Automatic pattern and pattern filing system fused with artificial intelligence

Through the combination of cascading high-resolution convolution model and SEB-ChOA algorithm, the problem of low accuracy and efficiency of pattern pattern feature extraction is solved, and more efficient feature extraction and optimization is achieved, premature convergence is avoided, and pattern pattern is improved Overall performance of automated archiving systems.

CN119066219AInactive Publication Date: 2024-12-03SHANGHAI THEATRE ACADEMY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411179984.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has problems with low feature extraction accuracy and low efficiency in pattern classification and archiving, and optimization algorithms tend to converge too early, making it difficult to find the global optimal solution in high-dimensional search space.

Method used

The cascading high-resolution convolution model is used to combine the wave pipeline strategy for feature extraction, and the model optimization is used using the SEB-ChOA algorithm, and the helical function and hybrid strategy are introduced to allocate calculation tasks to multiple GPUs for parallel processing to avoid premature convergence. And improve global search capabilities.

Benefits of technology

It significantly improves the accuracy and efficiency of pattern pattern feature extraction, avoids premature convergence, enhances the exploration ability in high-dimensional space, and shortens the calculation time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066219A_ABST
    Figure CN119066219A_ABST
Patent Text Reader

Abstract

The invention relates to the field of artificial intelligence, in particular to an artificial intelligence-fused pattern and pattern automatic filing system, which comprises a data collection and preprocessing subsystem, a feature extraction subsystem, a classification and label subsystem, a database management subsystem and an automatic filing subsystem, according to the scheme, multi-scale features in pattern data are captured by constructing a cascaded high-resolution convolution model, forward propagation and back propagation processes are divided into a plurality of stages by implementing a wave type pipeline strategy, and the stages are distributed to different GPUs for parallel calculation, so that the model training and reasoning efficiency is greatly improved; an SEB-ChOA algorithm is used for optimization of a cascade high-resolution convolution model, a spiral function and a hybrid strategy are introduced, the global search capability of the algorithm is improved, the problem of premature convergence is avoided, and the exploration capability of the algorithm in a high-dimensional space is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an automatic archiving system for patterns integrating artificial intelligence. Background Art

[0002] With the development of artificial intelligence, the classification and archiving of patterns have become more efficient and accurate, but general feature extraction methods cannot capture the detailed features in patterns, especially when dealing with complex and high-resolution patterns, resulting in the loss of feature information. At the same time, traditional methods require more computing resources and time to process the same task, and the training and reasoning processes are inefficient. Existing optimization algorithms have problems of premature convergence and insufficient search space. They are prone to fall into local optimal solutions during the search process, resulting in premature convergence and inability to find the global optimal solution. They also perform poorly in high-dimensional search spaces and find it difficult to fully explore the entire search space. Summary of the invention

[0003] In view of the above situation, the present invention provides an automatic archiving system for patterns and patterns integrating artificial intelligence. For general feature extraction methods, there are problems of low quality precision, low accuracy and low efficiency. This solution captures multi-scale features in pattern and pattern data by constructing a cascaded high-resolution convolutional model, so that the model can more comprehensively understand the local and global features of the image, and by implementing a wave-like pipeline strategy, the forward propagation and back-propagation processes are divided into multiple stages, and are assigned to different GPUs for parallel computing, which greatly improves the efficiency of model training and reasoning, shortens the calculation time, and improves efficiency; in view of the problems of premature convergence and insufficient search space in existing optimization algorithms, this solution uses the SEB-ChOA algorithm (Spiral Evolutionary Behavior-Chimpanzee Optimization Algorithm) for the optimization of the cascaded high-resolution convolutional model, introduces spiral functions and hybrid strategies, adjusts parameters at each stage, and continuously updates individual positions, thereby improving the global search capability of the algorithm, avoiding the problem of premature convergence, improving the exploration capability of the algorithm in high-dimensional space, and increasing the coverage of the search space.

