Digital interconnection complex graph reading and identification system

Abstract

The invention provides a digital interconnection complex graph reading and recognition system, and relates to the technical field of complex graph reading and recognition. The system adopts a multi-module architecture and comprises a graph input module, a feature extraction module, a graph simplification module, a recognition module and an output module. The key technology comprises edge detection using a Sobe l operator, parameter optimization based on deep learning automatic adjustment, multi-scale segmentation and adaptive noise reduction processing, and advanced feature extraction of a deep learning algorithm. The system supports various graphic file formats and can output various forms such as graphs, texts and data files. According to the method, the characteristics of the complex graph are extracted and recombined, the complex graph is simplified, read and recognized, the processing speed and quality of the complex graph are remarkably improved, and the method has a wide application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of complex pattern reading and recognition, and in particular to a digitally interconnected complex pattern reading and recognition system. Background Art

[0002] In the field of complex graphic reading and recognition technology, existing systems usually face multiple challenges; for example, complex graphics have diverse shapes, textures and color features, making it difficult for traditional image processing algorithms to accurately extract and recognize these features; the image processing methods commonly used in the prior art often lose important detail information when processing graphic simplification and noise reduction, resulting in inaccurate recognition results.

[0003] In addition, existing graphic recognition systems mostly rely on manually designed features, have low recognition accuracy, and cannot meet the needs of high-precision recognition; especially when processing complex graphics, the performance of these systems is often unsatisfactory. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the deficiencies of the prior art, the present invention provides a digitally interconnected complex graphic reading and recognition system, which solves the problem of how to simplify complex graphics and quickly recognize graphics.

[0006] (II) Technical solution

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a digitally interconnected complex graphic reading and recognition system, including a graphic input module, a feature extraction module, a graphic simplification module, a recognition module and an output module; wherein the graphic input module is used to receive complex graphics uploaded by users and perform standardization processing; the feature extraction module performs feature extraction processing on the standardized complex graphics, extracts key features that are conducive to identifying complex graphics, including shape, texture and color features; the graphic simplification module reorganizes the key features in the feature extraction module through multi-scale segmentation and adaptive noise reduction and reduces redundant information to simplify the complex graphics; the recognition module reads and recognizes the reorganized and simplified graphics, and the recognition module introduces a self-attention mechanism into the convolutional neural network model through a deep learning algorithm, recognizes the graphics, and captures the connection between the graphic features; the output module outputs the recognition results of deep learning and supports multiple output formats, including graphics, text and data files.

[0008] Preferably, the graphic input module enables users to upload or specify graphic files to be processed through a user interaction interface or an API interface; the graphic input module supports multiple graphic file formats to meet the needs of different users and graphic files from different sources; the formats of graphic files include vector graphics file formats, such as SVG, EPS, and common graphic file formats, such as JPEG, PNG, GIF, and BMP; the graphic file is divided into multiple data slices, and each data slice is assigned a unique serial number; the segmentation of the data slice is dynamically adjusted according to the network transmission efficiency; when the data slice begins to be transmitted, the feature extraction module reorganizes the data slice according to the serial number to reconstruct the complete graphic file; during the transmission process, if a data slice is lost or a transmission error occurs, the graphic input module identifies the data slice by the serial number and requests to retransmit the corresponding data slice; after the graphic file is successfully loaded and reorganized, the graphic input module performs preprocessing; first, the graphics are standardized to unify the size and color of the graphics, and then the graphics are processed by the graphics input module. The graphics are scaled and standardized to meet the input size of the deep learning model. Secondly, the color information of the graphics is converted into grayscale information, that is, the grayscale value is calculated through the formula Y=0.299R+0.587G+0.114B, where R, G, and B represent the red, green, and blue colors in the graphics, respectively, and Y represents the converted grayscale value. Converting multi-channel color graphics into single-channel grayscale graphics helps to simplify the subsequent feature extraction process.

