Automatic processing device for visual information of historical map

Abstract

The invention belongs to the technical field of digital humanity, computer vision and cartography, and discloses an automatic processing device for historical map visual information, which comprises a map image preprocessing module, a map fragmentation module, a feature extraction and optimization module, a map type generation module and a visual evolution analysis module. According to the method, the map image is decomposed into basic units (mapels), explainable feature representation is constructed based on comparative learning, and automatic visual style analysis of a large-scale historical map set is achieved. High efficiency: a large-scale historical map set can be automatically processed, and the analysis efficiency is remarkably improved. And interpretability: the feature space generated through comparative learning has clear semantic interpretation, so that the evolution of map styles can be understood conveniently. Interdisciplinary application: the method is suitable for multiple fields such as history, geography and cultural research, and a new method is provided for quantitative research of cartography.

Description

Technical Field

[0001] The present invention belongs to the technical fields of digital humanities, computer vision, and cartography, and particularly relates to an automated processing device for visual information of historical maps. Background Art

[0002] As a product of culture and technology, historical maps reflect geographical, social, and cultural changes in different periods. With the progress of digital technology, a large number of historical maps have been digitized and made public, providing unprecedented opportunities for the automated processing of maps. However, existing research methods mostly rely on manual analysis, making it difficult to process large-scale atlases and lacking a systematic method for quantifying visual styles.

[0003] Through the above analysis, the problems and defects of the existing technology are as follows:

[0004] Existing research methods mostly rely on manual analysis, making it difficult to process large-scale atlases and lacking a systematic method for quantifying visual styles. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention provides an automated processing device for visual information of historical maps.

[0006] The present invention is implemented as follows. An automated processing device for visual information of historical maps includes:

[0007] A map image preprocessing module, a map segmentation module, a feature extraction and optimization module, a map type generation module, and a visual evolution analysis module;

[0008] The map image preprocessing module, connected to the map segmentation module, is used to digitize historical maps and remove background noise; use the OCRNet neural network to perform semantic segmentation on the maps and extract geographical content areas;

[0009] The map segmentation module, connected to the feature extraction and optimization module, is used to divide the map image into basic units of 50×50 pixels, and each unit represents a local visual feature of the map;

[0010] The feature extraction and optimization module, connected to the map type generation module, is used to extract 13 candidate visual features, optimize the feature space using a genetic algorithm, select the feature combination that can best distinguish different map series, and construct a 6-dimensional feature space;

[0011] The map type generation module, connected to the visual evolution analysis module, is used to generate map types mapotype by iteratively dividing the feature space, and each mapotype represents a class of mapels with similar visual features;

[0012] The visual evolution analysis module, connected to the map type generation module, is used to analyze the visual style evolution of maps in different periods and at different scales based on the distribution of mapotype.

[0013] Furthermore, the map image preprocessing module:

[0014] Collect more than 10,000 historical maps from multiple libraries and archives.

[0015] Furthermore, the map fragmentation module:

[0016] Through random sampling and repositioning, ensure that each mapel covers different areas of the map and exclude areas with low information content.

[0017] Furthermore, the candidate visual features include color distribution, morphology, texture, graphic load, line width, and direction.

[0018] Furthermore, the map type generation module:

[0019] Use the t-SNE algorithm to reduce the high-dimensional feature space to 2D for easy visual analysis.

[0020] Furthermore, the visual evolution analysis module:

[0021] Quantify the long-term change trend of the map style through statistical methods.

[0022] Another object of the present invention is to provide an automated processing method for historical map information, including:

[0023] Step 1, digitally process the historical map through the map image preprocessing module to remove background noise; use the OCRNet neural network to perform semantic segmentation on the map and extract the geographical content area;

[0024] Step 2, segment the map image into basic units of 50×50 pixels through the map fragmentation module, and each unit represents a local visual feature of the map;

[0025] Step 3, extract 13 candidate visual features through the feature extraction and optimization module, use the genetic algorithm to optimize the feature space, select the feature combination that can best distinguish different map series, and construct a 6D feature space;

[0026] Step 4, use the map type generation module to iteratively divide the feature space to generate the map type mapotype, and each mapotype represents a class of mapels with similar visual features;

[0027] Step 5, analyze the visual style evolution of maps in different periods and at different scales through the visual evolution analysis module based on the distribution of mapotype.