[0004] The present invention provides an automatic filing system for patterns and patterns integrated with artificial intelligence, including a data collection and preprocessing subsystem, a feature extraction subsystem, a classification and labeling subsystem, a database management subsystem and an automatic filing subsystem, which specifically includes the following contents:

[0005] The data collection and preprocessing subsystem collects pattern data and performs preprocessing;

[0006] The feature extraction subsystem uses a cascaded high-resolution convolutional model architecture to extract features from the preprocessed pattern data;

[0007] The classification and labeling subsystem classifies patterns and generates descriptive labels.

[0008] The database management subsystem establishes an image index;

[0009] The automatic filing subsystem defines the rules and strategies for automatic filing, and files according to categories, dates and tags.

[0010] Furthermore, in the feature extraction subsystem, a cascaded high-resolution convolutional model architecture is used to extract features from the preprocessed pattern data, specifically including the following steps:

[0011] Step S1: image collection and preprocessing, collecting pattern images for building a cascaded high-resolution convolutional model and performing denoising, normalization and cropping;

[0012] Step S2: construct a cascaded high-resolution convolutional model;

[0013] Step S3: Model training: a data set is established based on the pattern data used to construct the cascaded high-resolution convolutional model, and is divided into a training set, a validation set, and a test set in a ratio of 6:2:2. The cascaded high-resolution convolutional model is trained using the training set, and a wave pipeline strategy is implemented;

[0014] Step S4: Model optimization, using the SEB-ChOA algorithm to optimize the trained cascade high-resolution convolutional model;

[0015] Step S5: Fine edge detection, using a cascaded high-resolution convolutional model to detect edges in the pattern image, through multi-level feature extraction of the cascade structure.

[0016] Furthermore, step S2 specifically includes the following contents:

[0017] Multi-scale representation: Use convolution kernels of different sizes to perform convolution operations on the pattern image that constructs the cascaded high-resolution convolution model. Design multiple convolution layers, each layer uses convolution kernels of different sizes, and concatenate the outputs of these convolution layers in the channel dimension to generate a multi-scale feature map, which is transmitted to the classification and labeling subsystem;

[0018] High resolution preservation: Image features are extracted layer by layer through a multi-layer convolutional network, and a batch normalization layer is used after each convolutional layer;

[0019] Use convolutional layers for feature extraction, and use batch normalization layers and ReLU activation functions to process feature maps;

[0020] Cascade structure: Combine the output of each layer with the output of the previous layer, concatenate the output of the current layer with the output of the previous layer in the channel dimension, add the output of the current layer with the output of the previous layer element by element, and use the combined features as the input of the next layer;

[0021] Furthermore, in step S3, a wave pipeline strategy is implemented, which specifically includes the following steps:

[0022] Step S31: Forward propagation division, the forward propagation process of the cascaded high-resolution convolutional model is divided into five stages, specifically including the following:

[0023] Convolution layer processing stage: perform convolution operations, including convolution calculations and activation functions, extract local features in the pattern image, and generate local feature maps;

[0024] Pooling layer stage: Perform the maximum pooling operation to reduce the spatial dimension of the local feature map and send it to the feature map flattening stage;

[0025] Feature map flattening stage: flatten the local feature map output by the pooling layer into a one-dimensional vector and transmit it to the fully connected layer;

[0026] Fully connected layer stage: Set the fully connected layer weight matrix, perform matrix multiplication on the one-dimensional vector in the feature map flattening stage and the fully connected layer weight matrix, apply the activation function, and generate the feature vector of the fully connected layer;

[0027] Output layer stage: Set the output layer weight matrix, perform matrix multiplication on the feature vector of the fully connected layer and the weight matrix of the output layer, and apply the activation function to generate the output layer prediction value;

[0028] Step S32: back propagate partitioning, set the loss function, calculate the cross entropy loss of the output layer prediction value, and calculate the gradient of the loss function to the weights of each layer by the chain rule;

[0029] Step S33: Optimization part, updating the parameters of the cascaded high-resolution convolutional model according to the gradient of the loss function to the weights of each layer;

[0030] Step S34: stage division, the convolution layer processing stage is stage 1, the pooling layer stage and the feature map flattening stage are stage 2, the full connection layer stage and the output layer stage are stage 3, the back propagation is divided into stage 4, and the optimization part is stage 5;