[0009] Preferably, after the graphics are scaled and color converted, the feature extraction module uses an edge detection algorithm to identify the edge information of the graphics, thereby locating the contour of the graphics; the edge detection algorithm uses the Sobel operator to calculate the change in the gradient of the graphics, and detects the edge points of the graphics based on the change in the gradient of the graphics; after identifying the edge information, the feature extraction module uses the contour tracking technology to extract the closed contour of the graphics; the contour tracking technology selects an unprocessed edge point as the initial edge point, traverses the edge points in a clockwise order, and connects adjacent edge points to form a complete closed contour; for each closed contour, the feature extraction module calculates its geometric features, including area, perimeter, circularity and rectangularity; the area and perimeter are calculated by integration, while the circularity and rectangularity are evaluated by comparing the similarity of the geometric shape of the contour with the standard circle and rectangle; after completing the extraction of the geometric features, the feature extraction module analyzes the internal shape of the graphics by structural analysis technology The structure of the part is used to identify the texture rules of the graphics; the structural analysis technology obtains the filtered image by applying each filter to the normalized grayscale image; the filtering operation is implemented through convolution calculation, that is, the filter is slid on the image and calculation is performed at each position to extract local texture features; for the filtering results of each direction and frequency, the local energy is calculated, which is achieved by calculating the sum of the squares of each filtering result; the local energy reflects the texture intensity of the graphics in different directions and frequencies; the mean and variance of the filtering results are calculated as the texture features of the graphics; the mean represents the overall texture intensity, and the variance represents the degree of texture change; through statistical features, the texture characteristics of the graphics are effectively described, providing data support for subsequent graphics recognition, classification and processing.

[0010] The edge detection algorithm uses the Sobel operator to identify edge information and applies the Sobel operator to process the grayscale graphics. Calculate the gradient change of the local area to identify the edge; there are two Sobel operators, one is Gx for detecting horizontal edges, and the other is Gy for detecting vertical edges; the gradient of the figure is calculated by the Sobel operator, and the size and direction of the gradient reflect the edge information of the figure; apply the threshold operation to identify the area with a gradient greater than the threshold as the edge; the threshold setting steps are:

[0011] Step 1. Process the image using an initial threshold and evaluate the quality of the result.

[0012] Step 2. Determine whether to increase or decrease the threshold based on the evaluation results;

[0013] Step 3. Adjust the threshold according to the set step size;

[0014] Step 4. Iteration: Repeat steps 1-3 until the image quality stops improving.

[0015] Contour detection is performed on the image to connect the detected edge points into a continuous contour. Contour detection uses the cv2.findContours function to extract closed contours from the edges of the image; the cv2.findContours function accepts the processed graphics as input and returns a list of all detected contours, as well as the hierarchical information of each contour; each contour is represented by a numpy array, which contains the coordinates of each point on the contour; the contour points are used to form a closed contour, and the extracted contours are further processed to calculate their geometric features; the specific steps include calculating the area, perimeter, circularity and rectangularity of each contour; the area is calculated using the cv2.contourArea function, which accepts the contour point array as input and returns the area of ​​the area enclosed by the contour; the perimeter is calculated using the cv2.arcLength function, which also accepts the contour point array as input and chooses whether to close the contour to calculate its perimeter; the circularity is calculated using the relationship between area and perimeter, through circularity = 4π×area / perimeter 2 The circularity is calculated between 0 and 1. The closer the value is to 1, the closer the contour is to a circle. The rectangularity is calculated by fitting the contour into a minimum enclosing rectangle and using the cv2.minAreaRect function to get the width and height of the enclosing rectangle. The rectangularity is defined as the ratio of the area of ​​the minimum enclosing rectangle to the area of ​​the contour.

