtained as follows;

[0144] The adaptability adjustment submodule uses the efficiency evaluation results to adjust the adaptability of the big data image format, focusing on optimizing the big data image processing process, and improving the overall data processing efficiency and user experience by optimizing the format and structure of the image display. By analyzing the image recognition log and efficiency evaluation results, the image format is adjusted to better adapt to different usage scenarios and needs. The optimized image format will be more intuitive and easier to understand, while maintaining efficient data processing capabilities, and the image recognition log is obtained using the formula:

[0145]

[0146] Among them, A O represents the adaptive optimization score, β m is the adaptive weight for adjusting m, ΔF m is the effect of the format change, and h is the number of parameters to be adjusted.

[0147] See also Fig. 9 , a method for big data image and text recognition based on artificial intelligence, the method for big data image and text recognition based on artificial intelligence is performed based on the above-mentioned big data image and text recognition system based on artificial intelligence, and includes the following steps:

[0148] S1: Scan the big data image at the pixel level, identify and record the color value and brightness level of each pixel, calculate the color distribution and texture characteristics of the differentiated area, and generate a visual element index;

[0149] S2: Scan the big data text according to the visual element index, record the frequency of each word, and analyze the correlation between the word and the visual elements in the image data to generate a semantic correlation index;

[0150] S3: Create a graph network through semantic association indexing. Each node represents a graphic element. The connection strength between nodes is calculated. The structure of the graph network is adjusted with reference to the context association of the vocabulary and the fit of the visual elements to generate an optimized semantic network.

[0151] S4: Use the optimized semantic network to monitor the changes of graphics and texts in the real-time data stream, dynamically adjust the nodes and connections in the network, optimize the relationship between nodes according to the real-time update of the graphics and text content, and generate a dynamic semantic model;

[0152] S5: Based on the dynamic semantic model, evaluate the decision path of the image and text content, optimize the image and text recognition path, adjust the decision path according to the recognition efficiency and consistency, adjust the adaptability of the image and text format, and generate the image and text recognition log.

[0153] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A big data image and text recognition system based on artificial intelligence, characterized in that: The system comprises: The element identification module performs pixel analysis on big data images, analyzes the relationship between objects in the image and visual labels through visual feature matching, and generates a visual element index; The text keyword extraction module uses artificial intelligence to scan the input big data text data according to the visual element index, performs word frequency analysis and semantic parsing, associates the visual element index with the text, and establishes an initial semantic link; The mapping network construction module constructs a graph network from the initialized semantic links, wherein the nodes represent the graph and text elements, and generates an optimized semantic network by calculating the semantic relevance and adjusting the connection strength between the nodes; The image and text perception adjustment module uses the optimized semantic network to monitor the semantic changes of big data images and texts in real time through artificial intelligence, adjusts the semantic network, and obtains a dynamically adjusted semantic model; The decision diagram construction module extracts decision-making data from the dynamically adjusted semantic model, adjusts the decision path according to changes in the image and text content, optimizes the image and text recognition efficiency, and generates a classification decision diagram; The format adaptability adjustment module uses the classification decision diagram to evaluate the consistency and efficiency of big data image and text recognition through artificial intelligence, adjusts the adaptability of the image and text format, and generates an image and text recognition log.

2. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The visual element index includes the object boundaries, color features and shape descriptions identified in the big data image, the initialized semantic links include the association tags, frequencies and contextual fits between keywords and visual elements, the optimized semantic network includes strength-adjusted node connections, node attribute adjustment data and key path displays, the dynamically adjusted semantic model includes real-time updated semantic strength parameters, semantic evolution tracking records and adjusted node influence assessments, the classification decision diagram includes decision node configurations, path selection logic and image-text matching metrics, and the image-text recognition log includes recorded format adjustment history, recognition efficiency optimization and format compatibility evaluation results.

3. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The element identification module includes: The pixel analysis submodule analyzes the pixels of the big data image, uses convolutional neural network technology to identify and distinguish the objects and background in the image, extracts the basic elements of the image content, and identifies the spatial positioning information of the object to obtain analytical positioning data; The feature matching submodule identifies the visual features of the big data image based on the analysis and positioning data, matches the shape, color, and texture of the object with the preset visual tags, associates the object features with the visual tags, and records the matching association strength to obtain feature association data; The formatting processing submodule records the matching status of each object and the associated visual label according to the feature association data, performs formatting processing, synchronizes the consistency of big data image recognition, and obtains a visual element index.

4. The big data image and text recognition system based on artificial intelligence according to claim 3 is characterized in that: The formula of the convolutional neural network technique is as follows: F l =σ((W l ·k+D l )*F l-1 ·a+b l +C l ) Among them, F l is the feature map of the lth layer, W l is the convolution kernel weight of the lth layer, D l is the dynamic adjustment coefficient, k is the learning rate, * is the convolution operation, F l-1 is the feature map of the previous layer, α is the attenuation factor, b l is the bias term, C l is the additional bias, and σ is the nonlinear activation function.

5. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The text keyword extraction module includes: The data scanning submodule uses the visual element index and artificial intelligence to scan the input big data text data, identify key information in the text, extract text fragments associated with the visual elements, perform initial text segmentation and labeling, and obtain key text fragments; The word frequency analysis submodule uses the text key fragments to perform word frequency analysis on the big data text data, identifies the keywords and phrases in the text by statistically analyzing the frequency of the keywords, and associates the keywords with the tags in the visual index to obtain keyword frequency data; The semantic analysis submodule performs semantic analysis through the keyword frequency data, analyzes the semantic relationship between words, combines the visual element index, associates the text keywords with the visual elements, and obtains the initialization semantic link.

6. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The mapping network building module includes: The graph network initialization submodule uses the initialized semantic links to create a graph network, where each node represents a graph and text element. Basic semantic attributes and connection weights are assigned through node initialization, node attribute marking and connection settings are performed, and an initialized graph network structure is obtained; The relevance calculation submodule performs semantic relevance calculation between nodes based on the initialization graph network structure, measures the semantic fit and relevance between multiple nodes, adjusts the connection weights, and performs semantic fit update and weight adjustment to obtain adjusted node relevance data; The network optimization submodule performs optimization adjustment of the network structure through the adjusted node association data, optimizes the semantic recognition efficiency of the overall network by adjusting the connection strength between nodes and the network layout, and obtains an optimized semantic network.

7. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The image and text perception adjustment module includes: The real-time monitoring submodule uses artificial intelligence based on the optimized semantic network to perform real-time monitoring of big data graphic content, capture semantic changes in graphic streams, and record dynamic features of graphic data in real time, perform data stream analysis and change capture, and obtain graphic change records; The change assessment submodule performs semantic change analysis of big data graphics based on the graphics and text change records, identifies key semantic change points and patterns, analyzes the impact of the changes on the entire network, and evaluates the semantic criticality and analytical impact of the change points to obtain semantic change analysis results; The weight adjustment submodule uses the semantic change analysis results to adjust the connection strength and semantic weight between nodes, match the semantic changes captured in real time, optimize the semantic connection and node response strategy, and obtain a dynamically adjusted semantic model.

8. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The decision diagram building module includes: The data extraction submodule identifies key image-text association data and change trends from the dynamically adjusted semantic model, classifies and organizes semantic features and structural data that affect decision-making, and obtains a decision-making data set; The path adjustment submodule adjusts the decision path according to the decision-making data set, and performs real-time update and logic reconstruction of the decision path according to the real-time changes and semantic dynamic adjustment of the big data graphic content to obtain the adjusted decision path; The efficiency optimization submodule optimizes the image and text recognition efficiency based on the adjusted decision path, optimizes the processing time and recognition efficiency of big data images and texts, performs performance adjustment and efficiency optimization, and obtains a classification decision diagram.

9. The big data image and text recognition system based on artificial intelligence according to claim 1 is characterized in that: The format adaptability adjustment module includes: The image and text evaluation submodule performs consistency evaluation of the image and text content of the big data according to the classification decision diagram, analyzes the degree of image and text matching and the consistency level, measures the matching degree between the image and text, records the areas and factors of identification deviation, and obtains the consistency evaluation result; The efficiency calculation submodule uses the consistency evaluation results to evaluate the efficiency of big data image and text recognition, calculate the speed and response time of image and text processing, record efficiency bottlenecks and potential risk points, and obtain efficiency evaluation results; The adaptability adjustment submodule uses the efficiency evaluation result to adjust the adaptability of the big data image and text format, optimizes the big data image and text processing flow, and obtains the image and text recognition log by optimizing the format and structure of the image and text display.

10. A big data image and text recognition method based on artificial intelligence, characterized in that: The big data image and text recognition system based on artificial intelligence according to any one of claims 1 to 9 comprises the following steps: Scan big data images at the pixel level, identify and record the color value and brightness level of each pixel, calculate the color distribution and texture characteristics of differentiated areas, and generate a visual element index; According to the visual element index, the big data text is scanned, the frequency of each word is recorded, and the relevance between the word and the visual element in the image data is analyzed to generate a semantic association index; By using the semantic association index, a graph network is created, each node represents a graphic element, the connection strength between nodes is calculated, and the structure of the graph network is adjusted with reference to the context association of the vocabulary and the fit of the visual elements to generate an optimized semantic network; Using the optimized semantic network, monitor the changes of graphics and texts in the real-time data stream, dynamically adjust the nodes and connections in the network, optimize the relationship between nodes according to the real-time update of the graphics and text content, and generate a dynamic semantic model; Based on the dynamic semantic model, the decision path of the image and text content is evaluated, the image and text recognition path is optimized, the decision path is adjusted according to the recognition efficiency and consistency, the adaptability of the image and text format is adjusted, and the image and text recognition log is generated.

Images