[0031] Step S35: wave-like layout, with forward propagation and reverse propagation stages performed alternately;

[0032] Step S36: GPU task allocation, allocating the five stages to five different GPUs, GPU1 corresponds to stage 1, GPU2 corresponds to stage 2, GPU3 corresponds to stage 3, GPU4 corresponds to stage 4, and GPU5 corresponds to stage 5;

[0033] Step S37: synchronous communication, using synchronous communication operations to ensure synchronization between GPUs;

[0034] Step S38: Error handling and recovery, setting up error detection mechanisms in all five stages and designing error recovery strategies;

[0035] Further, step S4 specifically includes the following steps:

[0036] Step S41: Initialization: randomly initialize a group of individuals as a population, each individual represents a parameter combination of a cascaded high-resolution convolutional model, and set the individual position, population size and maximum number of iterations;

[0037] Step S42: fitness evaluation, defining a fitness function;

[0038] Step S43: Optimization process, introducing spiral function and hybrid strategy;

[0039] Step S44: Optimization terminates. When the maximum number of iterations is reached, the optimization process is terminated.

[0040] Step S45: output the optimal solution;

[0041] Further, step S43 specifically includes the following steps:

[0042] Step S431: In the driving phase, the individual position is updated using the spiral function. The formula used is as follows: ;

[0043] in, is the position of the individual at the next moment, is the individual's current location, is the amplitude of the spiral function, and the range is set to [5, 10], is the contraction speed of the spiral function, and the range is set to [0.5, 1.0], is the current iteration number;

[0044] Step S432: In the blocking stage, the spiral function is used to make fine adjustments to the local range, and the spiral function is further set. The range is [2,5], set The range of is [0.2,0.5];

[0045] Step S433: Chasing stage, approaching the optimal parameter combination, setting The range is [1,2], set The range of is [0.1,0.2];

[0046] Step S444: Attack phase, by adjusting , Range, set The range is [0.1,1], set The range of is [0.01,0.1], and the current optimal parameter combination is obtained.

[0047] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0048] (1) In view of the above situation, the present invention provides an automatic archiving system for patterns and patterns that integrates artificial intelligence. The general feature extraction methods have the problems of low quality precision, low accuracy and low efficiency. This solution captures the multi-scale features in the pattern data by constructing a cascaded high-resolution convolutional model, so that the model can more comprehensively understand the local and global features of the image. In addition, by implementing a wave-like pipeline strategy, the forward propagation and back-propagation processes are divided into multiple stages and assigned to different GPUs for parallel computing, which greatly improves the efficiency of model training and reasoning, shortens the computing time and improves efficiency.

[0049] (2) The present invention uses the SEB-ChOA algorithm (Spiral Evolutionary Behavior-Chimpanzee Optimization Algorithm) to optimize the cascaded high-resolution convolutional model, introduces spiral functions and hybrid strategies, adjusts parameters at each stage, and continuously updates individual positions, thereby improving the global search capability of the algorithm, avoiding the problem of premature convergence, improving the algorithm's exploration capability in high-dimensional space, and increasing the coverage of the search space. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A schematic diagram of an automatic archiving system for patterns and designs integrated with artificial intelligence provided by the present invention.

[0051] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all 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.

[0053] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying 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 direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0054] Example 1, see Figure 1 The present invention provides an automatic filing system for patterns and patterns integrated with artificial intelligence, including a data collection and preprocessing subsystem, a feature extraction subsystem, a classification and labeling subsystem, a database management subsystem and an automatic filing subsystem, which specifically includes the following contents:

[0055] The data collection and preprocessing subsystem collects pattern data and performs preprocessing;

[0056] The feature extraction subsystem uses a cascaded high-resolution convolutional model architecture to extract features from the preprocessed pattern data;

[0057] The classification and labeling subsystem classifies patterns and generates descriptive labels.

[0058] The database management subsystem establishes an image index;

[0059] The automatic filing subsystem defines the rules and strategies for automatic filing, and files according to categories, dates and tags.