[0016] Preferably, the graphic simplification module performs multi-scale segmentation processing on the extracted key features. The multi-scale segmentation technology performs multi-scale segmentation on the graphic features by using filters and segmentation algorithms of different scales. The filters at each scale can capture the details and structural information of the graphic features at different scales. At each scale, edge detection segmentation is used to segment the graphic features into different areas by detecting the edge information in the shape features. After completing the multi-scale segmentation, the graphic simplification module performs adaptive denoising processing on the segmented graphic features. The adaptive denoising technology dynamically adjusts the denoising parameters according to the texture features and color features of the graphics, removes the noise information, and retains the edge information, texture, and color information. The adaptive denoising first performs a local statistical characteristic analysis on each feature area after segmentation to obtain the noise level and texture complexity in the area. The denoising parameters are dynamically adjusted according to the local statistical characteristics. The denoising parameters include the filter. The size, strength and number of iterations of the noise reducer are adjusted. Relatively stronger noise reduction parameters are used in areas with high noise levels, and relatively weaker noise reduction parameters are used in areas with high texture complexity to avoid loss of details. Edge details are retained by combining spatial domain and grayscale information through bilateral filtering. On the basis of multi-scale segmentation and adaptive noise reduction, the graphics simplification module reorganizes the processed key features. During the reorganization process, the graphics simplification module deletes redundant information according to the feature similarity after segmentation and noise reduction, retains the main features of the graphics, and simplifies the graphics. By setting a similarity threshold, such as 0.8, the similarity threshold exceeding 0.8 is deleted. The reorganization of features includes the application of feature fusion and feature selection algorithms. Feature fusion is used to fuse feature information at different scales to form a comprehensive feature representation. The feature selection algorithm selects important features and deletes redundant features to ensure that the key information after graphics simplification is retained intact.

[0017] Preferably, the recognition module receives the simplified graphics from the graphics simplification module, and preprocesses the input simplified graphics. The preprocessing steps include graphics normalization, resizing and data enhancement. Graphic normalization normalizes the eigenvalues ​​of the graphics to a fixed range, such as 0 to 1, and resizing scales the graphics to a fixed size required by the model. Data enhancement generates a variety of training samples by random rotation, flipping and cropping. After the preprocessing is completed, the simplified graphics are input into the deep learning model for feature extraction and recognition. The recognition module extracts features through a convolutional neural network model. The convolution layer scans the input image by using multiple filters or convolution kernels. Each filter is responsible for detecting features in the graphics, such as edges, corners or complex texture structures. The filter moves on the input graphics in a sliding window manner, calculates the weighted sum of each local area and the filter weight, and generates a set of feature maps, each feature map corresponding to the response of a filter. Self-attention is introduced in the convolution layer. The force mechanism is used to capture the local features and short-distance dependencies of the feature map, so that the convolutional neural network model considers global information and captures the long-distance dependencies between any two positions in the image, so that its recognition module can read the contextual information of the recognized graphics and easily understand the graphics content; the main purpose of the pooling layer is to reduce the spatial size of the feature map, thereby reducing the amount of calculation of the subsequent layers and increasing the invariance to the displacement of the graphics; the pooling operation uses average pooling to calculate the average value of the area; the pooling layer downsamples the input feature map by calculating the average value of all elements in the covered area, reducing the size of the feature map but retaining important information; the pre-trained model can extract features from the graphics, including edges, textures, and shapes; after the feature extraction is completed, the recognition module inputs the extracted feature vector into the fully connected layer for classification; the fully connected layer maps the feature vector to different category labels through linear transformation and nonlinear activation function, and finally outputs the recognition result of the graphics.

[0018] Preferably, the output module parses the recognition results transmitted by the recognition module, and determines the output content and structure according to the parsing results; the output module converts the parsed recognition results according to the output format option selected by the user; for graphic output, the output module uses the graphic processing library Matplotlib to generate a graphic file with recognition information so that the graphics can be output clearly; for text output, the output module generates a text file containing the recognition results; for data file output, the output module uses the data processing library Pandas or Json to generate a data file in the format selected by the user.

[0019] (III) Beneficial effects

[0020] The present invention provides a digitally interconnected complex graphic reading and recognition system, which has the following beneficial effects:

[0021] 1. The present invention simplifies the complex graphics by extracting the features of the complex graphics, and the system reads and recognizes the simplified graphics, thereby enhancing the processing speed of the system and reducing computing resources.

[0022] 2. The present invention outputs graphic files in various formats through an output module, thereby improving the user's further processing of complex graphics.

[0023] 3. During the loading and preprocessing of graphic files, the present invention improves the efficiency of graphic processing through standardization processing and edge detection, and realizes real-time reading and recognition of complex graphics. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] The image input module allows users to upload or specify the graphic files to be processed through the user interaction interface; the graphic input module supports a variety of graphic file formats to meet the needs of different users and graphics from different sources, including vector graphics file formats or common graphic file formats; the graphic file is divided into multiple data slices, each data slice is assigned a serial number, and when the data slice begins to be transmitted, the feature extraction module reorganizes it according to the serial number to obtain the graphic file; after the graphic file is successfully loaded, the graphic input module performs preprocessing work, including standardization of graphics, unification of graphics size and color; among them, through The graphics are scaled and standardized; the color information of the graphics is converted into grayscale information through Y=0.299R+0.587G+0.114B, and the feature extraction module is used to accurately extract features through scaling and color conversion.