[0028] Another object of the present invention is to provide a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the method for automatically processing historical map information.

[0029] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor is caused to execute the steps of the method for automatically processing historical map information.

[0030] Another object of the present invention is to provide an information data processing terminal for implementing the automatic processing device for historical map visual information.

[0031] Combined with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by the present invention are as follows:

[0032] For a long time, the research on historical maps has mainly relied on manual methods for sorting, comparison, and style analysis, resulting in low data processing efficiency and difficulty in supporting systematic research on large-scale historical map collections. The automatic map processing device proposed by the present invention constructs interpretable feature representations based on the contrastive learning method, enabling the system to analyze historical maps batch by batch and intelligently, and solving the problems of inefficiency, subjectivity, and difficulty in standardization faced by traditional research methods in large-scale data processing. This breakthrough allows scholars to analyze tens of thousands of historical maps in a short period of time, thus promoting the quantitative research of cartography.

[0033] The style features of historical maps are complex, and traditional feature extraction methods based on artificial rules or machine learning often have difficulty accurately expressing the core features of map style evolution. The present invention constructs an interpretable feature space based on the contrastive learning method on the basis of unsupervised learning, enabling each map to find the historical map type (mapotype) with the closest style in the high-dimensional feature space. This method ensures that when the system extracts map features, it can automatically identify core visual elements such as map boundaries, line styles, fonts, and color patterns, and provide clear semantic explanations based on data-driven, filling the gap in the research on automatic classification of historical map styles.

[0034] In the past, the analysis of maps in disciplines such as history, geography, and cultural studies mostly remained at the level of qualitative description, lacking effective quantitative research tools. Through technical means such as automated feature extraction, machine learning classification, and visual style evolution analysis, the map data of the present invention can be used for interdisciplinary research such as historical event analysis, geographical environment changes, urban development trajectories, and cultural dissemination paths. This innovation enables cartographic research to no longer be limited to traditional map cartography, but to expand to the quantification, pattern recognition, and large-scale statistical analysis of historical maps, filling an important technical gap in interdisciplinary research.

[0035] Historical maps often have a wide range of sources, with different formats, scales, and drawing styles. It is difficult to construct a standardized analysis framework in traditional research. Through the map unitization method based on mapels, the present invention decomposes historical maps of different sources and styles into unified basic units of 50×50 pixels, and conducts comparative learning and clustering analysis in the feature space. This method can not only be applied to maps of different eras and drawing methods, but also establish a quantitative standard for similarity between different maps through feature mapping, so as to uniformly analyze various types of historical map data such as hand-drawn maps, printed maps, and digital maps.

[0036] Traditional map evolution analysis mainly relies on manual comparison, which is not only time-consuming and laborious, but also difficult to extract deep-seated laws. The visual evolution analysis module proposed by the present invention, based on mapotype clustering and time series analysis, can automatically detect the visual style changes of maps in different periods. For example, it can analyze the change trend of map styles in a certain country or region from the 16th century to the 20th century, and identify font evolution, changes in cartographic symbols, and changes in the complexity of boundary lines. This innovative method enables cartographers and historians to more intuitively and systematically understand the evolution of map styles over time, providing data-driven quantitative support for historical research.

[0037] The processing of large-scale historical map data often involves high-dimensional feature analysis. Traditional machine learning methods have limitations in terms of computational efficiency and classification accuracy. The present invention uses a genetic algorithm to optimize the feature space, reducing the dimensionality from 13-dimensional candidate features to a 6-dimensional optimal feature space. This not only significantly reduces the computational complexity, but also improves the accuracy of map classification. This optimization strategy enables the system to improve the operation efficiency while ensuring high accuracy, and can complete the automated classification and analysis of a large number of historical maps in a short time, breaking through the computational bottleneck of traditional methods in large-scale data processing.

