Artwork reproduction feature point recognition system and method based on AI technology

Abstract

The present invention relates to the field of artwork restoration technology, and specifically discloses a system and method for identifying feature points of artwork reproduction based on AI technology, wherein the method comprises: obtaining a plurality of artwork images that have been feature-annotated by AI automatic annotation technology to form a feature model; forming a plurality of corresponding sets of recognizable feature points in a coded definition manner according to all feature information possessed by each feature model; forming a feature table with the sets of recognizable feature points, calling the feature table based on demand information generated in an artwork image recognition process, the feature table being configured to provide a corresponding set of recognizable feature points with the demand information; forming a feature recognition network with the sets of recognizable feature points, constructing one or more optional paths for the feature recognition network, and implementing the calling and applicable comparison of feature information through the optional paths, thereby realizing the recognition of feature points that can be reproduced in the artwork image.

Description

Technical Field

[0001] The present invention relates to the technical field of artwork restoration, and in particular to an AI-based feature point recognition system and method for artwork reproduction. Background Art

[0002] Traditionally, artwork restoration is done by human experts based on their experience. In practical teaching of artworks, it is often only possible to explain them through some mature cases. Due to the diverse forms and expressions of artworks, it is difficult to fully explain different styles and factions. Therefore, repeated practical work is required to complete the training of students. However, artworks are important physical assets with certain economic and research values. Whether or not to perfectly repair and reproduce incomplete and damaged artworks requires very professional technical means.

[0003] Some existing technologies use 3D scanning and 3D reconstruction methods to repair artworks, but these technologies can only repair artworks that are slightly damaged or have few missing parts, and it is difficult to form practical cases with teaching characteristics during the restoration process. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a system and method for identifying feature points of artwork reproduction based on AI technology.

[0005] The main purpose of this application is to provide an AI-based feature point recognition system for artwork reproduction, including: a feature model construction module for acquiring multiple artwork images that have been feature-annotated by AI automatic annotation technology to form a feature model;

[0006] A feature point set forming module, used for forming a plurality of corresponding replica feature point sets that can be identified according to all feature information possessed by each feature model in a coded definition manner;

[0007] A feature table forming module, used to form a feature table with the set of replicated feature points, and embed the feature table into a processor;

[0008] Input module, obtains any artwork image and inputs it into the processor;

[0009] The processor is provided with:

[0010] A logic program, the logic program is used to call the feature table based on the demand information generated by the artwork image recognition process, the feature table is configured to provide a corresponding set of replication feature points based on the demand information;

[0011] A feature recognition network generator, which forms a feature recognition network with the set of replicated feature points, wherein each feature information in the set of replicated feature points is configured as a node in the feature recognition network;

[0012] The configurator randomly constructs one or more optional paths for the feature recognition network in response to the demand information through the configurator, and realizes the calling and applicable comparison of feature information through the optional paths, thereby realizing the identification of feature points that can be reproduced in the artwork image.

[0013] Furthermore, the AI ​​automatic annotation technology is formed according to the following method:

[0014] Get a large number of artwork images;

[0015] Using human experts to segment and annotate the feature information in the artwork image to form an annotated image;

[0016] Extracting the annotated image and feature information corresponding to the annotated image to form an annotated sample;

[0017] Classifying the labeled samples based on the expression approximation of the feature information as a basis for semantic extraction, and building a feature information semantic library according to the classification;

[0018] By constructing control logic and semantic logic, linking the semantic logic to the feature information semantic library, and linking the semantic logic to the control logic, an AI automatic labeling program is formed;

[0019] The AI ​​automatic labeling technology is formed by the AI ​​automatic labeling program.

[0020] Furthermore, the coding definition includes:

[0021] Setting a unique identification code I for each of the characteristic information;

[0022] Identification codes with similar feature information have the same encoding expression rules II;

[0023] Feature information with the same encoding expression rules are linked to each other through expressions III.

[0024] Furthermore, the expression is a regular expression formed by coding expression.

[0025] Furthermore, a register is provided in the processor, and the feature table is stored in the register.