[0060] Embodiment 2: Based on the above embodiment, in the feature extraction subsystem, a cascaded high-resolution convolution model architecture is used to extract features from the preprocessed pattern data, specifically including the following steps:

[0061] Step S1: image collection and preprocessing, collecting pattern images for building a cascaded high-resolution convolutional model and performing denoising, normalization and cropping;

[0062] Step S2: construct a cascaded high-resolution convolutional model;

[0063] Step S3: Model training: a data set is established based on the pattern data used to construct the cascaded high-resolution convolutional model, and is divided into a training set, a validation set, and a test set in a ratio of 6:2:2. The cascaded high-resolution convolutional model is trained using the training set, and a wave pipeline strategy is implemented;

[0064] Step S4: Model optimization, using the SEB-ChOA algorithm to optimize the trained cascade high-resolution convolutional model;

[0065] Step S5: Fine edge detection, using a cascaded high-resolution convolutional model to detect edges in the pattern image, through multi-level feature extraction of the cascade structure.

[0066] Embodiment 3, based on the above embodiment, step S2 specifically includes the following contents:

[0067] Multi-scale representation: Use convolution kernels of different sizes to perform convolution operations on the pattern image that constructs the cascaded high-resolution convolution model. Design multiple convolution layers, each layer uses convolution kernels of different sizes, and concatenate the outputs of these convolution layers in the channel dimension to generate a multi-scale feature map, which is transmitted to the classification and labeling subsystem;

[0068] High resolution preservation: Image features are extracted layer by layer through a multi-layer convolutional network, and a batch normalization layer is used after each convolutional layer;

[0069] Use convolutional layers for feature extraction, and use batch normalization layers and ReLU activation functions to process feature maps;

[0070] Cascade structure: The output of each layer is combined with the output of the previous layer, the output of the current layer is concatenated with the output of the previous layer in the channel dimension, the output of the current layer is added element by element with the output of the previous layer, and the combined features are used as the input of the next layer.

[0071] Embodiment 4: This embodiment is based on the above embodiment. In step S3, a wave pipeline strategy is implemented, which specifically includes the following steps:

[0072] Step S31: Forward propagation division, the forward propagation process of the cascaded high-resolution convolutional model is divided into five stages, specifically including the following:

[0073] Convolutional layer processing stage: design a convolution layer with a convolution kernel size of 3x3, a step size of 1, and a padding method of SAME. Perform convolution calculation on the preprocessed pattern image to extract local features. Apply the ReLU activation function to perform nonlinear transformation on the convolution feature map. The feature map obtained after the convolution operation represents the local features of the pattern and generates a local feature map.

[0074] Pooling layer stage: Design a pooling layer with a pooling window size of 2x2 and a step size of 2, perform a maximum pooling operation, reduce the spatial dimension of the local feature map and send it to the feature map flattening stage;

[0075] Feature map flattening stage: flatten the local feature map output by the pooling layer into a one-dimensional vector and transmit it to the fully connected layer;

[0076] Fully connected layer stage: Set the fully connected layer weight matrix, perform matrix multiplication on the one-dimensional vector in the feature map flattening stage and the fully connected layer weight matrix, apply the activation function, and generate the feature vector of the fully connected layer;

[0077] Output layer stage: Set the output layer weight matrix, perform matrix multiplication on the feature vector of the fully connected layer and the weight matrix of the output layer, and apply the Softmax activation function to generate the output layer prediction value;

[0078] Step S32: back propagate partitioning, set the loss function, calculate the cross entropy loss of the output layer prediction value, and calculate the gradient of the loss function to the weights of each layer by the chain rule;

[0079] Step S33: Optimization part, updating the parameters of the cascaded high-resolution convolutional model according to the gradient of the loss function to the weights of each layer;

[0080] Step S34: stage division, the convolution layer processing stage is stage 1, the pooling layer stage and the feature map flattening stage are stage 2, the full connection layer stage and the output layer stage are stage 3, the back propagation is divided into stage 4, and the optimization part is stage 5;

[0081] Step S35: wave-like layout, with forward propagation and reverse propagation stages performed alternately;

[0082] Step S36: GPU task allocation, allocating the five stages to five different GPUs, GPU1 corresponds to stage 1, GPU2 corresponds to stage 2, GPU3 corresponds to stage 3, GPU4 corresponds to stage 4, and GPU5 corresponds to stage 5;

[0083] Step S37: synchronous communication, using synchronous communication operations to ensure synchronization between GPUs;

[0084] Step S38: Error handling and recovery: error detection mechanisms are set up in all five stages, and error recovery strategies are designed.