[0026] After the graphics are scaled and color converted, the feature extraction module uses an edge detection algorithm to identify the edge information of the graphics, thereby locating the contour of the graphics; the closed contour is extracted through contour tracking technology, and the geometric features of the contour are calculated, including area, perimeter, circularity, and rectangularity; the internal structure of the graphics is analyzed through structural analysis technology to identify the texture regularity of the graphics; including detecting repeated elements, directionality, and roughness in the graphics, thereby generating texture feature parameters, including regularity, directionality, and roughness.

[0027] The edge detection algorithm uses Sobel to identify edge information and applies the Sobel operator to process the grayscale image. Calculate the gradient change of the local area to identify the edge; there are two Sobel operators, one is Gx for detecting horizontal edges, and the other is Gy for detecting vertical edges; the gradient of the figure is calculated by the Sobel operator, and the size and direction of the gradient reflect the edge information of the figure; apply the threshold operation to identify the area with a gradient greater than the threshold as the edge; the threshold setting steps are:

[0028] Step 1. Process the image using an initial threshold and evaluate the quality of the result.

[0029] Step 2. Determine whether to increase or decrease the threshold based on the evaluation results;

[0030] Step 3. Adjust the threshold according to a certain step size;

[0031] Step 4. Iteration: Repeat steps 1-3 until the image quality stops improving.

[0032] Use contour detection to connect the detected edge points into a continuous contour. Contour detection uses the cv2.findContours function to extract closed contours from the edges. The contour returned by the cv2.findContours function is a numpy array containing the coordinates of each point on the contour. The contour points are used to form a closed contour. The geometric features of the extracted contour, such as area, perimeter, circularity, and rectangularity, are calculated.

[0033] The graphic simplification module performs multi-scale segmentation processing on the extracted key features. The multi-scale segmentation technology performs multi-scale segmentation on the graphic features by using filters and segmentation algorithms of different scales. The filters at each scale can capture the details and structural information of the graphic features at different scales. At each scale, edge detection segmentation is used to segment the graphic features into different regions by detecting the edge information in the shape features. After completing the multi-scale segmentation, the graphic simplification module performs adaptive denoising processing on the segmented graphic features. The adaptive denoising technology dynamically adjusts the denoising parameters according to the texture and color features of the graphics, removes noise information, and retains important features. The adaptive denoising first performs a local statistical characteristic analysis on each feature region after segmentation to obtain the noise level and texture complexity in the region. The denoising parameters are dynamically adjusted according to the local statistical characteristics. The parameters include the size, strength and number of iterations of the filter. Relatively stronger noise reduction parameters are used in areas with high noise levels, and relatively weaker noise reduction parameters are used in areas with high texture complexity to avoid loss of details. The edge details are retained by combining spatial domain and grayscale information through bilateral filtering. On the basis of multi-scale segmentation and adaptive noise reduction, the graphics simplification module reorganizes the processed key features. During the reorganization process, the graphics simplification module deletes redundant information according to the feature similarity after segmentation and noise reduction, retains the main features of the graphics, and simplifies the graphics. The reorganization of features includes the application of feature fusion and feature selection algorithms. Feature fusion is used to fuse feature information at different scales to form a comprehensive feature representation. The feature selection algorithm selects important features and deletes redundant features. After feature reorganization, the complex graphics retains key features and achieves a simplified effect.