[0038] The automated processing device for historical map visual information of the present invention has successfully solved six core technical problems that have long remained unsolved in the research of historical maps through breakthroughs in multiple technical links such as map decomposition, feature learning, classification optimization, and evolution analysis. It provides an efficient, interpretable, standardizable, and quantifiable analysis tool for multiple fields such as cartography, history, and geography, greatly promoting the quantitative research and automated processing capabilities of historical maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a structural block diagram of the automated processing device for historical map visual information provided by an embodiment of the present invention.

[0040] Figure 2 is a flowchart of the method of the map segmentation module provided by an embodiment of the present invention.

[0041] Figure 3 is a flowchart of the method for automated processing of historical map information provided by an embodiment of the present invention.

[0042] Figure 1 In the figure: 1. Map image preprocessing module; 2. Map segmentation module; 3. Feature extraction and optimization module; 4. Map type generation module; 5. Visual evolution analysis module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] As Figure 1 shown, an automated processing device for historical map visual information provided by an embodiment of the present invention includes:

[0045] Map image preprocessing module 1, map segmentation module 2, feature extraction and optimization module 3, map type generation module 4, visual evolution analysis module 5;

[0046] The map image preprocessing module 1 is connected to the map segmentation module 2 and is used for digitizing the historical map to remove background noise; using the OCRNet neural network to perform semantic segmentation on the map to extract the geographical content area;

[0047] The map segmentation module 2 is connected to the feature extraction and optimization module 3 and is used for segmenting the map image into basic units of 50×50 pixels, and each unit represents a local visual feature of the map;

[0048] The feature extraction and optimization module 3, connected to the map type generation module 4, is used to extract 13 candidate visual features, optimize the feature space using a genetic algorithm, select the feature combination that can best distinguish different map series, and construct a 6-dimensional feature space;

[0049] The map type generation module 4, connected to the visual evolution analysis module 5, is used to generate map types mapotype by iteratively partitioning the feature space, where each mapotype represents a class of mapels with similar visual features;

[0050] The visual evolution analysis module 5, connected to the map type generation module 4, is used to analyze the visual style evolution of maps at different times and scales based on the distribution of mapotype.

[0051] The map image preprocessing module provided by the embodiment of the present invention: collects more than 10,000 historical maps from multiple libraries and archives.

[0052] The map image preprocessing module of the present invention first collects more than 10,000 historical maps from multiple libraries and archives and converts them into high-resolution digital images. Since historical maps are usually affected by paper aging, ink blurring, scanning noise, and color fading, etc., this module uses image enhancement and denoising algorithms (such as adaptive histogram equalization, filter denoising, color normalization) to improve the readability of the maps. Subsequently, the system uses OCRNet (Optical Character Recognition Network) for semantic segmentation to automatically extract the geographical content areas in the maps and remove irrelevant background information, such as watermarks, binding lines, annotations in non-map areas, etc., thus laying a foundation for subsequent map structured analysis.

[0053] On the preprocessed map image, the map fragmentation module divides the entire map into basic units of 50×50 pixels, so that each unit contains only one local visual feature. This method ensures that the local information of the map can be analyzed independently, and at the same time helps to maintain spatial consistency and comparability. Subsequently, the feature extraction and optimization module calculates the visual features of these fragments and extracts 13 candidate visual features (such as line density, texture pattern, color distribution, edge direction histogram, etc.). However, not all features are useful for map classification, so this module uses a genetic algorithm (Genetic Algorithm, GA) to optimize the feature space, automatically selects the feature combination that can best distinguish different map types, reduces the dimension from 13 dimensions to 6-dimensional optimal feature space, thereby reducing redundant data and improving the classification accuracy.