[0026] The present invention also provides a method for identifying feature points of artwork reproduction based on AI technology, which uses the above-mentioned system for identifying feature points of artwork reproduction based on AI technology to perform identification, including the following steps: acquiring a plurality of artwork images that have been feature-annotated by AI automatic annotation technology to form a feature model;

[0027] Forming a plurality of corresponding sets of identifiable replicated feature points according to all the feature information of each feature model in a coded definition manner;

[0028] forming a feature table with the set of replicated feature points, and embedding the feature table into a processor;

[0029] Obtain any artwork image and input it into the processor, and call the feature table based on the requirement information generated by the artwork image recognition process, wherein the feature table is configured to provide a corresponding set of replicated feature points based on the requirement information;

[0030] The replicated feature point set is used to form a feature recognition network, wherein each feature information in the replicated feature point set is configured as a node in the feature recognition network;

[0031] Through the configurator and in response to the demand information, one or more optional paths are randomly constructed for the feature recognition network, and the feature information is called and applicable compared through the optional paths, thereby realizing the recognition of the feature points that can be reproduced in the artwork image.

[0032] This application uses AI automatic labeling technology to accurately identify all the characteristic information on the artwork. Therefore, it is possible to identify the characteristic information around the missing or damaged parts of the artwork, and use the characteristic information of the surrounding areas to infer the characteristic information of the missing or damaged parts. In this way, the corresponding standard samples can be derived from the constructed characteristic information semantic library, and the corresponding reproduction and repair can be performed using the standard samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is the system framework principle diagram of the present invention;

[0034] Figure 2 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] Example 1

[0037] The present invention provides an AI-based feature point recognition system for artwork reproduction, comprising:

[0038] A feature model building module is used to obtain multiple artwork images that have been feature-annotated using AI automatic annotation technology to form a feature model;

[0039] A feature point set forming module, used for forming a plurality of corresponding replica feature point sets that can be identified according to all feature information possessed by each feature model in a coded definition manner;

[0040] A feature table forming module, used to form a feature table with the set of replicated feature points, and embed the feature table into a processor;

[0041] Input module, obtains any artwork image and inputs it into the processor;

[0042] The processor is provided with:

[0043] A logic program, the logic program is used to call the feature table based on the demand information generated by the artwork image recognition process, the feature table is configured to provide a corresponding set of replication feature points based on the demand information;

[0044] A feature recognition network generator, which forms a feature recognition network with the set of replicated feature points, wherein each feature information in the set of replicated feature points is configured as a node in the feature recognition network;

[0045] The configurator randomly constructs one or more optional paths for the feature recognition network in response to the demand information through the configurator, and realizes the calling and applicable comparison of feature information through the optional paths, thereby realizing the identification of feature points that can be reproduced in the artwork image.

[0046] In the above, the AI ​​automatic annotation technology is formed according to the following method:

[0047] Get a large number of artwork images;

[0048] Using human experts to segment and annotate the feature information in the artwork image to form an annotated image;

[0049] Extracting the annotated image and feature information corresponding to the annotated image to form an annotated sample;

[0050] Classifying the labeled samples based on the expression approximation of the feature information as a basis for semantic extraction, and building a feature information semantic library according to the classification;

[0051] By constructing control logic and semantic logic, linking the semantic logic to the feature information semantic library, and linking the semantic logic to the control logic, an AI automatic labeling program is formed;

[0052] The AI ​​automatic labeling technology is formed by the AI ​​automatic labeling program.

[0053] This application applies AI automatic annotation technology to artwork recognition by constructing AI automatic annotation technology, which has many advantages. It can scan and annotate a large number of artworks, and can be used as a resource for scanning and annotation whether they need to be repaired or not; the obtained resource library can be used for the repair, reproduction and restoration of artworks, and has a very wide range of uses; during the restoration process, guided restoration can be carried out, as well as replica restoration; through the annotation of artificial experts, the annotation of artificial experts can be used as the main direction of teaching, allowing students to understand and learn the annotation methods of different features, so as to deepen their understanding of the expression of different features in artworks.