[0085] Embodiment 5, based on the above embodiment, step S4 specifically includes the following steps:

[0086] Step S41: Initialization: randomly initialize a group of individuals as a population, each individual represents a parameter combination of a cascaded high-resolution convolutional model, set the individual position, the population size is 50, and the maximum number of iterations is 1000;

[0087] Step S42: fitness evaluation, defining a fitness function, and calculating the accuracy of the cascaded high-resolution convolutional model corresponding to each individual on the validation set. The higher the accuracy, the higher the fitness.

[0088] Step S43: Optimization process, introducing spiral function and hybrid strategy;

[0089] Step S44: Optimization terminates. When the maximum number of iterations is reached, the optimization process is terminated.

[0090] Step S45: Output the optimal solution.

[0091] Embodiment 6, based on the above embodiment, step S43 specifically includes the following steps:

[0092] Step S431: In the driving phase, the individual position is updated using the spiral function. The formula used is as follows: ;

[0093] in, is the position of the individual at the next moment, is the individual's current location, is the amplitude of the spiral function, set to 7, is the contraction speed of the spiral function, set to 0.8, is the current iteration number;

[0094] Step S432: In the blocking stage, the spiral function is used to make fine adjustments to the local range, and the spiral function is further set. To 3, set is 0.3;

[0095] Step S433: Chasing stage, approaching the optimal parameter combination, setting is 1.5, set is 0.15;

[0096] Step S444: Attack phase, by adjusting , Range, set is 0.5, set is 0.05, and the current optimal parameter combination is obtained.

[0097] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0098] 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.

[0099] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An automatic archiving system for patterns and patterns integrating artificial intelligence, Features: It includes data collection and preprocessing subsystem, feature extraction subsystem, classification and labeling subsystem, database management subsystem and automatic archiving subsystem, specifically including the following contents: The data collection and preprocessing subsystem collects pattern data and performs preprocessing; The feature extraction subsystem uses a cascaded high-resolution convolutional model architecture to extract features from the preprocessed pattern data; The classification and labeling subsystem classifies patterns and generates descriptive labels. The database management subsystem establishes an image index; The automatic filing subsystem defines the rules and strategies for automatic filing, and files according to categories, dates and tags; In the feature extraction subsystem, a cascaded high-resolution convolutional model architecture is used to extract features from the preprocessed pattern data, specifically including the following steps: Step S1: image collection and preprocessing, collecting pattern images for building a cascaded high-resolution convolutional model and performing denoising, normalization and cropping; Step S2: construct a cascaded high-resolution convolutional model; Step S3: Model training: a data set is established based on the pattern data used to construct the cascaded high-resolution convolutional model, and is divided into a training set, a validation set, and a test set in a ratio of 6:2:

2. The cascaded high-resolution convolutional model is trained using the training set, and a wave pipeline strategy is implemented; Step S4: Model optimization, using the SEB-ChOA algorithm to optimize the trained cascade high-resolution convolutional model; Step S5: Fine edge detection, using a cascaded high-resolution convolutional model to detect edges in the pattern image, through multi-level feature extraction of the cascade structure.

2. According to claim 1, an automatic archiving system for patterns integrating artificial intelligence, Features: Step S2 specifically includes the following contents: Multi-scale representation: Use convolution kernels of different sizes to perform convolution operations on the pattern image that constructs the cascaded high-resolution convolution model. Design multiple convolution layers, each layer uses convolution kernels of different sizes, and concatenate the outputs of these convolution layers in the channel dimension to generate a multi-scale feature map, which is transmitted to the classification and labeling subsystem; High resolution preservation: Image features are extracted layer by layer through a multi-layer convolutional network, and a batch normalization layer is used after each convolutional layer; Use convolutional layers for feature extraction, and use batch normalization layers and ReLU activation functions to process feature maps; Cascade structure: The output of each layer is combined with the output of the previous layer, the output of the current layer is concatenated with the output of the previous layer in the channel dimension, the output of the current layer is added element by element with the output of the previous layer, and the combined features are used as the input of the next layer.