[0054] Through the optimized six-dimensional feature space, the map type generation module uses an iterative partitioning method to classify maps into different mapotypes. A mapotype represents a set of maps with similar visual features that belong to the same historical period, similar drawing styles, or share the same cartographic techniques. For example, hand-drawn maps from the 18th century form one mapotype, while printed maps from the early 19th century form another mapotype. The key task of this module is to automatically divide maps into different types using hierarchical clustering or K-Means clustering and create a hierarchical structure of map categories for subsequent analysis and retrieval.

[0055] Finally, based on the generated mapotypes, the visual evolution analysis module explores the evolution trends of historical maps in the time dimension and scale dimension. This module uses time series analysis to detect how the cartographic styles of maps change over time. For example:

[0056] Evolution of map boundaries and symbols: Analyze the changes in map boundary lines to detect the transition from hand-drawn maps to printed maps.

[0057] Changes in annotation fonts and symbols: Identify the annotation styles of different maps in historical periods and distinguish between handwritten, engraved, or modern printed styles.

[0058] Map color patterns: Observe the evolution trends of map colors, such as the transition from monochromatic drawing to color printing.

[0059] This analysis can help scholars and researchers understand the development of cartographic techniques and provide data support for map digitization, historical geography research, and analysis of the evolution of cartography.

[0060] In summary, the automated processing device for historical map visual information of the present invention realizes the automated classification, pattern recognition, and evolution trend analysis of historical maps through intelligent image processing, optimized feature extraction, machine learning classification, and time series analysis, improving the usability and research value of historical map data.

[0061] As Figure 2 shown, the map fragmentation module provided by the embodiment of the present invention:

[0062] S101, through random sampling and repositioning, ensure that each mapel covers different areas of the map and exclude areas with low information content.

[0063] The candidate visual features provided by the embodiment of the present invention include color distribution, morphology, texture, graphic load, line width, and direction.

[0064] The map type generation module provided by the embodiment of the present invention:

[0065] Use the t-SNE algorithm to reduce the high-dimensional feature space to 2D for visual analysis convenience.

[0066] The visual evolution analysis module provided by the embodiments of the present invention:

[0067] Quantify the long-term change trend of map styles through statistical methods.

[0068] As Figure 3 shown, a method for automatically processing historical map information provided by the embodiments of the present invention includes:

[0069] S201, digitally process the historical map through the map image preprocessing module to remove background noise; use the OCRNet neural network to perform semantic segmentation on the map and extract the geographical content area;

[0070] S202, segment the map image into basic units of 50×50 pixels through the map fragmentation module, and each unit represents a local visual feature of the map;

[0071] S203, extract 13 candidate visual features through the feature extraction and optimization module, optimize the feature space using the genetic algorithm, select the feature combination that can best distinguish different map series, and construct a 6D feature space;

[0072] S204, use the map type generation module to iteratively divide the feature space to generate the map type mapotype, and each mapotype represents a class of mapels with similar visual features;

[0073] S205, analyze the visual style evolution of maps in different periods and at different scales through the visual evolution analysis module based on the distribution of mapotype.

[0074] The present invention provides a method for automatically processing historical map information, aiming to efficiently analyze historical map information through digital and intelligent analysis technologies. First, digitally process the historical map through the map image preprocessing module, use a denoising algorithm to remove background noise, and use the OCRNet neural network for semantic segmentation to accurately extract the geographical content area from the map image to ensure the accuracy of subsequent analysis. On this basis, use the map fragmentation module to divide the map image into basic units of 50×50 pixels, and each unit corresponds to a local visual feature of the map to facilitate subsequent feature extraction and pattern recognition.

[0075] In terms of the analysis and optimization of the visual features of maps, the present invention designs a feature extraction and optimization module to extract 13 candidate visual features from map units, and uses a genetic algorithm to optimize the feature space, screen the feature combinations that can best distinguish different map series, and finally construct a 6-dimensional feature space. Based on this feature space, the map type generation module automatically generates map types (mapotype) by iteratively dividing the feature space. Each mapotype represents a class of mapels with similar visual features, thus realizing the intelligent classification of historical maps.