[0054] In the above, the AI ​​automatic annotation technology for feature annotation includes the following methods: under the control direction of the control logic, when the artwork image is recognized, the feature information in the artwork image is automatically extracted, and the feature information is semantically extracted under the semantic logic; based on the semantic information obtained by the semantic extraction, the annotated sample is obtained from the feature information semantic library, and feature annotation is performed through the annotated sample.

[0055] In the above, the coding definition includes: setting a unique identification code for each of the feature information I; identification codes with similar feature information have the same coding expression rules II; feature information with the same coding expression rules are linked to each other through expressions III.

[0056] Furthermore, the expression is a regular expression formed by coding expression.

[0057] Furthermore, a register is provided in the processor, and the feature table is stored in the register.

[0058] Furthermore, the logic program is an identification control program arranged in a processor with the processor as a carrier, and includes:

[0059] Recognition programs for identifying characteristic information contained in artwork images;

[0060] An extraction program for segmenting and extracting image blocks corresponding to feature information;

[0061] an analysis program for semantic analysis of feature information;

[0062] A loader that loads the feature table.

[0063] Further, semantic analysis is performed on the feature information to obtain demand information;

[0064] The demand information refers to the semantic information obtained by semantically analyzing the feature information.

[0065] Furthermore, the analysis program includes an AI automatic labeling program.

[0066] Furthermore, the method of providing a corresponding set of replicated feature points based on the demand information is:

[0067] Obtaining the labeled sample from a feature information semantic library according to the semantic information;

[0068] By labeling samples, feature points are obtained by labeling the extracted image blocks;

[0069] Obtain identification codes and coding expression rules in feature points;

[0070] Group image blocks with the same encoding expression rules into the same list;

[0071] The set of replicated feature points is composed of at least one list.

[0072] Furthermore, the method of forming a feature recognition network using the set of replicated feature points is as follows:

[0073] Get all lists of copied feature point sets;

[0074] All lists are expanded in parallel, and the feature points corresponding to the image blocks under the lists are expanded at the same time, and then the feature points under the same list are associated to obtain a parallel feature recognition network.

[0075] Example 2

[0076] The present invention also provides a method for identifying feature points of artwork reproduction based on AI technology, which uses the above-mentioned system for identifying feature points of artwork reproduction based on AI technology to perform identification, including the following steps: acquiring a plurality of artwork images that have been feature-annotated by AI automatic annotation technology to form a feature model;

[0077] Forming a plurality of corresponding sets of identifiable replicated feature points according to all the feature information of each feature model in a coded definition manner;

[0078] forming a feature table with the set of replicated feature points, and embedding the feature table into a processor;

[0079] Obtain any artwork image and input it into the processor, and call the feature table based on the requirement information generated by the artwork image recognition process, wherein the feature table is configured to provide a corresponding set of replicated feature points based on the requirement information;

[0080] The replicated feature point set is used to form a feature recognition network, wherein each feature information in the replicated feature point set is configured as a node in the feature recognition network;

[0081] Through the configurator and in response to the demand information, one or more optional paths are randomly constructed for the feature recognition network, and the feature information is called and applicable compared through the optional paths, thereby realizing the recognition of the feature points that can be reproduced in the artwork image.

[0082] In the above, the AI ​​automatic annotation technology is formed according to the following method:

[0083] Get a large number of artwork images;

[0084] Using human experts to segment and annotate the feature information in the artwork image to form an annotated image;

[0085] Extracting the annotated image and feature information corresponding to the annotated image to form an annotated sample;

[0086] Classifying the labeled samples based on the expression approximation of the feature information as a basis for semantic extraction, and building a feature information semantic library according to the classification;

[0087] By constructing control logic and semantic logic, linking the semantic logic to the feature information semantic library, and linking the semantic logic to the control logic, an AI automatic labeling program is formed;

[0088] The AI ​​automatic labeling technology is formed by the AI ​​automatic labeling program.

[0089] Furthermore, the method of providing a corresponding set of replicated feature points based on the demand information is:

[0090] Obtaining the labeled sample from a feature information semantic library according to the semantic information;

[0091] By labeling samples, feature points are obtained by labeling the extracted image blocks;

[0092] Obtain identification codes and coding expression rules in feature points;

[0093] Group image blocks with the same coding expression rules into the same list;

[0094] The set of replicated feature points is composed of at least one list.