3. An automatic archiving system for patterns and patterns incorporating artificial intelligence according to claim 2, Features: In step S3, a wave pipeline strategy is implemented, which specifically includes the following steps: Step S31: Forward propagation division, the forward propagation process of the cascaded high-resolution convolutional model is divided into five stages, specifically including the following: Convolution layer processing stage: perform convolution operations, including convolution calculations and activation functions, extract local features in the pattern image, and generate local feature maps; Pooling layer stage: Perform the maximum pooling operation to reduce the spatial dimension of the local feature map and send it to the feature map flattening stage; Feature map flattening stage: flatten the local feature map output by the pooling layer into a one-dimensional vector and transmit it to the fully connected layer; Fully connected layer stage: Set the fully connected layer weight matrix, perform matrix multiplication on the one-dimensional vector in the feature map flattening stage and the fully connected layer weight matrix, apply the activation function, and generate the feature vector of the fully connected layer; Output layer stage: Set the output layer weight matrix, perform matrix multiplication on the feature vector of the fully connected layer and the weight matrix of the output layer, and apply the activation function to generate the output layer prediction value; Step S32: back propagate partitioning, set the loss function, calculate the cross entropy loss of the output layer prediction value, and calculate the gradient of the loss function to the weights of each layer by the chain rule; Step S33: Optimization part, updating the parameters of the cascaded high-resolution convolutional model according to the gradient of the loss function to the weights of each layer; Step S34: stage division, the convolution layer processing stage is stage 1, the pooling layer stage and the feature map flattening stage are stage 2, the full connection layer stage and the output layer stage are stage 3, the back propagation is divided into stage 4, and the optimization part is stage 5; Step S35: wave-like layout, with forward propagation and reverse propagation stages performed alternately; Step S36: GPU task allocation, allocating the five stages to five different GPUs, GPU1 corresponds to stage 1, GPU2 corresponds to stage 2, GPU3 corresponds to stage 3, GPU4 corresponds to stage 4, and GPU5 corresponds to stage 5; Step S37: synchronous communication, using synchronous communication operations to ensure synchronization between GPUs; Step S38: Error handling and recovery: error detection mechanisms are set up in all five stages, and error recovery strategies are designed.

4. According to claim 3, an automatic archiving system for patterns integrating artificial intelligence, Features: Step S4 specifically includes the following steps: Step S41: Initialization: randomly initialize a group of individuals as a population, each individual represents a parameter combination of a cascaded high-resolution convolutional model, and set the individual position, population size and maximum number of iterations; Step S42: fitness evaluation, defining a fitness function; Step S43: Optimization process, introducing spiral function and hybrid strategy; Step S44: Optimization terminates. When the maximum number of iterations is reached, the optimization process is terminated. Step S45: Output the optimal solution.

5. According to claim 4, an automatic archiving system for patterns integrating artificial intelligence, Features: Step S43 specifically includes the following steps: Step S431: In the driving phase, the individual position is updated using the spiral function. The formula used is as follows: ; in, is the position of the individual at the next moment, is the individual's current location, is the amplitude of the spiral function, the range is set to [5,10], is the contraction speed of the spiral function, and the range is set to [0.5, 1.0], is the current iteration number; Step S432: In the blocking stage, the spiral function is used to make fine adjustments to the local range, and the spiral function is further set. The range is [2,5], set The range of is [0.2,0.5]; Step S433: Chasing stage, approaching the optimal parameter combination, setting The range is [1,2], set The range of is [0.1,0.2]; Step S444: Attack phase, by adjusting , Range, set The range is [0.1,1], set The range of is [0.01,0.1], and the current optimal parameter combination is obtained.