[0076] To further study the evolution trend of historical maps, the present invention introduces a visual evolution analysis module to analyze the visual style evolution of maps at different times and scales based on the distribution of mapotype. This analysis method can reveal the changes in cartographic styles, symbol expressions, color uses, etc. of maps in different eras, and thus assist in historical geography research, map restoration, and historical information mining. This method can be widely applied in the fields of historical geography research, archaeological cartography, museum digital management, etc.

[0077] The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the program is executed by the processor, it realizes each step of the automated processing of historical map information. In addition, the present invention also covers a computer-readable storage medium for storing the computer program, enabling it to run on an information data processing terminal, realizing the intelligent and automated parsing of historical map information, and improving the efficiency and accuracy of researchers in processing historical maps.

[0078] The present invention realizes the digitization of map information, feature extraction, type generation, and visual evolution analysis by automatically processing historical maps. In the preprocessing stage of map images, the historical maps are first digitized, and an image denoising algorithm is used to remove background noise (such as borders, titles, legends, etc.) to ensure the clarity of the main map information. Subsequently, the OCRNet neural network is used to perform semantic segmentation to accurately extract the geographical content areas in the maps and eliminate irrelevant elements, providing high-quality input data for subsequent analysis.

[0079] In the map fragmentation stage, the map is divided into basic units (mapels) of 50×50 pixels, and each mapel represents the local visual features of the map. To ensure representativeness, the system ensures that the mapels evenly cover the entire map through random sampling and repositioning, while excluding low-information areas (such as blank areas). This fragmentation strategy can ensure the full expression of the local features of the map and provide a basis for subsequent feature extraction and pattern analysis.

[0080] In the feature extraction and optimization stage of the present invention, 13 candidate visual features are designed, covering multiple dimensions such as color distribution, morphology, texture, graphic load, line width, and direction, to comprehensively characterize the visual style of maps. To screen the most discriminative feature combinations, a genetic algorithm is used to optimize the feature space, and a 6-dimensional optimal feature space is selected through iterative calculation. This optimization strategy not only reduces redundant features but also improves the discrimination ability for different map series, making automatic classification more accurate and reliable.

[0081] In the map type (mapotype) generation stage, the system automatically generates map types (mapotypes) based on the optimized feature space using an iterative partitioning strategy. Each mapotype represents a group of mapels with similar visual features. At the same time, the t-SNE dimensionality reduction algorithm is used to map high-dimensional features to a two-dimensional visualization space, making the relationships between map types more intuitive and facilitating the analysis and interpretation of the differences and evolution patterns between map types.

[0082] The visual evolution analysis module of the present invention deeply studies the changes in visual styles at different historical periods and different map scales based on the distribution of mapotypes. The system uses statistical methods (such as the Shapiro-Wilk normality test and Kendall correlation coefficient) to quantitatively analyze the long-term change trends of map styles and reveal the evolution paths of historical maps in aspects such as cartographic style, color use, and symbol expression. Through these quantitative analyses, researchers can explore the development of cartographic techniques in different eras and further understand the change patterns of map visual information.

[0083] The present invention realizes an efficient, interpretable, and interdisciplinary automated processing solution for historical maps. First, this method can automatically process large-scale historical map collections, greatly improving data processing efficiency and being applicable to application scenarios such as archives, historical research, and geographical analysis. Second, through the optimized feature space and classification model, the generated mapotypes have clear semantic interpretations, which helps to understand the evolution process of map styles. In addition, this method can be widely applied to multiple disciplinary fields such as cultural research, cartography, archaeology, and museum digital management, providing a new technical tool for quantitative cartographic research.