[0095] Furthermore, the method of forming a feature recognition network using the set of replicated feature points is as follows:

[0096] Get all lists of copied feature point sets;

[0097] All lists are expanded in parallel, and the feature points corresponding to the image blocks under the lists are expanded at the same time, and then the feature points under the same list are associated to obtain a parallel feature recognition network.

[0098] This application uses AI automatic labeling technology to accurately identify all the characteristic information on the artwork. Therefore, it is possible to identify the characteristic information around the missing or damaged parts of the artwork, and use the characteristic information of the surrounding areas to infer the characteristic information of the missing or damaged parts. In this way, the corresponding standard samples can be derived from the constructed characteristic information semantic library, and the corresponding reproduction and repair can be performed using the standard samples.

[0099] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. The AI-based feature point recognition system for artwork reproduction is characterized by: include: A feature model building module is used to obtain multiple artwork images that have been feature-annotated using AI automatic annotation technology to form a feature model; A feature point set forming module, used for forming a plurality of corresponding replica feature point sets that can be identified according to all feature information possessed by each feature model in a coded definition manner; A feature table forming module, used to form a feature table with the set of replicated feature points, and embed the feature table into a processor; Input module, obtains any artwork image and inputs it into the processor; The processor is provided with: A logic program, the logic program is used to call the feature table based on the demand information generated by the artwork image recognition process, the feature table is configured to provide a corresponding set of replication feature points based on the demand information; A feature recognition network generator, which forms a feature recognition network with the set of replicated feature points, wherein each feature information in the set of replicated feature points is configured as a node in the feature recognition network; The configurator randomly constructs one or more optional paths for the feature recognition network in response to the demand information through the configurator, and realizes the calling and applicable comparison of feature information through the optional paths, thereby realizing the identification of feature points that can be reproduced in the artwork image.

2. The AI-based feature point recognition system for artwork reproduction according to claim 1 is characterized in that: The AI ​​automatic annotation technology is formed according to the following method: Get a large number of artwork images; Using human experts to segment and annotate the feature information in the artwork image to form an annotated image; Extracting the annotated image and feature information corresponding to the annotated image to form an annotated sample; Classifying the labeled samples based on the expression approximation of the feature information as a basis for semantic extraction, and building a feature information semantic library according to the classification; By constructing control logic and semantic logic, linking the semantic logic to the feature information semantic library, and linking the semantic logic to the control logic, an AI automatic labeling program is formed; The AI ​​automatic labeling technology is formed by the AI ​​automatic labeling program.

3. The AI-based feature point recognition system for artwork reproduction according to claim 1 is characterized in that: The coding definition includes: Setting a unique identification code I for each of the characteristic information; Identification codes with similar feature information have the same encoding expression rules II; Feature information with the same encoding expression rules are linked to each other through expressions III.

4. The AI-based feature point recognition system for artwork reproduction according to claim 3 is characterized in that: The expression is a regular expression formed by coding expression.

5. The AI-based feature point recognition system for artwork reproduction according to claim 1 is characterized in that: The processor is provided with a register, and the feature table is stored in the register.

6. A method for identifying feature points of artwork reproduction based on AI technology, using the system for identifying feature points of artwork reproduction based on AI technology as claimed in any one of claims 1 to 5 for identification, characterized in that: Acquire multiple artwork images that have been feature-annotated using AI automatic annotation technology to form a feature model; Forming a plurality of corresponding sets of identifiable replicated feature points according to all the feature information of each feature model in a coded definition manner; forming a feature table with the set of replicated feature points, and embedding the feature table into a processor; Obtain any artwork image and input it into the processor, and call the feature table based on the requirement information generated by the artwork image recognition process, wherein the feature table is configured to provide a corresponding set of replicated feature points based on the requirement information; The replicated feature point set is used to form a feature recognition network, wherein each feature information in the replicated feature point set is configured as a node in the feature recognition network; Through the configurator and in response to the demand information, one or more optional paths are randomly constructed for the feature recognition network, and the feature information is called and applicable compared through the optional paths, thereby realizing the recognition of the feature points that can be reproduced in the artwork image.

Images