[0034] The recognition module receives the simplified graphics from the graphics simplification module and preprocesses the input simplified graphics. The preprocessing steps include image normalization, resizing and data enhancement. Image normalization normalizes the feature values ​​of the graphics to a fixed range, such as 0 to 1, and scales the graphics to the fixed size required by the model according to the size requirements of the model input. Data enhancement generates a variety of training samples through random rotation, flipping and cropping. After the preprocessing is completed, the simplified graphics are input into the deep learning model for feature extraction and recognition. The recognition module extracts features through the pre-trained model of the convolutional neural network in deep learning. The convolution layer scans the input image by using multiple filters or convolution kernels. Each filter is responsible for detecting features in the graphics, such as edges, corners or complex texture structures. The filter moves on the input graphics in a sliding window manner, calculates the weighted sum of each local area and the filter weight, and generates a set of feature maps. Each feature map corresponds to the response of a filter. In the convolution layer, The self-attention mechanism is introduced to capture the local features and short-distance dependencies of the feature map, so that the convolutional neural network model takes global information into consideration and captures the long-distance dependencies between any two positions in the image, so that its recognition module can read the contextual information of the recognized graphics and easily understand the graphics content; the main purpose of the pooling layer is to reduce the spatial size of the feature map, thereby reducing the amount of calculation of the subsequent layers and increasing the invariance to the displacement of the graphics; the pooling operation average pooling calculates the average value of the area; the pooling layer downsamples the input feature map by calculating the average value of all elements in the covered area, reducing the size of the feature map but retaining important information; through continuous convolution and pooling operations, the convolutional neural network extracts high-level features of the graphics layer by layer; after the feature extraction is completed, the recognition module inputs the extracted feature vector into the fully connected layer for classification; the fully connected layer maps the feature vector to different category labels through linear transformation and nonlinear activation function, and finally outputs the recognition result of the graphics.

[0035] The output module performs a detailed analysis of the recognition results and determines the final recognized object by decoding the high-level features or classification labels output by the recognition module. The analyzed recognition results determine the output content and structure based on the specific needs of the user. For example, if the user needs to annotate specific objects in a graphic, the output module combines the recognition results with the coordinate information of the graphic to generate an annotated graphic.

[0036] The output module converts the parsed recognition results accordingly according to the output format option selected by the user; for graphic output, the output module uses a graphic processing library such as Matplotlib to generate a graphic file with recognition information; this is convenient for users to further analyze and display; for text output, the output module generates a text file containing the recognition results. The output module converts the recognition results into a text format that is easy to read and process, including the name of the recognized object, recognition confidence, and related attributes; the generation of the text file is convenient for users to view and record the recognition results, and is also convenient for data exchange with other systems or applications; for data file output, the output module uses a data processing library such as Pandas or Json to generate a data file in the format selected by the user. The output module organizes the recognition results into a structured data format, such as a table or JSON object, so that users can easily analyze, store or transmit data.

[0037] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A digitally interconnected complex pattern reading and recognition system, characterized in that: It includes a graphics input module, a feature extraction module, a graphics simplification module, a recognition module and an output module; wherein the graphics input module is used to receive complex graphics uploaded by users and perform standardization processing; the feature extraction module performs feature extraction processing on the standardized complex graphics, extracts key features that are conducive to identifying complex graphics, including shape, texture and color features; the graphics simplification module reorganizes the key features in the feature extraction module through multi-scale segmentation and adaptive noise reduction, and reduces redundant information to simplify the complex graphics; the recognition module reads and recognizes the reorganized and simplified graphics, and the recognition module uses a deep learning algorithm to introduce a self-attention mechanism into the convolutional neural network model to recognize graphics and capture the connection between graphic features; The output module outputs the recognition results of deep learning and supports multiple output formats, including graphics, text and data files.

2. A digitally interconnected complex pattern reading and recognition system according to claim 1, characterized in that: The image input module receives the graphic file through the user interaction interface; The graphic input module supports a variety of graphic file formats to meet the needs of different users and graphics from different sources; The graphic file is divided into multiple data slices, each of which is assigned a serial number. When the data slices start to be transmitted, the feature extraction module reorganizes them according to the serial numbers to obtain the graphic file; After the graphic file is successfully loaded, the graphic input module performs preprocessing, including standardization of the graphic, unifying the size and color of the graphic; The graphics are scaled and standardized; the color information of the graphics is converted into grayscale information through Y=0.299R+0.587G+0.114B, and the feature extraction module is used to accurately extract features through scaling and color conversion.