[0084] Historical maps usually contain complex visual elements such as boundary lines, place name annotations, contour lines, rivers, and building symbols. At the same time, due to problems such as aging, fading, and blurred ink marks in historical maps, traditional manual recognition methods are difficult to efficiently extract the core information of maps. Therefore, the present invention uses OCRNet (Optical Character Recognition Network) to perform automated semantic segmentation on maps to accurately distinguish map areas from non-map areas and extract the main geographical elements.

[0085] 1) Data input and preprocessing

[0086] Collect historical map images and perform denoising, contrast enhancement, and color normalization on the images to improve readability.

[0087] Use edge detection algorithms (such as the Canny algorithm) and histogram equalization to enhance the main geographical structures in the map.

[0088] 2) OCRNet semantic segmentation

[0089] Adopt OCRNet with a deep learning convolutional neural network (CNN) structure, input the processed map image, and perform semantic segmentation.

[0090] Set multiple classification labels, including geographical regions (such as rivers, mountains, roads), text annotations (such as place names, scales), and background noise.

[0091] The network calculates features through multiple layers of convolution and combines global context information to optimize the map region segmentation results.

[0092] 3) Extract the main content of the map

[0093] The semantic segmentation results are used to screen the geographical content areas and remove background noise (such as binding lines, watermarks, stains, etc.).

[0094] Adopt a morphological filtering algorithm to further refine the map boundaries and make the geographical elements clearer.

[0095] 4) Output structured data

[0096] After being processed by OCRNet, the system extracts the main areas of the map and generates a structured data format for subsequent processing, such as map classification, feature extraction, and visual style analysis.

[0097] Through deep learning methods, the degree of automation is high, and large-scale historical map data can be efficiently processed.

[0098] Semantic segmentation can accurately extract the key content of the map and improve the accuracy of map analysis.

[0099] It is applicable to historical maps of different types and ages, improving the standardization ability of map data.

[0100] There are significant differences in the drawing styles and information expression methods of historical maps. For example, maps of different periods are different in terms of line density, annotation fonts, color patterns, etc. Traditional manual classification methods are inefficient and difficult to accurately distinguish different map series. Therefore, the present invention adopts a genetic algorithm (GA) to optimize the feature selection of maps and achieve automatic classification.

[0101] Specific working principle:

[0102] 1) Feature extraction

[0103] The map is divided into basic units of 50×50 pixels, and 13 candidate visual features are extracted from each unit, including:

[0104] Texture features (Gabor filtering, LBP local binary pattern).

[0105] Morphological features (boundary density, direction histogram).

[0106] Color features (gray histogram, HSV color distribution).

[0107] These feature data are used as input for subsequent classification optimization.

[0108] 2) Genetic algorithm for optimizing the feature space

[0109] Initializing the population: Randomly select different feature combinations as the initial feature set, and each individual represents a different feature selection scheme.

[0110] Fitness function calculation:

[0111] Calculate the classification accuracy of each feature combination, perform clustering analysis on the map using a classifier (such as K-Means or SVM), and evaluate the classification performance.

[0112] Genetic operations (crossover and mutation):

[0113] Combine different features through crossover operations to explore better feature combinations.

[0114] Select different feature subsets through mutation operations to enhance the diversity of the search and prevent getting stuck in local optima.

[0115] Select the optimal feature set: After iterative calculations, screen out the 6D optimal feature space that can best distinguish different map types and use it for map classification.

[0116] 3) Map classification based on optimized features

[0117] Use K-Means clustering or hierarchical clustering to classify the optimized 6D feature space and generate different mapotypes (map types).

[0118] Each mapotype represents a map category with similar visual features, such as "hand-drawn map", "modern printed map", etc.

[0119] 4) Output classification results

[0120] The system automatically classifies maps into different categories and can be used for the analysis of visual style evolution.

[0121] Technological progress and advantages:

[0122] Automatically optimize feature selection, reduce manual intervention, and improve classification accuracy.

[0123] A classification method based on machine learning can process historical maps of different styles on a large scale.

[0124] Improve the structured management ability of map data, facilitating the archiving, querying, and analysis of historical maps.

[0125] These two working principles respectively solve the problems of content extraction and automatic classification optimization of historical maps:

[0126] 1) OCRNet semantic segmentation can automatically remove noise from historical maps and extract core geographical content, making subsequent analysis more accurate.

[0127] 2) The feature classification method optimized by genetic algorithms makes map classification more intelligent, reduces human errors, and improves the usability of historical map data.

[0128] The combination of these two methods enables the automatic processing device for historical map visual information of the present invention to more efficiently digitize, classify, and analyze historical maps, providing important technical support for map research, archaeology, and historical geography.

[0129] It should be noted that the embodiments of the present invention can be implemented through hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated designed hardware. Those of ordinary skill in the art can understand that the above devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code is provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuits of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above hardware circuits and software, such as firmware.

[0130] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be covered by the protection scope of the present invention.

Claims

1. An automatic processing device for visual information of historical maps, characterized in that: include: A map image preprocessing module, connected to the map fragmentation module, is used to digitize the historical map and remove background noise; Use OCRNet neural network to perform semantic segmentation on the map and extract geographic content areas; The map fragmentation module is connected to the feature extraction and optimization module and is used to segment the map image into basic units of 50×50 pixels, each of which represents a local visual feature of the map; The feature extraction and optimization module is connected to the map type generation module to extract 13 candidate visual features, optimize the feature space using a genetic algorithm, select the feature combination that can best distinguish different map series, and construct a 6-dimensional feature space; The map type generation module is connected to the visual evolution analysis module and is used to generate mapotypes by iteratively dividing the feature space. Each mapotype represents a class of mapels with similar visual features. The visual evolution analysis module, connected with the map type generation module, is used to analyze the evolution of visual styles of maps of different periods and scales based on the distribution of mapotypes.

2. The automatic processing device for historical map visual information according to claim 1, characterized in that: The map image preprocessing module: A collection of over 10,000 historical maps from multiple libraries and archives.

3. The automatic processing device for historical map visual information according to claim 1, characterized in that: The map fragmentation module: By random sampling and repositioning, we ensure that each mapel covers a different area of ​​the map and excludes areas with low information content.

4. The automatic processing device for historical map visual information according to claim 1, characterized in that: The candidate visual features include color distribution, morphology, texture, graphic load, line width and direction.

5. The automatic processing device for historical map visual information according to claim 1, characterized in that: The map type generation module: The t-SNE algorithm is used to reduce the high-dimensional feature space to 2D for easy visualization analysis.

6. The automatic processing device for historical map visual information according to claim 1, characterized in that: The visual evolution analysis module: Quantify the long-term trend of map style changes through statistical methods.

7. A method for automatically processing historical map information using the device for automatically processing historical map visual information as claimed in any one of claims 1 to 6, characterized in that: The method for automatically processing historical map information comprises: Step 1: Use the map image preprocessing module to digitize the historical map and remove background noise; use the OCRNet neural network to perform semantic segmentation on the map and extract the geographic content area; Step 2: Segment the map image into basic units of 50×50 pixels through the map fragmentation module, where each unit represents a local visual feature of the map; Step 3: 13 candidate visual features are extracted through the feature extraction and optimization module, the feature space is optimized using a genetic algorithm, the feature combination that can best distinguish different map series is selected, and a 6-dimensional feature space is constructed; Step 4, the map type generation module uses iterative partitioning of the feature space to generate mapotypes, each mapotype represents a class of mapels with similar visual features; Step 5: Analyze the evolution of visual styles of maps of different periods and scales based on the distribution of mapotypes through the visual evolution analysis module.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for automatic processing of historical map information as claimed in claim 7.

9. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method for automatically processing historical map information as claimed in claim 7.

10. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the automatic processing device of historical map visual information as described in any one of claims 1-6.

Images