3. A digitally interconnected complex pattern reading and recognition system according to claim 2, characterized in that: The feature extraction module uses an edge detection algorithm to identify the edge information of the graphic after the graphic is scaled and the color is converted, thereby locating the outline of the graphic; extracts the closed contour through contour tracking technology, and calculates the geometric features of the contour, including area, perimeter, circularity, and rectangularity; The internal structure of the graphic is analyzed through structural analysis technology to identify the texture regularity of the graphic; including detecting repeated elements, directionality and roughness in the graphic, thereby generating texture feature parameters, including regularity, directionality and roughness.

4. A digitally interconnected complex pattern reading and recognition system according to claim 3, characterized in that: The edge detection algorithm uses Sobel to identify edge information and applies the Sobel operator to process the grayscale graphics. Calculate the gradient change of the local area to identify the edge; There are two Sobel operators, one is Gx for detecting horizontal edges, and the other is Gy for detecting vertical edges; Sobel operator dynamically adjusts Gx and Gy through deep learning; The gradient of the graph is calculated by the Sobel operator, and the size and direction of the gradient reflect the edge information of the graph; Apply the threshold operation to identify the area with a gradient greater than the threshold as an edge; The threshold setting steps are as follows: Step 1. Process the image using an initial threshold and evaluate the quality of the result. Step 2. Determine whether to increase or decrease the threshold based on the evaluation results; Step 3. Adjust the threshold according to the set step size; Step 4. Iteration: Repeat steps 1-3 until the image quality stops improving. Use contour detection to connect the detected edge points into a continuous contour. Contour detection uses the cv2.findContours function to extract closed contours from the edges. The contour returned by the cv2.findContours function is a numpy array, which contains the coordinates of each point on the contour. The contour points are used to form a closed contour. The geometric features of the extracted contour are calculated.

5. A digitally interconnected complex pattern reading and recognition system according to claim 1, characterized in that: The graphic simplification module performs multi-scale segmentation processing on the extracted key features. The multi-scale segmentation technology performs multi-scale segmentation on the graphic features through filters and segmentation algorithms of different scales; at each scale, edge detection segmentation is used to segment the graphic features into different areas by detecting edge information in the shape features; the graphic simplification module performs adaptive denoising processing on the segmented graphic features. The adaptive denoising technology dynamically adjusts the denoising parameters according to the texture features and color features of the graphic, removes noise information, and retains the features of the edge, texture, and color; A similarity threshold is set, and the graphic simplification module reorganizes the processed key features. The graphic simplification module deletes part of the information whose similarity exceeds the similarity threshold according to the feature similarity after segmentation and denoising, retains the main features of the graphic, and simplifies the graphic. The reorganization of features includes the application of feature fusion and feature selection algorithms. Feature fusion is used to fuse feature information at different scales to form a comprehensive feature representation. The feature selection algorithm is used to select important features and delete redundant features.

6. A digitally interconnected complex pattern reading and recognition system according to claim 1, characterized in that: The recognition module receives the simplified graphics from the graphics simplification module and preprocesses the input simplified graphics. After the preprocessing is completed, the simplified graphics are input into the deep learning model for feature extraction and recognition. The recognition module extracts features through the pre-trained model of the convolutional neural network, introduces a self-attention mechanism in the convolution layer, and uses multiple filters or convolution kernels to scan the input image. A set of feature maps are generated according to the scanned input graphics. The self-attention mechanism enables the feature map to capture local features and short-distance dependencies, and enables the model to consider global information and capture long-distance dependencies between any two positions in the graphics. The pooling layer uses average pooling to calculate the average value of the area. The pooling layer downsamples the input feature map by calculating the average value of all elements in the coverage area to reduce the size of the feature map, thereby reducing the amount of calculation. The pre-trained model can extract high-level features in the graphics, including edges, textures, and shapes. After feature extraction is completed, the recognition module inputs the extracted feature vector into the fully connected layer for classification; the fully connected layer maps the texture, edge, and color of the features to different category labels through linear transformation and nonlinear activation function, and finally outputs the recognition result of the graphic.

7. A digitally interconnected complex pattern reading and recognition system according to claim 6, characterized in that: The output module analyzes the recognition results transmitted by the recognition module, and determines the content and structure of the output graph according to the analysis results; The output module converts the parsed recognition results according to the output format option selected by the user.

Images